Human Cytoplasmic Aconitase (Iron Regulatory Protein 1) Is Converted into Its [3Fe-4S] Form by Hydrogen Peroxide in Vitro but Is Not Activated for Iron-responsive Element Binding*

Iron regulatory protein 1 (IRP1) regulates the synthesis of proteins involved in iron homeostasis by binding to iron-responsive elements (IREs) of messenger RNA. IRP1 is a cytoplasmic aconitase when it contains a [4Fe-4S] cluster and an RNA-binding protein after complete removal of the metal center by an unknown mechanism. Human IRP1, obtained as the pure recombinant [4Fe-4S] form, is an enzyme as efficient toward cis-aconitate as the homologous mitochondrial aconitase. The aconitase activity of IRP1 is rapidly lost by reaction with hydrogen peroxide as the [4Fe-4S] cluster is quantitatively converted into the [3Fe-4S] form with release of a single ferrous ion per molecule. The IRE binding capacity of IRP1 is not elicited with H2O2. Ferrous sulfate (but not other more tightly coordinated ferrous ions, such as the complex with ethylenediamine tetraacetic acid) counteracts the inhibitory action of hydrogen peroxide on cytoplasmic aconitase, probably by replenishing iron at the active site. These results cast doubt on the ability of reactive oxygen species to directly increase IRP1 binding to IRE and support a signaling role for hydrogen peroxide in the posttranscriptional control of proteins involved in iron homeostasis in vivo.

Over the last 10 years, proteins known as iron regulatory proteins (IRPs) 1 have been recognized as the major regulatory factors for iron uptake and storage in most mammalian cells (1,2). These proteins bind to specific stem-loop structures (3), the iron-responsive elements (IREs), on the mRNA of the target genes to repress translation or enhance mRNA stability. Ferritin subunits, transferrin receptors, and erythroid 5-aminolevulinic acid synthetase are regulated by IRPs, but other proteins, such as subunit b of Drosophila succinate dehydrogenase and mitochondrial aconitase, the involvement of which in iron metabolism is less direct, display IREs on their mRNA. IRPs may thus have more widespread regulatory functions, all the more so as they have been shown to respond to stress mediators and signaling agents, such as nitrogen monoxide, cytokines, and hydrogen peroxide, in addition to iron concentration.
IRP1 belongs to an isomerase family of proteins containing one [4Fe-4S] cluster and including mitochondrial aconitase (4). It exhibits aconitase activity (5). Extensive disassembly of the cluster is required to allow the protein to interact with IRE (6,7). The reciprocal control of both activities ("iron-sulfur switch") appears as a key mechanism in the regulatory function of IRP1. Low concentrations of H 2 O 2 induce a fast and potent activation of IRP1 in intact cells, but not in lysates (8 -11). The latter results seemingly contradict the notion that reactive oxygen species (ROS) 2 in general, and H 2 O 2 in particular, directly target and destroy iron-sulfur clusters (12,13). H 2 O 2 is a by-product of aerobic metabolism and serves as an effector molecule in cell defense; it modulates cellular functions either as a direct oxidation agent or as a signaling molecule (14), and the exact role of H 2 O 2 in the conversion of IRP1 activities needs to be fully defined. The efficient heterologous production of human IRP1 (rhIRP1) with its full content of [4Fe-4S] cluster has allowed us to assess the consequences of reactions with oxygen derivatives on the functional properties of the cytoplasmic aconitase.

MATERIALS AND METHODS
Electrophoretic Mobility Shift Assay-All experiments using the IRE probe were carried out in diethyl pyrocarbonate-treated solutions. A human ferritin H-chain IRE probe was 33 P-labeled as described (15) using T7 RNA polymerase (Roche Molecular Biochemicals) and a mixture of ribonucleoside triphosphates containing [␣-33 P]UTP synthesized from the XbaI-linearized plasmid I12CAT (16). The reaction mixture was separated by denaturing electrophoresis on a 20% acrylamide gel containing 7 M urea in 0.5ϫ Tris borate-EDTA buffer, and the labeled probe was extracted in 0.5 M ammonium acetate, 0.5 mM EDTA, 0.1% SDS before ethanol precipitation. IRP1 binding to the probe was carried out in an anaerobic chamber ensuring an oxygen concentration of less than 2 ppm (Jacomex, Livry-Gargan, France). 10-l reactions were carried out for 20 min in 10 mM Hepes, pH 7.6, 3 mM MgSO 4 , 40 mM KCl with 0.4 unit of human placental ribonuclease inhibitor (Roche Molecular Biochemicals). 5 mg/ml of heparin were then added, and incubation was continued for 10 min. 1 l of 50% glycerol with 0.05% bromphenol blue was added to each reaction mixture. The samples were separated (outside the anaerobic chamber in a cold room) on a nondenaturing 5% acrylamide gel in 0.25ϫ Tris borate-EDTA buffer after 1 h of preelectrophoresis at 80 V. The gel was run for 75 min at a constant voltage of 80 V and dried under vacuum. The radioactive bands were analyzed by quantitative autora-* 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 U.S.C. Section 1734 solely to indicate this fact.
Heterologous Production of Human IRP1-Plasmid pT7-His-hIRF contains the complete cDNA for human IRP1 fused downstream of a histidine anchor (16). The tag was removed as follows. A 283-base pair DNA fragment corresponding to the intact 5Ј-coding sequence of hIRP1 was amplified with the NS1 (5Ј-ggatccatcagtacatatgagcaacccattcgc) and NS3 (5Ј-cccgtaaagtcctgcaggatgacac) oligonucleotides. NS1 was designed to introduce a NdeI recognition sequence (catatg) overlapping the start codon. The amplified fragment was digested with NdeI and PstI and cloned into NdeI-PstI-cleaved pT7-7 (17). The resulting plasmid was cleaved with HindIII and PstI and ligated with the PstI-HindIII fragment from pT7-His-hIRF. The sequence of the final plasmid was checked by dideoxynucleotide termination sequencing and corresponded to the intact coding sequence of the human IRP1 cDNA inserted into the multiple cloning site of pT7-7.
The bacterial synthesis of human IRP1 was evidenced by [ 35 S]Cys labeling under different growth conditions, as described (18). In addition, 5 g of induced Escherichia coli K38/pGP1-2 extracts containing the pT7-7-based plasmid with the intact coding sequence of the human IRP1 displayed a strong retarded band in the standard IRE binding assay similar in intensity to that produced by 5 g of E. coli BL21(DE3)/ pT7-His-hIRF extracts. E. coli K38/pGP1-2 extracts did not give rise to any complex with labeled IRE (not shown). The large scale (20 l) bacterial production of hIRP1 was carried out in Terrific broth (Life Technologies, Inc.). The bacteria were grown at 30°C until they reached early stationary phase. Synthesis of T7 RNA polymerase was induced for 1 h at 42°C and the production of hIRP1 lasted for 3-5 h after cooling the medium back to 30°C. Cells were harvested by centrifugation and kept at Ϫ80°C until use.
Purification of hIRP1-All purification steps and subsequent experiments were carried out in 20 mM Tris-Cl, pH 7.0, 0.2 mM citrate, 0.05 mM phenylmethylsulfonyl fluoride (Buffer A) under strictly anaerobic conditions using argon lines or an anaerobic chamber (19). Both activities of IRP1 (aconitase and IRE binding) were measured. The lysate obtained by ultrasonic treatment was fractionated by 25-65% saturation ammonium sulfate precipitation at 4°C. The 65%-pellet was dialyzed overnight and loaded successively on two ion-exchange columns, 30 ml of CM52 followed by 50 ml of DE52 (Whatman) equilibrated in Buffer A. The active fraction did not bind and was chromatographed on a 200 ml phenyl-Sepharose CL 4B column (Amersham Pharmacia Biotech) after adding 20% saturated ammonium sulfate. rhIRP1 was eluted by Buffer A and loaded on a 150-ml hydroxyapatite column (Bio-Gel HTP, Bio-Rad). 50 mM potassium phosphate, pH 7.0, with 0.2 mM citrate were used to recover rhIRP1, which was gel-filtrated on a 500-ml column of Sephacryl S-200 HR (Amersham Pharmacia Biotech) in Buffer A. A final purification step was carried out on a PL-SAX (Polymer Laboratories) high pressure liquid chromatography column equilibrated in Buffer A and developed with a 0 -0.1 M sodium chloride gradient to give the material characterized in Fig. 1.
Enzymatic and Other Assays-Pig heart mitochondrial aconitase was obtained from Sigma and further purified by a combination of cation exchange and gel filtration steps. The partially active fraction was activated by incubation with 5 mM dithiothreitol, 25 M FeSO 4 , and 12.5 M Na 2 S, desalted, and quickly analyzed for protein concentration and enzymatic activity. The aconitase activities were measured either by monitoring NADPH production in the coupled assay with isocitrate dehydrogenase (20) for partially purified fractions or by recording the decrease of absorbance at 240 nm corresponding to the decrease of cis-aconitate used as substrate (21) with pure enzymes. Protein concentrations were measured by established methods (22,23) and agreed within 20%. Using an extinction coefficient at 400 nm of 16000 M Ϫ1 cm Ϫ1 for pure rhIRP1 gave similar values.
For the experiments assessing the reactivity of rhIRP1 with H 2 O 2 , a 30% H 2 O 2 solution in water (Aldrich) was filtrated, right before use, through a mixed bed AG 501-X8 resin (Bio-Rad) to remove any contaminating ionized material. Caution: this procedure is potentially hasardous and care has been taken to minimize the amounts of processed reactants. After thorough removal of oxygen by sparging the solution with argon, the H 2 O 2 concentration was determined spectrophotometrically (⑀ 240 nm ϭ 40 M Ϫ1 cm Ϫ1 ). These deionized solutions could not reduce cytochrome c. The reactions between 20 M pure rhIRP1 and H 2 O 2 were in 50 mM Tris-Cl, pH 7.4, in the absence of substrate; the aconitase activity was periodically measured with 25 g of protein and saturating amounts (200 M) of cis-aconitate. The rates of cis-aconitate conversion were initially irreproducible; it was then realized that experimental conditions implementing argon lines and liquid handling with stainless steel needles were prone to leak uncontrolled amounts of metals in the presence of H 2 O 2 . Consequently, kinetic experiments have all been carried out in an anaerobic chamber with disposable plastic dispensing equipment. As an example, this procedure yielded rate constants for inactivation of IRP1 by H 2 O 2 agreeing within 20% for different experiments. The conditions for recording UV-visible and EPR spectra were as described (19). The spin trapping agent DEPMPO was obtained from Oxis International, Inc. (Portland, OR).

Molecular Properties of the rhIRP1
Identity of the Purified Protein-Numerous reports have previously described the production of IRP1 in many different systems (7, 16, 24 -27). None of the expression and purification strategies were designed to recover the protein as pure cAcn with its full content of [4Fe-4S] cluster. The scheme described under "Materials and Methods" afforded a protein migrating as a single band of apparent molecular weight around 95,000 on SDS-polyacrylamide gel electrophoresis (Fig. 1A). Amino acid sequencing of this material gave the first eleven residues expected for human IRP1 starting with the serine following the initial methionine.
Spectroscopic Properties- Fig. 1B shows the electronic absorption spectrum of rhIRP1. In addition to the protein maximum at 280 nm, a broad band centered at about 400 nm gave a yellow-brownish color to the sample and indicated the presence of a chromophore. The visible band accounted for about one tenth of the intensity at 280 nm. UV-visible spectra of cytoplasmic aconitases purified from natural sources have not been reported, but the visible absorbance of pure bovine mAcn (see, for example, Ref. 28) is very similar to that of rhIRP1.
The EPR spectrum of rhIRP1 is axial and virtually identical to previously reported spectra of bovine cAcn (5,28) with the characteristic [3Fe-4S] ϩ g values of 2.03 and 2.014 (Fig. 1C). EPR double integration, using a copper EDTA reference, of several preparations yielded values of less than 0.1 spin/protein molecule, indicating that only a minority of the rhIRP1  (46). B, electronic absorption spectrum. C, X-band EPR spectrum of 63 M rhIRP1: microwave frequency, 9.653 GHz; power, 10 W; modulation frequency, 100 kHz; modulation amplitude, 1 millitesla; temperature, 10 K. molecules binds a [3Fe-4S] cluster. In contrast, purifications of mammalian mAcn were designed to quantitatively recover the stable but inactive [3Fe-4S] form that can be reactivated in vitro (29).
Enzymatic Properties of the rhIRP1-The rhIRP1 molecules devoid of metals do not contribute to the visible part of the spectrum in Fig. 1B. The concentration of those displaying IRE binding activity was assessed with increasing amounts of rhIRP1 assayed under anaerobic conditions. The activity of 20 ng of the protein as isolated was hardly detected, whereas addition of 2% 2ME, which fully reveals the IRE binding activity (30), to as little as 2 ng gave a clear retarded band on gels (not shown). Approximately 50-fold more rhIRP1 in the absence of 2% 2ME was necessary to obtain as much IRP1-IRE complex as formed in the presence of the reducing thiol. It can therefore be estimated conservatively that the IRE-binding active form of rhIRP1 as isolated represents less than 5% of the total preparation.
The high visible/UV absorbance ratio for a protein of this size combined with the very small concentrations of molecules with [3Fe-4S] clusters or displaying IRE binding activity indicate that rhIRP1 is mainly (Ͼ80%) isolated as cytoplasmic aconitase containing one [4Fe-4S] 2ϩ cluster. This conclusion is borne out by the enzymatic properties of purified rhIRP1. The catalytic efficiency of rhIRP1 measured as the k cat /K m ratio in the conversion of cis-aconitate was not significantly different from that of reactivated mitochondrial aconitase and did not change under conditions favoring the insertion of [4Fe-4S] clusters (Table I).

Reactivity of the rhIRP1 toward Oxygen Derivatives
Superoxide Inactivates cAcn-In addition to switch activities as a function of the intracellular iron concentration, IRP1 has been shown to respond to a variety of cellular conditions, many of them contributed by ROS and referred to as oxidative stress (2,(31)(32)(33). In contrast to many other Fe-S proteins (34), the aconitase activity of rhIRP1 was insensitive to exposure of the enzyme in air for more than 90 min. However, superoxide ion production aerobically (from the xanthine/xanthine oxidase reaction) or anaerobically (from KO 2 ) rapidly bleached the aconitase activity ( Fig. 2A). Similar results have already been reported for other aconitases (12,35,36), including IRP1 (28,(37)(38)(39) and, because of the efficient inhibition of aconitase activity by superoxide anion, all other reactions reported herein have been carried out under strictly anaerobic conditions.
Hydrogen Peroxide Inactivates cAcn, but Ferrous Ions Prevent the Inactivation-The reactivity of the aconitase form of IRP1 with hydrogen peroxide was investigated. When 100 M deionized H 2 O 2 was added to 20 M rhIRP1 as isolated, the aconitase activity decreased (Fig. 2). However, adding ferrous sulfate to the medium inhibited the loss of activity until no inactivation was detected for 90 min with 100 M FeSO 4 . Relatively low concentrations of ferrous sulfate decreased the rate of cAcn inactivation when added before H 2 O 2 (Fig. 2) and the inhibition was stopped when large concentrations of metal were added after some reaction with hydrogen peroxide had already occurred (Fig. 2B). Glutathione (1 mM) did not resume inhibition.
In contrast to ferrous sulfate, 100 M ferric chloride, sodium sulfate, or ferrozine could not protect rhIRP1 from the inhibitory effect of 100 M H 2 O 2 , and 100 M ferrous EDTA hardly interfered with the reaction (Fig. 2A). These observations on aconitase activity in the presence of ferrous ions may well explain the inconsistency of our initial measurements carried out with stainless steel needles (see under "Materials and Methods"). The presence of 100 M either citrate or transaconitate did not afford protection of cAcn against inactivation by H 2 O 2 .
Hydrogen Peroxide Does Not Activate IRE Binding of rhIRP1-Although rhIRP1 as isolated is marginally active in binding IRE (Fig. 3, lane 2), the protein can be efficiently activated to bind to IRE, for instance by incubation with high concentrations of 2ME (Fig. 3, lane 1). In this respect, the recombinant preparation described herein compares with previously reported ones for which high IRE binding activities were obtained (7, 16, 24 -27). The usual description of the IRE-binding activation of IRP1 is that of the destruction of the iron-sulfur cluster liberating its three cysteine ligands and exposing other residues of the active site to interaction with IRE (6). Numerous reports (8 -11, 39) have shown activation of IRP1 by treatment of whole cells with H 2 O 2 . H 2 O 2 reaction with rhIRP1, under conditions that almost completely inhibit aconitase activity (Fig. 2B), failed to reveal the IRE binding activity (Fig. 3, lanes 3-6), although the aconitase-inactive protein kept the ability to be activated for IRE binding by 2ME (Fig. 3, lane 7). The presence of 100 M FeSO 4 in the reaction medium, which prevents aconitase inhibition, did not change this pattern (Fig. 3, lanes 8 -14). Therefore, direct reaction of H 2 O 2 with rhIRP1 failed to convert the protein into its IREbinding form, even under conditions that suppress its aconitase activity.
Hydrogen Peroxide Quantitatively Generates [3Fe-4S] rhIRP1-Besides its two (aconitase and IRE-binding) active forms, other molecular species of IRP1 may occur, such as proteins with oxidized cysteines unable to interact with IRE (40 -42) or proteins in which the [4Fe-4S] cluster is converted into the inactive [3Fe-4S] cluster (28). The fate of metal ions   Under the conditions of Fig. 2, the inactivation of rhIRP1 was not correlated with any changes of the visible absorbance at 400 nm. However, the difference spectrum obtained by subtracting spectra before reaction with H 2 O 2 from those recorded after displayed two maxima at 335 and 485 nm (not shown). This difference spectrum is strikingly similar to that of Azotobacter vinelandii ferredoxin I minus Clostridium pasteurianum ferredoxin taken as representative examples of [4Fe4S] 2ϩ [3Fe-4S] ϩ and 2[4Fe-4S] 2ϩ proteins, respectively. When the same reaction was followed by EPR, the characteristic signal of [3Fe-4S] ϩ clusters as in Fig. 1C built up to reach a plateau at 1 spin/protein molecule as cAcn activity disappeared. Only a minor increase of the high spin ferric EPR signal at g ϭ 4.3 was detected (Fig. 4B).
This conversion of [4Fe-4S] 2ϩ into [3Fe-4S] ϩ releases iron. The experiment between 20 M rhIRP1 and 100 M H 2 O 2 was carried out in the presence of ferrozine, a strong chelator of ferrous ions. The calculated concentration of Fe(Fz) 3 2ϩ was slightly smaller than that of the protein when aconitase activity was abolished, and it slowly decreased afterward (Fig. 5), probably as the result of oxidation by excess H 2 O 2 . When ferrozine was replaced by desferal, a strong chelator of ferric ions, the EPR spectra elicited a high spin ferric signal at g ϭ 4.3, the intensity of which was correlated with that of the [3Fe-4S] ϩ signal in the g ϭ 2 region (Figs. 4 and 5). The same correlation was found with the visible absorbance due to ferrioxamine (Fig.  5). When desferal was added after the reaction between the protein and hydrogen peroxide was completed, the maximal g ϭ 4.3 signal intensity was developed after a few additional minutes (Fig. 4C). The presence of cis-aconitate at the onset of the reaction also helped developing the g ϭ 4.3 signal, probably arising from ferric citrate in this case (Fig. 4D).
Increasing the H 2 O 2 concentration did not generate other EPR signals or any visible modification of the cluster nuclearity. The [3Fe-4S] cluster signal was only bleached after extensive reaction of the protein with 1000-fold (100 mM) more H 2 O 2 . In a similar way, no other EPR signals than that assigned to the [3Fe-4S] ϩ cluster were detected by reaction of rhIRP1 with KO 2 , in agreement with previously reported data (28).
Hydrogen Peroxide, Not Hydroxyl Radical, as the Reactive Species-Our experimental conditions with ferrous cations released from rhIRP1 and H 2 O 2 should produce other ROS, such as hydroxyl radicals, in what is often called the Fenton reaction (43). The generation of radicals in the reaction between cAcn and H 2 O 2 was evidenced by DEPMPO trapping (Fig. 6). The EPR spectra formed with this pyrroline 1-oxide derivative are clearly different for superoxide and hydroxyl spin adducts (44). Only DEPMPO-OH was detected in the reaction between rhIRP1 and H 2 O 2 (Fig. 6C) and the radical species in frozen solutions (Fig. 6B) coexisted with the [3Fe-4S] ϩ cluster (Fig. 6,  A and B). In agreement with the observations of Fig. 2, adding 100 M FeSO 4 to the reaction greatly decreased the amounts of [3Fe-4S] ϩ and of DEPMPO-OH formed (not shown).
In order to probe the role of these hydroxyl radicals in the inactivation of cAcn by H 2 O 2 , dimethyl sulfoxide was used as a radical scavenger (45). 100 mM dimethyl sulfoxide displayed no effect on the cAcn activity of rhIRP1. When added to the reaction between rhIRP1 and H 2 O 2 , dimethyl sulfoxide hardly decreased the rate and extent of inactivation of cAcn ( Fig. 2A). Furthermore, glutathione, which regenerates the ferrous ions needed to produce hydroxyl radicals from hydrogen peroxide, had no effect on the rate of inactivation of cAcn (see above). It is thus likely that the hydroxyl radicals are a product, rather than an active intermediate, of the conversion of [4Fe-4S] into [3Fe-4S] rhIRP1. low fields (left panels), temperature, 4 K; microwave power, 1 mW; modulation amplitude, 1 millitesla; high fields (right panels), temperature, 10 K; microwave power, 0.1 mW; modulation amplitude, 1 millitesla. The low field spectra were recorded with a 2-fold higher receiver gain than the high field region, but the same gain for a each field region was used for all spectra. Hydrogen Peroxide-reacted rhIRP1 Can Be Reactivated-The reversibility of the conversion of [4Fe-4S] rhIRP1 into the inactive [3Fe-4S] form by H 2 O 2 was assessed. The enzyme was gel-filtrated to remove low molecular weight reactants after interaction with hydrogen peroxide, and its aconitase activity was assayed. Less than 10% of the initial (before reaction with H 2 O 2 ) specific activity was recovered by this procedure, but incubation with molar excess of reductant (dithiothreitol) and ferrous sulfate afforded an enzyme with about half of the starting specific activity (data not shown).

DISCUSSION
Bacteria Can Produce Active Human cAcn-The occurrence and function of iron regulatory proteins have so far only been described for metazoans (2,31), although the presence of mRNA sequences resembling IRE has been noticed in bacteria (46). No binding of proteins from E. coli to the human H-ferritin IRE has been evidenced under our experimental conditions. The bacteria were perfectly able to insert the [4Fe-4S] cluster of human IRP1, because more than 80% of the recombinant protein was recovered as active aconitase. The suitability of bacteria to synthesize iron-sulfur proteins of eukaryotic, including human, origin is not without precedent (see, for example, Refs. 47 and 48), and some evidence from a previous report (41) indicated the bacterial production of holo-IRP1. Therefore, no specific cellular components in eukaryotes are needed to posttranslationally convert hIRP1 into active cAcn.
The enzymatic properties of rhIRP1 have been compared with those of the homologous mAcn (Table I). The endosymbiont hypothesis for the origin of mitochondria (49) implies that IRP1 evolved from mAcn. Significantly, the protein acquired the ability to bind to specific RNA sequences but did not loose much of its enzymatic capabilities.
Role of H 2 O 2 in IRP1 Activation-As a regulator of cellular iron metabolism, IRP1 can be considered as an iron sensor, although the details of the interaction between the protein and metals remain to be fully explored. The balance between assembly and complete destruction of the [4Fe-4S] cluster appears to be the main outcome of the regulatory effect (6,7).
The results reported herein confirm, after other studies (28,37), that superoxide inhibits cAcn activity. Fig. 2 shows that the same is true when H 2 O 2 substitutes for superoxide. However, these two major intermediates of oxygen reduction, at molar excess likely to have biological significance, are unable to convert cAcn into IRE-binding IRP1 (38) (Fig. 3). Therefore, the activation of IRP1 observed in whole cell studies probably obeys a more complex mechanism leading to the complete disruption of the cluster, whereas maintaining cysteines reduced to interact with the RNA recognition sequence. A signal transduction cascade initiated by H 2 O 2 has been proposed (10,11,50), but it remains to be fully characterized. Links between production or action of ROS and phosphorylating events have been evidenced in other instances (14). The present results show that activation of IRP1 by H 2 O 2 requires a specific triggering system to fully remove the [4Fe-4S] cluster. This could be achieved either by providing a more specific destructing reactant for the iron-sulfur cluster than superoxide ion or hydrogen peroxide, or by changing the conformation of the protein, through phosphorylation, for instance, so that a more readily access of the reactive species to the cluster may be possible. Indeed, H 2 O 2 activates isozymes of protein kinase C (51). IRP1 is a substrate for the kinase, although the [4Fe-4S] form of IRP1 does not seem to be efficiently phosphorylated (52). The relevance, if any, of these observations to the H 2 O 2 activation of IRP1 will have to be demonstrated.
Mechanism of Human cAcn Inactivation by H 2 O 2 -In contrast to their inability to activate IRE-binding IRP1, superoxide ions and H 2 O 2 are efficient inhibitors of cAcn (Fig. 2). In the reaction with H 2 O 2 , the protecting role of increasing concentrations of ferrous sulfate may seem surprising when considering the largely documented reactivity of these ions with H 2 O 2 to generate highly oxidizing compounds (see, for example, Refs. 43 and 53). The production of hydroxyl radicals in our reaction conditions, even without exogenous iron, has been demonstrated (Fig. 6), but these radicals do not change the nuclearity of the cluster beyond the [3Fe-4S] form. Significantly, strongly chelated iron compounds, such as Fe-EDTA, failed to afford the same protection as ferrous sulfate, indicating that rapidly mobilized ferrous ions are needed to interfere with the action of H 2 O 2 .
The data in Fig. 5 show that a single iron atom is removed by H 2 O 2 from the cluster of rhIRP1. Several lines of evidence point to that atom being a ferrous ion. First, no EPR signal at g ϭ 4.3 appears in the spectra after reaction (Fig. 4B), thus indicating that the released iron is either ferrous or is part of a complex in which ferric ions are coupled (e.g. ferric hydroxide polymers). Second, quantitative UV-visible titration of the ferrous complex with ferrozine correlates with the appearance of the [3Fe-4S] ϩ rhIRP1. Last, the production of hydroxyl radicals (Fig. 6) from hydrogen peroxide needs a reductant, and the most readily available sources of electrons under our experimental conditions are ferrous ions originating from the cluster. The resulting ferric ions can be quantitatively coordinated by desferal to give the EPR active ferrioxamine B (Fig. 4C), but the formation of the complex is delayed when the chelator is not present at the onset of the reaction. Overall, these data agree with the relevant part of the previously proposed scheme for the reaction of several [4Fe-4S] dehydratases with superoxide ion (12 vicinity of the cluster after reaction with H 2 O 2 : for instance, no obvious magnetic interactions between the radical and the cluster signals in Fig. 6 were detected. Because the reaction of rhIRP1 with H 2 O 2 quantitatively releases only one ferrous ion (see above), H 2 O 2 further generates hydroxyl radicals with Fe 2ϩ probably after some diffusion of the metal ion away from the active site has occurred. This may not be true in the reaction between [4Fe-4S] dehydratases and superoxide ions as extensive destruction of the clusters, with more than 3 iron atoms released by cluster, were monitored (12).
In this framework, the effect of exogenous ferrous ions is probably to replenish the labile iron atom of the cluster removed by the action of hydrogen peroxide, as long as these ferrous ions can be readily mobilized for this purpose (Fig. 2). Alternatively, adding more ferrous ions may help reducing hydrogen peroxide and decrease its deleterious effect. However, ferrous EDTA has the same ability to reduce H 2 O 2 but not the same protecting role against inactivation of cAcn (Fig. 2).
Physiological Implications-The deleterious effects of oxygen derivatives in the presence of reduced transition metals may be mechanistically complex, as concluded by analyzing ROS induced DNA oxidation (54,55). The reaction of H 2 O 2 with rhIRP1 is far less destructive than would be anticipated from the strong reactivity of the combination of H 2 O 2 and ferrous sulfate (the original Fenton catalyst) and the usual instability of iron-sulfur proteins (56). Iron-sulfur and other iron containing proteins have often been suggested as the primary intracellular targets of ROS (see, for example, Ref. 56), but the most prominent cytoplasmic iron-sulfur protein studied herein only shows a limited structural defect induced by hydrogen peroxide.
Previous studies have shown that the direct reaction of low concentrations (e.g. 10 M) of externally applied hydrogen peroxide with IRP1 is unlikely, although these conditions yield high IRE binding activation (8 -11). This signaling role of H 2 O 2 must be contrasted with its possible intracellular occurrence at high concentrations. Defective mitochondrial respiratory chains are a major site of hydrogen peroxide production (57,58). H 2 O 2 may diffuse out the mitochondrial membrane and react with IRP1 as presented here, because our observations qualitatively apply to a large range of H 2 O 2 /IRP1 ratios. Therefore, an H 2 O 2 burst, in the absence of other contributing factors, does not increase the IRE binding activity of IRP1, it does rapidly inactivate its aconitase function, and it has the ability to provide one, but not more, iron atom per molecule of IRP1 to other cellular components or to amplify the oxidative stress condition via production of hydroxyl radicals. This detailed mechanistic information exemplifies the use of pure iron-replete rhIRP1 to unravel the interferences between intracellular iron metabolism and ROS action(s), including conditions, such as hypoxia, for which some uncertainties remain (59,60).