Of Two Cytosolic Aconitases Expressed in Drosophila, Only One Functions as an Iron-regulatory Protein*

In mammalian cells, iron homeostasis is largely regulated by post-transcriptional control of gene expression through the binding of iron-regulatory proteins (IRP1 and IRP2) to iron-responsive elements (IREs) contained in the untranslated regions of target mRNAs. IRP2 is the dominant iron sensor in mammalian cells under normoxia, but IRP1 is the more ancient protein in evolutionary terms and has an additional function as a cytosolic aconitase. The Caenorhabditis elegans genome does not contain an IRP2 homolog or identifiable IREs; its IRP1 homolog has aconitase activity but does not bind to mammalian IREs. The Drosophila genome offers an evolutionary intermediate containing two IRP1-like proteins (IRP-1A and IRP-1B) and target genes with IREs. Here, we used purified recombinant IRP-1A and IRP-1B from Drosophila melanogaster and showed that only IRP-1A can bind to IREs, although both proteins possess aconitase activity. These results were also corroborated in whole-fly homogenates from transgenic flies that overexpress IRP-1A and IRP-1B in their fat bodies. Ubiquitous and muscle-specific overexpression of IRP-1A, but not of IRP-1B, resulted in pre-adult lethality, underscoring the importance of the biochemical difference between the two proteins. Domain-swap experiments showed that multiple amino acid substitutions scattered throughout the IRP1 domains are synergistically required for conferring IRE binding activity. Our data suggest that as a first step during the evolution of the IRP/IRE system, the ancient cytosolic aconitase was duplicated in insects with one variant acquiring IRE-specific binding.

Iron is required for aerobic metabolism and is actively sensed, sequestered, and regulated by organisms, including hosts and pathogens, that often compete for metal acquisition (1). In multicellular organisms, signals exist for systemic control of iron metabolism (2). One of the central regulators of cellular iron metabolism is the IRP/IRE system (3). In brief, absence of iron is sensed by IRP1 and IRP2, which then bind to iron-responsive elements (IREs) 4 in the 5Ј-untranslated regions of mRNAs encoding several storage or consumer iron proteins, inhibiting their translation. IRP1 and IRP2 also bind to IREs in the 3Ј-untranslated regions of mRNAs encoding proteins that function in cellular iron import, stabilizing the transcripts and enhancing iron sequestration. There is much interest in defining how IRP1 and IRP2 sense iron levels. IRP1 interconverts between an iron-sulfur protein with aconitase activity and the apoprotein that binds to the IRE (4). IRP2 has an additional 73-amino acid domain inserted in the protein and has no aconitase activity (5,6). Extensive analysis of knock-out mice has led to the conclusion that IRP2 dominates mammalian cellular iron metabolism (7). IRP1 also contributes to the regulation of cellular iron metabolism, albeit to a lesser extent, because its iron-sulfur cluster is efficiently repaired and the protein functions mostly as an aconitase (8 -10). It is also well established that both IRPs also sense and respond to oxidative stress (3,11).
The ancestral protein that gave rise to the IRP/IRE system is likely an aconitase, which is broadly distributed in all phyla (12) and has binding affinity to nucleic acids (13). Saccharomyces cerevisiae contains only a single aconitase gene, encoding a mitochondrial enzyme, though low amounts of aconitase can be detected in the cytosol (14). Aconitase activity in both mitochondria and cytosol is also documented for the more primitive protozoan parasites Trypanosoma brucei and Plasmodium falciparum, each of which contains a single aconitase gene (15,16). Conversely, mitochondrial and cytosolic aconitases are encoded by different genes in Caenorhabditis elegans (17,18). Cytosolic aconitases from these lower eukaryotes have little or no IRE binding activity (18 -20), and no functional IREs have been found in their genomes. In contrast, the insects Manduca sexta, Aedes aegypti, Anopheles gambiae, and Calpodes ethlius contain cytosolic aconitases that can bind specifically to endogenous or mammalian IREs, and they may exemplify an intermediate step in the evolution of the IRP/IRE system (21)(22)(23).
Drosophila melanogaster was the first insect discovered to possess IRE binding activity (24) that can regulate endogenous levels of succinate dehydrogenase subunit B in an iron-dependent manner (25)(26)(27). Together with an alternatively spliced transcript of the ferritin 1 heavy chain homolog (Fer1HCH), these are the only two genes in the Drosophila genome that were shown to possess an IRE (28 -30). Although electrophoretic analysis of fly extracts suggests the presence of a single protein binding to IRE and a single protein possessing cytosolic aconitase activity (31), Drosophila actually contains two highly homologous IRP1-like proteins (IRP-1A and IRP-1B) encoded by different genes (17,32). Herein, we have purified heterologously expressed IRP-1A, IRP-1B,  and hybrid IRP-1A/IRP-1B proteins and determined that although both  IRPs are functional aconitases, only IRP-1A binds IREs. We have also generated transgenic flies that overexpress either protein in a tissuespecific manner and discovered that overexpression of IRP-1A can be lethal.

EXPERIMENTAL PROCEDURES
Construction of Plasmids and Transgenic Flies-To express recombinant protein in bacteria, the cDNAs corresponding to IRP-1A and IRP-1B (GenBank TM accession code AJ252016 and AJ252017) were cloned into the expression vectors pQE 30 (Qiagen) and into pTrc 99A (Amersham Biosciences), respectively (33). Both expression plasmids (pEX-IRP1A and pEX-IRP1B) introduce a His 6 tag in the N terminus of the protein.
To generate UAS-IRP-1A and UAS-IRP-1B flies, expressed sequence tags AT02682 and GM05743 were obtained by the Berkeley Drosophila Genome Project and subcloned into pUAST. The resulting plasmids were verified by sequencing and injected into embryos using conventional techniques (34). All phenotypes reported were confirmed with independent transgene insertions, which were mapped to different chromosomes.
Protein Expression and Purification-Recombinant IRP-1A was produced in Escherichia coli M15, transformed with the pEX-IRP1A. The bacteria were grown in Luria Bertani medium containing 50 g/ml of carbenicillin and 20 g/ml of kanamycin to a density of A 600 ϭ 1 and then treated with 2 mM isopropyl ␤-D-thiogalactopyranoside for 2 h at 25-27°C to induce expression of the plasmid. The cells were harvested by centrifugation at 3,000 ϫ g for 10 min at 4°C. Large-scale batch purification of His 6 -IRP-1A under native condition was performed according to the manufacturer for TALON™ Metal Affinity Resin (Clontech). The recombinant protein was then eluted with sonication buffer (20 mM Tris, pH 8.0, 100 mM NaCl) containing 50 mM imidazole and dialyzed against N buffer (24 mM Hepes, 150 mM KOAc, 1.5 mM MgCl 2 , 5% glycerol, pH 7.6) overnight. The protein suspension was concentrated by using Centricon-30 (Amicon), and the protein concentration was determined by the Bradford method. The protein was used directly or stored at Ϫ80°C. The isolation of His 6 -IRP-1B and chimeric His 6 -IRP-A to His 6 -IRP-E constructs was done as described above for His 6 -IRP-1A with the following modifications: E. coli BL21, transformed with the construct pEX-IRP1B or any of the chimeric constructs, was grown in Luria Bertani medium containing 50 g/ml of ampicillin.
Electrophoretic Mobility Shift Assay (EMSA)-EMSA was performed as previously published (29). The recombinant IRPs were reduced with 2% 2-mercaptoethanol (2-Me) before addition of the probe. Where indicated, purified IRP-1A and 1B were treated instead with 10 mM cysteine, 50 M FeSO 4 , and 5 mM EDTA. For the competition experi-ment, 100-fold molar excess of cold oligonucleotide was mixed with the radioactive IRE before the protein was added. The gels were subjected to autoradiography or to imaging plates for a fluorescent image analyzer (FLA-3000 system; Fuji Film). The radioactivity was quantified by using the software Multi Gauge v2.2 (Fuji Film).
The competitor RNAs (wild type and mutant) have been transcribed from the vectors I-12.CAT and I-19.CAT, respectively (33). I-12 is derived from human ferritin IRE, whereas I-19 is identical to I-12 except for a cytosine deletion in the loop, which will prevent binding of IRP.
Aconitase Assay-Aconitase activity was measured by the NADPcoupled aconitase assay according to Drapier and Hibbs (35). The assay was performed at 30°C in a total volume of 500 l in the presence of 5 g of recombinant Drosophila IRP-1A or IRP-1B and an excess of isocitrate dehydrogenase (aconitase-free; Sigma), which catalyzes the oxidation of isocitrate into 2-oxoglutarate coupled to the production of NADPH, which is followed at 340 nm. Sample without citrate was used as blank (⌬A 340 Ͻ0.01/h). For the measurements performed in wholefly lysates, cis-aconitate (Sigma) was used as a substrate for measuring activity; the in-gel aconitase assay is described in Ref. 10.

Expression and Purification of Recombinant IRP-1A and IRP-1B-
IRP-1A and IRP-1B cDNAs were cloned into inducible vectors to overexpress recombinant IRP1s with N-terminal His 6 tag extensions in bacteria. Soluble proteins with a size of ϳ100 kDa were extracted from E. coli by sonication and centrifugation and were isolated by a metal affinity batch method to Ͼ90% purity. The efficacy of successive purification steps for isolation of recombinant IRP-1A and IRP-1B was analyzed using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 1). The recovery per liter of culture was ϳ0.2 mg for IRP-1A and 1 mg for IRP-1B. Initiation of protein translation at the correct methionine was confirmed by N-terminal protein sequencing.
Recombinant IRP-1A and IRP-1B Have Aconitase Activity-In silico analysis of IRP-1A and IRP-1B has shown that all amino acids known to participate in the catalytic center of the aconitase enzymes were conserved (17), suggesting that both proteins should be active as aconitases. We tested this hypothesis and confirmed that both IRP-1A and IRP-1B exhibit aconitase activity as purified ( Fig. 2A). The bacteria were able to produce a fraction of recombinant IRPs with an intact iron-sulfur cluster, and aconitase activity was in the same range as purified recombinant human IRP1 expressed in a similar prokaryotic system (36,37). Thus, D. melanogaster is the first organism shown to contain two active cytosolic aconitases.
Only IRP-1A Binds to the IRE-Next we asked whether both IRPs can bind to an IRE. Purified recombinant IRP-1A was able to bind in vitro both Drosophila ferritin IRE and human ferritin IRE in the presence of a reducing agent (Figs. 2B and 4). In contrast, IRP-1B did not show any affinity toward either Drosophila or human ferritin IRE. No IRE binding activity of IRP-1B was detected even when the protein was added in high molar excess or pretreated by different reducing agents, including 2-ME, cysteine, and dithiothreitol (Figs. 2, 4, and 5, and data not shown).
To test for the binding specificity of IRP-1A to IRE, 100-fold molar excess of non-radioactive IRE was added and allowed to compete for binding. The results showed that cold IRE was able to reduce the binding, whereas the same amount of non-functional IRE did not affect the IRP-1A/IRE radioactive signal, thus demonstrating an IRE-specific binding activity for IRP-1A (Fig. 3). This result combined with previous information (24 -27, 31) suggests that the IRE binding activity of Drosophila extracts should be attributed exclusively to IRP-1A.
IRP-1A Acquisition of IRE Binding Activity Correlates with Loss of Aconitase Activity-IRP-1A as purified from bacteria has aconitase activity and only binds to IRE upon incubation with reductant (either 2-Me in Figs. 2, 3, and 5 or cysteine in Fig. 4A) by a mechanism that remains to date unclear (38,39). Incubation with cysteine in the presence of iron has no effect on the purified enzyme (Fig. 4). The protective property of iron with respect to cluster disassembly is shown by further addition of EDTA, which again results in loss of aconitase activity and concomitant gain in IRE binding activity, all in agreement with previous work on mammalian IRP1 (26,40). In contrast, no IRE binding activity was detected in similar experiments with purified IRP-1B, but aconitase activity of IRP-1B was also dependent on the presence of iron and the absence of highly reducing conditions (Fig. 4B).
Domain-swap Experiments-We constructed chimeric IRPs to identify the regions of IRP-1A that contain the critical residues for IRE binding activity. Different fragments of IRP-1B were swapped with fragments from IRP-1A ( Fig. 5A and Table 1). Five different chimeras (His-IRP-A to His-IRP-E) were produced with N-terminal His tag in E. coli and purified as previously described for His-IRP1A and His-IRP-1B. Equal amounts (250 ng) of the purified recombinant proteins were then subjected to EMSA, and their IRE binding activity was compared with the activity of His-IRP-1A, which was set to 100% (Fig. 5B). The chimeras His-IRP-B, His-IRP-C, and His-IRP-E were able to bind both human and Drosophila ferritin IRE with an activity of 26 Ϯ 6, 6 Ϯ 2, and 24 Ϯ 5%, respectively. The chimeras His-IRP-A and His-IRP-D did not show binding activity even when 2.5 g of purified protein was loaded on the gels (data not shown).
Overexpression of IRP-1A and IRP-1B in Vivo-To confirm the findings we obtained in vitro, we also generated transgenic flies in which temporal and tissue-specific expression of the endogenous IRPs was achieved by means of the UAS/Gal4 system (41). Overexpression of IRP-1A in a single tissue (the fat bodies) resulted in a marked increase of IRE binding activity of whole-fly extracts, whereas no similar effect was seen when IRP-1B was overexpressed (Fig. 6A). At the same time, total aconitase activity of these extracts was increased by 10% in the case of IRP-1A overexpression ( p, 0.10) and by 20% when IRP-1B was overexpressed ( p, 0.027). Simultaneous overexpression of both proteins resulted in flies with high IRE binding and high aconitase activities (Fig. 6B). Separation of cellular aconitase activities by electrophoresis (10,31) confirmed that the modest increase in total activity was due to specific increase from the overexpressed cytosolic aconitases (Fig. 6C). We confirmed that the two proteins were overexpressed at similar levels by Western blotting using cross-species-reacting rabbit antisera raised against mouse IRP1 (Fig. 6D).
We have expressed IRP-1A and IRP-1B with a variety of Gal4 drivers, chosen on the basis of their differential tissue-specific expression patterns and scored the flies for phenotypes (Table 2). Remarkably, ubiquitous overexpression of IRP-1A results in pre-adult lethality. In contrast, flies that ubiquitously overexpress IRP-1B are viable and show no obvious phenotypes. This result shows that the biochemical differences between the two proteins can have vital significance. We asked whether we could reproduce the lethal phenotype of ubiquitous IRP-1A overex-   JULY 7, 2006 • VOLUME 281 • NUMBER 27 pression by targeted expression of IRP-1A to a single tissue. Although high expression of IRP-1A in brain, motor neurons, fat bodies, midgut, oenocytes, epidermis, and imaginal disks did not affect viability, expression in muscle was detrimental to development with no flies ever surviving into adulthood or to the third instar larval stage.  Table 1. B, Coomassie-stained gel after SDS-PAGE of 1 g of each purified recombinant protein (His-IRP-1A, His-IRP-1B, His-IRP-A to His-IRP-E). EMSA was used to test the IRE binding activity of the different chimeras. 250 ng of protein was reduced by 2% 2-Me in a total volume of 12 l before addition of radioactive human ferritin IRE. The IRE binding activity was compared with the activity of His-IRP-1A, which was set to 100%. The experiment was repeated three times with human ferritin IRE and three times with Drosophila ferritin IRE. As no difference was observed between the two IREs, the results were pooled and quantified by using the software Multi Gauge v2.2 (Fuji Film). ND, not detected. 1% of the mixture was incubated with radioactive Drosophila ferritin IRE, and IRE binding activity was monitored by EMSA. The rest of the mixture was assayed for aconitase activity in a coupled aconitase/isocitrate dehydrogenase reaction. The activity is expressed as the change in A 340 nm per h and g of recombinant protein. No IRE binding activity was detected for His-IRP-1B.

DISCUSSION
This report describes the biochemical properties of two highly homologous enzymes that belong to the family of cytosolic aconitases or IRPs, encoded by different loci in the D. melanogaster genome. Despite 86% identity and 93% similarity between the two proteins (17), only IRP-1A is functionally similar to the mammalian IRP1, as it alternates between a cytosolic aconitase or an IRE-binding protein, probably through the assembly and disassembly of a cubane iron-sulfur cluster cofactor. Conversely, IRP-1B is more similar to the cytosolic aconitase of C. elegans, which is evolutionarily more ancient and lacks high affinity for IREs. We have shown that in vivo overexpression of IRP-1A and IRP-1B can result in profoundly different phenotypes, suggesting that the IRE binding property of IRP-1A influences cellular metabolism in Drosophila and that IRP-1A represents a genuine evolutionary molecular adaptation.
Efforts from a number of laboratories to identify domains of IRP1 that are responsible for IRE binding have been summarized by Dupuy et al. (42). These authors have modeled the docking of the IRE to IRP1, showing that most of the cleft between domains 1-3 and domain 4 is occupied by the IRE. The notion that many different contact points between the IRE and the IRP1 are required for binding and that these points are scattered in different domains of the IRP1 is consistent with our domain-swap experiments, which suggest that replacement of extended portions of IRP-1B with the corresponding IRP-1A sequences provides only partial IRE binding activity to the resulting IRP hybrid. Furthermore, such large domains can span very different portions of the IRP-1A protein (Fig. 5).
We took advantage of the recent annotation of the genome sequence of Drosophila pseudoobscura (43) and identified IRP-1A and IRP-1B homologs in this species. We then hypothesized 1) that the two proteins have the same biochemical properties as in D. melanogaster, 2) that the two IRP-1As should share residues important for IRE binding, and 3) that these residues should differ in the two IRP-1Bs. By applying the last two criteria to sequence comparisons of the two IRP-1As and the two IRP-1Bs, we identified 30 amino acids scattered in all four domains of FIGURE 6. IRE binding and aconitase activity of extracts from flies overexpressing IRPs. A, 8 g of total protein from whole-fly extracts of the genotypes indicated were incubated with radioactive human ferritin IRE and analyzed by EMSA. B, specific aconitase activities were calculated (per g of total protein) for the extracts. Experiments were replicated with three individual biological samples. Overexpression of IRP-1B (alone or in conjunction with IRP-1A) in fat bodies increased total aconitase activity of the flies by 20%. C, in-gel aconitase activities (10) indicate that the modest increase in total aconitase activity is due to a specific increase of cytosolic aconitase activity. D, Western blots of the fly extracts probed with antibodies against mouse IRP1 and mouse anti-␣-tubulin (7) show that IRP-1A and IPR-1B were overexpressed to similar levels.   (Table 3). Interestingly, many of these amino acids overlap or are in close proximity with regions that have been previously implicated in IRE binding on human IRP1 (44 -46) ( Table 3). We have also labeled on the human IRP1 structure the homologous amino acids that differ between IRP-1A and IRP-1B (Fig. 7A). Based on current knowledge (42), most of these residues are not facing the predicted RNA binding surfaces. Possible exceptions are the amino acids Asn 735 and His 412 (Fig. 7A, arrows). A negatively charged glutamate (Glu 745 ) in IRP-1B instead of Asn 735 (Ser 748 in IRP-1A) may repel the similarly charged IRE. In addition, a bulky hydrophobic phenylalanine (Phe 421 ) in IRP-1B, which replaces His 412 (Ser 424 in IRP-1A), could also obstruct binding. Moreover, our analysis suggests a previously unidentified hot spot defined by a stretch of amino acids, 341-350, carrying five substitutions between IRP1-A and IRP-1B (Table 3) and clustering in close proximity on the three-dimensional structure (Fig. 7A, circled  area). The significance of this finding is not known. IRP-1A and IRP-1B share, respectively, 78 and 79% similarity with human IRP1 at the protein level. We have generated three-dimensional models of IRP-1A and IRP-1B based on the recently solved structure of human IRP1 (42), using the Swiss Model server (47). Superimposition of the models with IRP1 revealed that the conserved residues involved in aconitase activity (17), including three arginines that were also implicated in IRE binding (48), perfectly overlap with the respective amino acids in the human IRP1 (Fig. 7B, asterisks). Therefore, we assume that both Drosophila IRPs have an almost identical core structure to human IRP1. Notable differences between IRP-1A and IRP-1B are only found on the external surfaces of the protein (Fig. 7B). Both IRP-1A and IRP-1B contain an extension of a loop in domain 1 (Gln 73 -Asp 77 in IRP-1B) as compared with the human IRP1. IRE binding is thought to involve a conformational change that causes an outward movement of domain 4 from domains 1-3 (42). This movement might be hampered in IRP-1B because the extended loop of domain 1 points toward domain 4 (boxed in Fig. 7B), whereas in IRP-1A this loop is directed outwards. Interestingly, a minor structural alteration of IRP-1B (Phe 875 -Glu 878 ) is observed in domain 4 in close proximity to the above-mentioned loop. In contrast, IRP-1A perfectly aligns with the human IRP1 structure in this region. It seems possible that these differences provide more space for IRP-1A and human IRP1 to "breathe" during IRE binding. The other two differences predicted by modeling to differ between IRP-1A and FIGURE 7. Differences between the Drosophila IRPs modeled on the human IRP1 crystal structure (Protein Data Bank accession code 2B3X). A, residues highlighted in red are conserved between D. melanogaster IRP-1A and Drosophila pseudoobscura IRP-1A but differ from their respective IRP-1B proteins (see also Table 3). The iron-sulfur cluster is shown in yellow. Arrows point to residues at the presumed entrance site of the cleft, where the IRE binds. No differences were found within the cleft. Amino acids 341-350 also accumulated changes between IRP-1A and IRP-1B (circle). B, overlay of IRP-1A and IRP-1B models on the human IRP1 structure. The three asterisks depict Arg 536 , Arg 541 , and Arg 780 , implicated in aconitase activity and IRE binding (48). Major structural differences between all three proteins predicted by modeling are color-coded black for hIRP1, red for IRP-1A, and blue for IRP-1B and are described under "Discussion." The boxed area is likely to be affected in putative movements of domain 4. Visualization used the Deep View spdbv 3.7 software (47).

Evolution of the IRP/IRE System
IRP-1B (Asp 197 -Ser 202 and Ser 852 -Leu 854 in IRP-1B) are located farther away from the hinge linker. Further experimental work is required to unravel the set of amino acid residues in IRP-1A that transform it into an IRE-binding protein in contrast to its "sister" IRP-1B. All present evidence suggests that several domains of the protein will be implicated in this transformation.
During eukaryotic evolution, enzymes that function in iron metabolism and redox reactions were required in more than one subcellular compartment (31, 49 -51). Information for the targeting of proteins to the proper location is usually contained in signal sequences at the N terminus, and alternative splicing can change the N termini encoded by single genes. Alternatively, duplication of genes and subsequent divergence allow targeting of proteins of similar function to different subcellular compartments, offering the possibility for independent evolution through acquisition of new mutations and molecular adaptations (52,53). Evolution of the IRP/IRE system was likely facilitated by a series of gene duplications that initially allowed cytosolic aconitase to evolve independently from mitochondrial aconitase, and a subsequent duplication of cytosolic aconitase produced two distinct cytosolic aconitases in insects. One of the two cytosolic aconitases was then free to specialize in binding of the IRE.
A common property of mRNA is to form stem-loop structures, as exemplified by the IRE. The question is how such inherently unstable structures become stabilized in evolution, because the non-coding mRNA sequences in the 5Ј-and 3Ј-untranslated regions are not constrained by the need to produce a functional protein. However, interaction of a foreign protein with a newly formed stem-loop structure of an mRNA could result in an overall positive effect for cellular functions and organism survival. Such a mechanism would produce adaptations for natural selection and offer a mechanism to explain how the IRP/IRE system evolved.
We speculate that the first example of an IRE-containing gene during evolution is either succinate dehydrogenase subunit B or Fer1HCH. Given that 1) aconitases catalyze the interconversion of citrate to isocitrate and thus require the presence of substrate to function, 2) aconitase function in the citric acid cycle is more ancient than a direct role of the protein in iron metabolism, 3) eukaryotic cells need to regulate intermediary metabolism both in the mitochondria and in the cytosol, utilizing citrate and isocitrate for different purposes in each compartment, 4) IRP1 control of the citric acid enzyme mitochondrial aconitase is present in mammalian cells (54,55), and 5) the IRE of Fer1HCH is contained in an alternatively spliced exon, enabling a bypass of IRP-1Adependent regulation of ferritin (29,30), we speculate that the initial IRE evolved in succinate dehydrogenase subunit B. After generation by gene duplication of two cytosolic aconitases, one of them acquired the ability to bind to the succinate dehydrogenase subunit B transcripts, which would down-regulate the citric acid cycle and secure higher citrate availability for the cytosol. Gaining the capacity to regulate the compartmentalization of citrate (used for the production of ATP in mitochondria or for the production of glutamate or lipids in the cytosol) could operate as the evolutionary driving force. Whether enzyme competition for the same substrate has a role in this process is at present unclear.
Because the control of iron-sulfur cluster stability is an iron-and oxygen-dependent process (8,10), the generation of a cytosolic molecular switch that could regulate translation could be further developed to control the flux of iron in the system. In this respect, it is intriguing that the IRE in Fer1HCH is found in a separate exon (29,30), as if it moved "appropriately" to a site with biological significance. Ferritin is the major iron buffer in cells and thus a very good initial target for iron regulation.
Subsequent evolutionary molecular events included an insertion of several amino acids in one of the cytosolic aconitases (defined as IRP2) and disruption of its activity as an aconitase by divergence of critical amino acid residues important for catalytic activity. Danio rerio, a nonmammalian vertebrate, is the lowest organism in evolutionary terms that is known to express IRP2 (56). Development of elaborate posttranslational control of IRP2 by oxygen and iron (3,57) transformed it into the dominant player in mammalian iron metabolism (7,8).