Epolones induce erythropoietin expression via hypoxia-inducible factor-1α activation

The glycoprotein hormone erythropoietin (Epo), produced by the embryonic liver and the adult kidney, is the main stimulator of erythropoiesis.1 Recombinant Epo is widely used to treat patients suffering from anemia. However, recombinant Epo is expensive and must be administered intravenously or subcutaneously. Thus, an orally active, small molecular weight compound that induces endogenous Epo production would be an alternative for the treatment of anemia not caused by deficient renal Epo production. Therefore, fungal products were screened for their ability to induce a reporter gene under the control of 6-kilobase (kb) 5′ and 0.3-kb 3′ flanking sequences derived from the Epo gene.2-4 Several compounds (the 3 sesquiterpene tropolones: pycnidione, epolone A and epolone B, and 8-methyl-pyridoxatin) were isolated, which induced reporter gene expression as well as Epo production in human Hep3B hepatoma cells.2,3 In addition, the related commercially available pyridone, ciclopirox olamine (CPX), was shown to have a similar function.2 Hereinafter, we refer to this class of substances, capable of inducing Epo expression, as “epolones.” The mechanism of Epo induction via epolones and the involvedcis-regulatory elements, however, remained unidentified.

The most powerful inducer of Epo expression is hypoxia (a deficiency in oxygen supply not matching oxygen consumption). The cellular oxygen concentration is measured by a putative heme oxygen sensor.5 Hypoxia results in the activation of hypoxia-inducible factor (HIF)-1, a heterodimeric transcription factor containing an oxygen-labile α subunit and a constitutively expressed β subunit previously known as aryl hydrocarbon receptor nuclear translocator (ARNT).6 Upon exposure to hypoxia, HIF-1α becomes protected from proteolytic degradation, translocates into the nucleus, and heterodimerizes with ARNT to form the HIF-1 complex.7-10 HIF-1 then binds to its DNA consensus binding site (HBS) present within cis-regulatory hypoxia-response elements (HREs) of oxygen-regulated genes.9 HREs confer enhanced transcription of oxygen-regulated genes by recruiting various other transcription factors in addition to HIF-1 along with transcriptional coactivators such as CBP/p300.7 These “oxygen(es)”9 include Epo, vascular endothelial growth factor (VEGF), glycolytic enzymes, and glucose transporter as well as transferrin and transferrin receptor. Apart from hypoxia, the transition cations Co²⁺ and Ni²⁺ as well as the iron chelator deferoxamine mesylate (DFX) are capable of inducing HIF-1α protein stability and hypoxia-dependent gene expression, probably by replacing or removing the central iron of the putative heme oxygen sensor.5,7,11 

In this work, we established an HIF-1–dependent hypoxia-reporter cell line, investigated the effects of epolones on HIF-1 activation, and unraveled the mechanism by which epolones activate Epo expression under normoxic conditions.

DFX, 2,2′-dipyridyl, CPX, neocuproine, ferrozine, luminol, and coumaric acid were purchased from Sigma (Buchs, Switzerland) and ferrous ethylenediammonium sulfate from Fluka (Buchs, Switzerland). Pycnidione and 8-methyl-pyridoxatin were isolated as described previously.2,3 Stock solutions (50 mmol/L) of the epolones (CPX, pycnidione, and 8-methyl-pyridoxatin) were prepared in methanol and stored at −30°C. Immediately before use, they were diluted in water to a working concentration of 1 mmol/L.

Chinese hamster ovary (CHO) cells were a kind gift of Peter J. Nielsen (Freiburg, Germany), and human hepatoma HepG2 cells were purchased from the American Type Culture Collection (HB-8065). Both cell lines were cultured in Dulbecco's modified Eagle's medium (DMEM; high glucose, Life Technologies) supplemented with 10% heat-inactivated fetal calf serum (FCS; Boehringer-Mannheim, Basel, Switzerland), 100-U/mL penicillin, 100-μg/mL streptomycin, 1 × nonessential amino acids, and 1-mmol/L Na-pyruvate (all purchased from Life Technologies, Basel, Switzerland) in a humidified atmosphere containing 5% CO₂ at 37°C. Oxygen partial pressures in the incubator (Forma Scientific, Illkirch, France) were either 140 mmHg (20% O₂ vol/vol, normoxia) or 7 mmHg (1% O₂ vol/vol, hypoxia). For transfection, 0.5 × 10⁶ cells in 350-μL medium without FCS were mixed with 50 μg of DNA in 50 μL of 10-mmol/L Tris-HCl (pH 7.4), and 1-mmol/L ethylenediaminetetraacetic acid and electroporated at 250 V and 960 μF (GenePulser, Bio-Rad, Glattbrugg, Switzerland).

A firefly luciferase reporter gene plasmid (pH3SVL) containing a total of 6 HBSs derived from the transferrin HRE12 was constructed by inserting 2 copies of the oligonucleotide TfHBSww into the SmaI site of the plasmid pGLTfHBSww.12 For stable transfection, pH3SVL was linearized with XmnI, mixed with the EcoRI-linearized neomycin expression vector pSV2neo at a molar ratio of 100:1, and coelectroporated into CHO cells. Following limited dilution and selection in 2-mg/mL G418 (Alexis, Läufelfingen, Switzerland), a hypoxia-reporter cell line (termed HRCHO5) was chosen based on the efficiency of hypoxic reporter gene induction. For transient transfection assays, pGLHIF1.3 containing 3 copies of the HBS derived from the Epo 3′ HRE13 was coelectroporated into HepG2 cells together with the β-galactosidase reference vector pCMVlacZ.12 The cells were split and incubated for 43 hours under normoxic or hypoxic conditions. Following stimulation, stably transfected HRCHO5 cells and transiently transfected HepG2 cells were lysed in reporter lysis buffer (Promega), and luciferase and β-galactosidase activities were determined according to the manufacturer's instructions (Promega, Catalys, Wallisellen, Switzerland) using a Lumat LB9501 luminometer (EG&G Berthold, Regensdorf, Switzerland) and a DigiScan 96-well plate photometer (ASYS, BioBlock, Illkirch, France), respectively. Differences in the transfection efficiency and extract preparation were corrected by normalization to the corresponding protein contents (Bradford assay, Bio-Rad) or β-galactosidase activities.

Following stimulation, cells were harvested and nuclear extracts were prepared as described previously,13 except that the cells were lysed with 0.02% Nonidet P-40 instead of dounce homogenization. For immunoblot assays, aliquots (30 μg) of nuclear extracts were fractionated by 7.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and electrotransferred to a nitrocellulose membrane (Schleicher & Schuell, Riehen, Switzerland). Staining with PonceauS (Sigma) confirmed equal loading and blotting efficiency. After blocking nonspecific binding sites with 4% defatted milk powder in phosphate-buffered saline (PBS), the blot was probed with the affinity-purified anti–HIF-1α monoclonal antibody mgc3 described previously14 (Affinity BioReagents, Lausen, Switzerland), followed by a horseradish peroxidase–coupled secondary goat antimouse antibody (Pierce, Socochim, Lausanne, Switzerland). Chemiluminescence detection was performed by incubation of the membrane with 100-mmol/L Tris-HCl (pH 8.5), 2.65-mmol/L H₂O₂, 0.45-mmol/L luminol, and 0.625-mmol/L coumaric acid for 1 minute, followed by exposure to x-ray films (SuperRX, Fuji, Dielsdorf, Switzerland). For immunofluorescence analysis, adherent cells were fixed with freshly prepared 4% paraformaldehyde in PBS (pH 7.4) for 10 minutes, washed with PBS, permeabilized with 0.5% Triton X-100 for 5 minutes, and rinsed again with PBS. After blocking nonspecific binding sites with 10% FCS in PBS for 30 minutes, the cells were incubated overnight with the anti–HIF-1α antibody mgc3 diluted 1:10 with 3% BSA in PBS, followed by a fluorescein isothiocyanate–coupled secondary donkey antimouse antibody diluted 1:100 with 3% BSA in PBS (Jackson, Milan Analytica, La Roche, Switzerland). After extensive washings in PBS and mounting in Mowiol (Calbiochem, Stehelin, Basel, Switzerland), the cells were analyzed by fluorescence microscopy.

Electrophoretic mobility shift assays (EMSAs) were carried out as described previously.15 A double-stranded, HBS-containing oligonucleotide derived from the Epo 3′ enhancer was used as probe. For supershift analysis, the anti–HIF-1α monoclonal antibody mgc3 was added to the completed DNA-protein binding reaction mixture and incubated for 16 hours at 4°C prior to loading.

Iron concentrations were determined by a colorimetric assay according to Fish.16 Briefly, iron was released from 100-μL samples by treatment with 50 μL of 142-mmol/L KMnO₄ and 600-mmol/L HCl at 60°C for 2 hours. Thereafter, 10 μL of 5-mol/L ammonium acetate, 2-mol/L ascorbic acid, 13.1-mmol/L neocuproine, and 6.5-mmol/L ferrozine were added and incubated at room temperature for 30 minutes. Iron was determined by measuring the absorption at 562 nm in a DU-65 spectrophotometer (Beckmann). Iron standards were prepared by dissolving ferrous ethylenediammonium sulfate in 10-mmol/L HCl. The detection limit of this assay was 2 ng of iron, and the standard curve was linear up to 800 ng of iron. It has been validated by demonstrating a 2-fold molar ratio of iron to protein in 10-μg samples of iron-saturated purified transferrin (Life Technologies).

To facilitate the screening for novel agonists/antagonists of the oxygen-regulated signaling pathway, CHO cells were stably transfected with the hypoxia-dependent reporter gene shown in Figure1A. This construct contains the firefly luciferase cDNA under the control of the SV40 promoter and 3 HREs derived from the hypoxia-responsive transferrin 5′ enhancer.12 Of note, the transferrin HRE contains 2 HBSs and the whole construct, hence, a total of 6 HBSs. One CHO clone (designated HRCHO5) was selected based on its high hypoxic inducibility of luciferase activity. Exposure of HRCHO5 cells to hypoxic conditions (1% oxygen) for 18 hours led to a 21.2 ± 4.2-fold induction (mean ± SD, n = 9) of luciferase activity compared with normoxic (20% oxygen) control cells (Figure 1B). The transition cations Co²⁺ and Ni²⁺ dose-dependently induced luciferase activity under normoxic conditions, approximately reaching the hypoxic values at concentrations of 50 μmol/L (CoCl₂) and 100 μmol/L (NiCl₂), respectively. Luciferase activity dropped again at higher concentrations, presumably because these agents are cytotoxic. Similar results were obtained with the iron chelator DFX, which maximally induced normoxic luciferase activity at concentrations of 50 to 200 μmol/L (Figure 1B, top graph). Under hypoxic conditions, CoCl₂ and DFX additionally induced reporter gene activity with a similar dosage dependence as found under normoxic conditions. In contrast, NiCl₂ had no additional effects under hypoxic conditions (Figure 1B, bottom graph).

Epolones have previously been identified as a family of fungal products capable of inducing Epo expression under normoxic conditions.2,3 Because the reporter gene used for Epo-inducing drug screening contained the 3′ HRE derived from the Epo gene, we reasoned that epolones could activate Epo transcription via an HBS, which is also present in the hypoxia-reporter cell line HRCHO5. Indeed, treatment of HRCHO5 cells with the epolones CPX, pycnidione, and 8-methyl-pyridoxatin dose-dependently induced luciferase expression under normoxic conditions (Figure2, top graph). While the induction was maximal at epolone concentrations of 4 to 8 μmol/L (16 μmol/L for CPX), luciferase activity decreased again at higher doses. Hypoxic exposure of HRCHO5 cells (Figure 2, bottom graph) had additive effects, but the decrease in reporter gene activity occurred already at lower epolone concentrations than under normoxic conditions.

We next analyzed whether epolones activated luciferase expression specifically via the HRE or whether other cis-regulatory elements were involved, as could be expected from the fact that hypoxia and epolones had additive effects. Therefore, luciferase reporter gene constructs containing 3 copies of the Epo 3′ HBS, either wild type (pGLHIF1.3) or mutant (pGLHIF1mt.3),13 were transiently transfected into HepG2 cells, which were split and exposed to hypoxia and/or CPX. As shown in Figure 3, CPX increased reporter gene activity also in this reporter gene–cell line combination under both normoxic and hypoxic conditions. However, the mutant HBSs conferred neither CPX nor hypoxic induction of reporter gene expression, suggesting that both stimuli activate reporter gene expression via similar mechanisms.

The binding of HIF-1 to an HBS is obligatory for the activation of an HRE and subsequent gene expression. The sequential steps in the formation of the HIF-1 complex include the stabilization of its α subunit followed by nuclear translocation and heterodimerization with ARNT to form a functional DNA-binding transcription factor complex. We thus investigated whether this pathway could be mimicked by epolones.

First, HIF-1α protein levels were determined by immunoblotting in HepG2 cells. As shown in the left part of Figure4, all 3 epolones efficiently induced HIF-1α protein expression in normoxic HepG2 cells to the same extent as CoCl₂. Hypoxic conditions (Figure 4, right part) also induced HIF-1α protein, which could be further enhanced by adding 8-methyl-pyridoxatin or CPX but not pycnidione or CoCl₂. Next, as shown by immunofluorescence, hypoxia as well as the epolone CPX induced nuclear accumulation of HIF-1α in HepG2 cells (Figure5). Finally, as shown by EMSA, CPX also activated DNA binding of HIF-1 to an oligonucleotide probe containing an HBS derived from the Epo HRE (Figure6). This induction was slightly more pronounced than with hypoxia or DFX or the combination of hypoxia with CPX. Supershift analysis using the monoclonal anti–HIF-1α antibody mgc314 confirmed that (most of) the HIF complexes present in HepG2 nuclear extracts contained the HIF-1α subunit.

Having established that epolone-dependent reporter gene activation followed HIF-1α induction, we addressed the question of how epolones might induce HIF-1α. Based on functional (Figures 1 and 2) and structural similarities between DFX and CPX, we followed the hypothesis that CPX also might act as an iron chelator. Feroxamine contains 1 atom of ferrous iron in the center of a hexadentate cluster formed by 3 hydroxamic acid groups.17 The epolone CPX contains a single hydroxamic acid group, suggesting a bidentate structure theoretically requiring 3 mol of CPX to chelate 1 mol of iron. We thus titered the concentration of iron necessary to block DFX- and CPX-induced reporter gene activation in HRCHO5 cells. As shown in Figure 7A, efficient inhibition of luciferase activity under normoxic and hypoxic conditions occurred at an iron:DFX molar ratio of 1:1, consistent with the ability of 1 molecule of DFX to stoichiometrically chelate 1 molecule of iron. Interestingly, CPX-induced luciferase activity was abolished at an iron:CPX ratio of 1:2 but not at 1:4, supporting the idea that CPX functions as an iron chelator with the expected stoichiometry of 3 mol of CPX per 1 mol of iron.

As shown in Figure 7B, the iron chelation–dependent induction of reporter gene activity in HRCHO5 cells can also be blocked by addition of AlCl₃. However, while inhibition of DFX activity again required stoichiometric concentrations of AlCl₃, the activity of CPX was already reduced by approximately 50% at a molar aluminium:CPX ratio of 1:8. This cannot be attributed to a higher toxicity of AlCl₃ compared with FeCl₂ because the actual AlCl₃ concentration inhibiting CPX function was only 1 μmol/L, whereas 50-μmol/L AlCl₃ had no significant effect on DFX-induced reporter gene activity. Thus, these data suggest that CPX-mediated activation of luciferase expression in HRCHO5 cells might be due to metal chelation.

To exclude the possibility of an inhibitory function of iron downstream of HIF-1α induction, we analyzed HIF-1α protein expression directly by immunoblotting. As shown in Figure 8, addition of iron inhibited HIF-1α induction by CPX as well as by the established activators DFX11 and 2,2′-dipyridyl.18 

The iron chelation stoichiometry suggested by the titration experiments shown in Figure 7 might be compromised by the (unknown) concentration of iron in the cell culture medium. To confirm our data, the iron saturation curves of DFX and CPX were determined by spectrophotometry. As shown in Figure 9A, DFX and CPX show iron-dependent (but not aluminium-dependent) absorption maxima at 430 nm19 and 421 nm, respectively, allowing the direct estimation of the iron chelation stoichiometry. Whereas DFX iron saturation was found to be completed at a 1:1 molar ratio, CPX was saturated with iron at a molar ratio of 1:3 (Figure 9B), thus confirming the functional results obtained in the HRCHO5 cell line.

In this context, it would be important to know the iron concentration in the cell culture medium. Using a colorimetric assay with a lower iron detection limit of 0.4 μmol/L (see “Materials and methods”), we determined an iron concentration in the FCS of 148 ± 19 μmol/L, whereas the iron concentration in the DMEM itself was 1.4 ± 0.35 μmol/L and, hence, close to the detection limit. Therefore, the medium containing 10% FCS contains 16-μmol/L iron. In conclusion, iron in the cell culture medium can be completely chelated with DFX but not CPX at their respective optimal concentrations, especially considering the fact that 3 mol of CPX are necessary to chelate 1 mol of iron, whereas 1 mol of DFX is sufficient for the same purpose.

In this study, we describe the construction of the HRCHO5 hypoxia-reporter cell line, which contains a stably integrated hypoxia-responsive luciferase reporter gene under the control of 6 HBSs. This allows easy and rapid monitoring of the activity of HIF-1. However, it cannot be formally excluded that one of the other HIF α family members, HIF-2α20-23 or HIF-3α,24might also be involved in the activation of reporter gene expression, because they display features very similar to HIF-1α.25,26 Originally thought to be expressed specifically in endothelial cells, HIF-2α has recently been found in many other cell lines of nonendothelial origin.25 Thus, CHO cells might also express HIF-2α or even HIF-3α, the expression pattern of the latter not being known yet. We therefore specified our results by detecting HIF-1α in a human cell line (HepG2) using the monoclonal antihuman HIF-1α antibody mgc3 that does not cross-react with the other family members.14 

The functioning of the HRCHO5 cell line has been validated by stimulation with hypoxia as well as the known hypoxia-mimetics Co²⁺, Ni²⁺, and DFX.5,11,27Interestingly, Co²⁺ and DFX, but not Ni²⁺, showed additive effects on HRCHO5 reporter gene activity when combined with hypoxia. We previously reported that hypoxia and Co²⁺also had additive effects on Epo secretion in hepatoma cell lines,28 which is not in agreement with other reports.5 Our results imply that Co²⁺ and DFX do not only interfere with the putative oxygen sensor but might have additional positive effects on the oxygen signaling pathway, for example, associated with reactive oxygen species production by a localized Fenton reaction probably involved in oxygen sensing and signaling.29,30 

Using the HRCHO5 cell line, we identified the 2 fungal epolones,2,3 pycnidione and 8-methyl-pyridoxatin, as well as the related pyridone, CPX, as potent inducers of HIF-1α activation. Most of the studies in this work were performed with the epolone CPX, which has been chosen because of its relatively simple molecular structure and because it is commercially available at a low price. Clinically, CPX is used as an antimycoticum in dermatologic and vaginal creams.31 The mechanisms of its antimicrobial action have not been completely resolved, but CPX seems to interfere with membrane integrity, a probable HIF-1α–independent effect. Apart from the epolones, several other substances were tested in the HRCHO5 cell line, which we suspected to influence HIF-1 activity: insulin and insulin-like growth factors I and II32-34; interleukin-1α, interleukin-1β, tumor necrosis factor-α35; lipopolysaccharide36; tumor growth factor-β37,38; ferrous ethylenediammonium sulfate, FeCl₂, FeCl₃11,39; ZnCl₂40; the angiogenesis inhibitor epigallocatechingallate found in drinking tea41; and taurin.42 These agents have been reported either to be induced by hypoxia or to interfere with the expression of hypoxia-inducible genes. However, none of these agents affected HIF-1–dependent reporter gene activity in normoxic or hypoxic HRCHO5 cells (data not shown).

The epolones attenuated luciferase activity in HRCHO5 cells when added at concentrations above 16 μmol/L under normoxic conditions and above 4 μmol/L under hypoxic conditions. We attribute this effect to a putative cytotoxicity of the epolones, which might increase when combined with the additional stress of exposure to hypoxia. Consistent with this notion, we observed a detachment of the cells with a concomitant decrease of cellular reporter gene activity after treatment with higher doses of epolones (data not shown). In support of this idea, it has been reported that CPX (as well as DFX) can block the cell cycle at the G1/S phase boundary.43,44 HIF-1α protein induction in HepG2 cells by hypoxia and epolones (Figure 4) was diminished compared with the corresponding luciferase activities in HRCHO5 cells (Figure 2). The reason for this finding is not completely clear but might be related to the presence of other members of the HIF family that contribute to reporter gene activation or be related to a higher sensitivity of HepG2 cells to hypoxic cell culture conditions than CHO cells.

Interestingly, the function of both the established iron chelator DFX, a sideramine obtained from Streptomyces pilosus,17 and of CPX could be blocked dosage dependently by adding ferrous iron salts. Ferrous iron is rapidly oxidized by ambient oxygen yielding the DFX-chelatable ferric iron. We hence do not know whether ferrous or ferric iron preferentially inhibits CPX-mediated HIF-1α activation. Because we discovered that CPX is an iron-dependent chromophore, we could directly determine the iron chelation stoichiometry to be 1:3 iron:CPX. This is in agreement with the presence of 1 hydroxamic acid group in CPX (bidentate) compared with 3 such groups in DFX (hexadentate), the latter chelating iron at a 1:1 stoichiometry.

The well-characterized, HIF-1α–inducing iron chelators DFX11 and 2,2′-dipyridyl18 are thought to mimic hypoxia by displacing iron from the porphyrin ring of the putative heme oxygen sensor5,39 or by removing iron from the reactive oxygen species-generating Fenton reaction.29,30 Therefore, CPX could induce HIF-1–dependent gene expression via similar mechanisms. The finding that AlCl₃ also blocked CPX-induced reporter gene activity suggests that epolones (as well as DFX) potentially might induce HIF-1α by chelation of a metal other than iron. This would have important implications for the nature of the oxygen sensor, but further experiments will be required to establish such a mechanism.

While 50- to 100-μmol/L DFX is necessary to maximally induce HIF-1–dependent reporter gene expression, the epolones already show maximal activation at 4 to 8 μmol/L. Because we found an iron concentration in the cell culture medium of 16 μmol/L, the concentration of DFX, but not of CPX, would be sufficient to completely chelate total iron in the medium. Considering that a 3-fold higher molarity of the bidentate CPX than of the hexadentate DFX is required to chelate the same quantity of iron, and regarding the fact that 8-methyl-pyridoxatin is even 10 times more potent than CPX (5-fold induction of Epo gene expression at 0.3 μmol/L compared with 3-μmol/L CPX),2 we conclude that the epolones cannot activate HIF-1α simply by chelating all iron in the medium as might be the case for DFX.

The hexadentate iron chelator DFX has been reported to be ineffective as an intracellular iron chelator (at concentrations similar to those used in our study), whereas the lipophilic bidentate hydroxypyridinone class of iron chelators efficiently chelated intracellular iron.45 In analogy, differences between the cellular permeability of DFX and epolones could explain their different concentration optima. Indeed, the bidentate iron chelator 1,2-diethyl-3-hydroxypyridin-4-1 (CP-94) induced VEGF messenger RNA (mRNA) expression in Hep3B hepatoma cells already at 10 μmol/L.37 However, CP-94 did not induce Epo mRNA at 10 μmol/L, and the highest VEGF and Epo mRNA induction was found at 200 μmol/L, which is clearly different from our results with the epolones.

In conclusion, fundamental differences appear to exist between CPX and other known hypoxia-mimicking iron chelators with respect to their structure; metal ion preference; iron chelation stoichiometry, affinity, and kinetics; cellular uptake; intracellular stability; and ability to chelate the intracellular labile iron pool. Therefore, differences might also exist in their interaction with the putative oxygen sensor iron center as well as their interference with the oxygen signaling pathway. Better understanding of these differences probably will help in the elucidation of the different steps involved in the regulation of oxygen-dependent gene expression.

We are grateful to Peter J. Nielsen for the gift of CHO cells, Franziska Parpan for technical assistance, and Christian Gasser for the artwork.