Ligand-induced Structural Alterations in Human Iron Regulatory Protein-1 Revealed by Protein Footprinting*

Human iron regulatory protein-1 (IRP-1) is a bifunctional protein that regulates iron metabolism by binding to mRNAs encoding proteins involved in iron uptake, storage, and utilization. Intracellular iron accumulation regulates IRP-1 function by promoting the assembly of an iron-sulfur cluster, conferring aconitase activity to IRP-1, and hindering RNA binding. Using protein footprinting, we have studied the structure of the two functional forms of IRP-1 and have mapped the surface of the iron-responsive element (IRE) binding site. Binding of the ferritin IRE or of the minimal regulatory region of transferrin receptor mRNA induced strong protections against proteolysis in the region spanning amino acids 80 to 187, which are located in the putative cleft thought to be involved in RNA binding. In addition, IRE-induced protections were also found in the C-terminal domain at Arg-721 and Arg-728. These data implicate a bipartite IRE binding site located in the putative cleft of IRP-1. The aconitase form of IRP-1 adopts a more compact structure because strong reductions of cleavage were detected in two defined areas encompassing residues 149 to 187 and 721 to 735. Thus both ligands of apo-IRP-1, the IRE and the 4Fe-4S cluster, induce distinct but overlapping alterations in protease accessibility. These data provide evidences for structural changes in IRP-1 upon cluster formation that affect the accessibility of residues constituting the RNA binding site.

Intracellular iron homeostasis in mammals and many other eukaryotes is controlled at the post-transcriptional level by two related cytoplasmic proteins, the iron regulatory proteins (IRP-1 and IRP-2) 1 (for reviews see Refs. 1 and 2). These two proteins bind to a conserved RNA stem-loop structure, the so-called iron-responsive element. An increasing number of mRNAs have been shown to carry one IRE motif in the 5Јuntranslated regions of mRNAs encoding ferritin (L-and Hchains), erythroid 5-aminolevulinate synthase, mitochondrial aconitase, and succinate dehydrogenase, as well as in the 3Ј-untranslated region of transferrin receptor (TfR) and Nramp2 mRNAs, thus modulating their expression according to iron availability (3)(4)(5)(6)(7)(8)(9)(10)(11)(12). For instance, the interaction of IRP with a single IRE in the 5Ј-untranslated region of ferritin mRNA inhibits translation by preventing ribosome binding (13)(14)(15). On the other hand, IRP binding to the five IREs present in the 3Ј-untranslated region of TfR mRNA stabilizes the message against degradation (10,16). Therefore, the specific interaction between IRP and mRNA ensures a coordinated regulation of the expression of proteins involved in the uptake (TfR), storage (ferritin), and utilization (erythroid 5-aminolevulinate synthase) of iron (2). Because the IRPs appear to regulate two enzymes of the citric acid cycle, they may also play an important role in mediating iron regulation of mitochondrial energy production (9,10).
IRP-1 and IRP-2 are affected by iron in different manners. Whereas iron increases the rate of degradation of IRP-2 (17,18), it inhibits the ability of IRP-1 to bind to IREs through formation of an iron-sulfur cluster. Interestingly, the 4Fe-4S cluster assembly confers aconitase activity to IRP-1 (19 -24). IRP-1 can therefore directly sense and respond to perturbations of iron levels and redox potential. Moreover, phosphorylation of IRP-1 may also be involved in the regulation of its two functions because it preferentially occurs in the apoprotein (25,26). Unlike other enzymes that recognize mRNA (1), the functional implication of the cytoplasmic aconitase activity of IRP-1 is not well understood. However, in addition to the functional similarities, sequence homologies between IRP-1 and the mitochondrial aconitase suggest that these proteins may adopt similar structures (23,27). In particular, a striking level of conservation exists for the amino acids located in the active site of mitochondrial aconitase. The crystal structure of mitochondrial aconitase has revealed that domains 1-3 are connected by a linker region to domain 4 (28), the iron-sulfur cluster being located in a cleft formed between domains 1 and 3 and domain 4. Furthermore, mutational analysis (29,30) and cross-linking data (31)(32)(33) have indicated that multiple contacts occur between IRP-1 and the IRE hairpin on both sides of the putative cleft. It is interesting to note that some arginine residues required for the aconitase activity are also critical for IRE binding (29,30). These data suggest that the RNA binding site and the catalytic center are overlapping and provide a rational explanation for the mutually exclusive functions of IRP-1. Based on the crystallographic structure of mitochondrial aconitase, it was also proposed that significant rearrangement of the IRP-1 may occur to permit the access of the IRE (for a review see Ref. 34).
In the present paper, we used a protein footprinting approach to compare the structure of human IRP-1 in its two functional states, the RNA-binding form (apoprotein) and the aconitase form (holoprotein). The results demonstrated structural changes in IRP-1 as a function of the presence or the absence of the iron-sulfur cluster. We also identified single amino acids of IRP-1 that are protected by either the minimal regulatory region of TfR mRNA (which contains three IREs) or of the ferritin IRE upon binding that are located in two defined distant regions of IRP-1 on both sides of the putative cleft. Interestingly, these two regions also become less accessible toward proteases in the holoprotein. Together our data provide direct experimental evidence for the structural rearrangements of IRP-1 that are assumed in current models (for a review see Ref. 2).

EXPERIMENTAL PROCEDURES
RNAs and IRP-1 Preparations-TfR (TRS-G165) and ferritin mRNAs were synthesized by in vitro transcription with T7 RNA polymerase from plasmids linearized with BamHI. TfR mRNA was transcribed with T7 RNA polymerase from polymerase chain reaction-generated DNA templates (35). RNAs were 5Ј-end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase (36). The RNAs were purified on an 8% polyacrylamide, 8 M urea gel electrophoresis, eluted overnight, and precipitated with ethanol. Prior to use, RNAs were first incubated at 90°C for 1 min in water and renatured by a slow cooling at 25°C in the appropriate buffer.
Recombinant human IRP-1 was purified from the Escherichia coli overproducing strain K12 pQE9-his-hIRF. The protein was isolated by affinity chromatography on Ni 2ϩ -NTA beads (Qiagen) as described previously (37). The protein was eluted with 50 mM imidazole and dialyzed against 20 mM Tris-HCl, pH 7.5, 1.5 mM MgCl 2 , 40 mM KCl, 5 mM 2-mercaptoethanol (Buffer A). The protein was then purified on a Mono-Q column equilibrated in Buffer A and eluted from the column with a salt gradient (40 mM to 1 M KCl). IRP-1 was then concentrated using a Centricon 30K (Amicon), washed several times in 20 mM Tris-HCl, pH 7.5, 1.5 mM MgCl 2 , 5 mM 2-mercaptoethanol, 5% glycerol (Buffer B), and stored at Ϫ80°C in small aliquots.
Reconstitution of the [4Fe-4S] cluster in the affinity-purified human IRP-1 (400 g) was performed in 250 l of buffer containing 50 mM Hepes-NaOH, pH 7.6, 150 mM potassium acetate, 1.5 mM MgCl 2 , 5% glycerol (Buffer N) in the presence of 3 mM FeSO 4 , 6 mM Na 2 S for 15 min at 37°C (38). The cytoplasmic aconitase form of IRP-1 was purified on a TSK2000 gel filtration column (fast protein liquid chromatography) equilibrated in Buffer B. Then the protein was concentrated using a Centricon 30K. The aconitase activity was determined with the method of Rose and O'Connell (39) by following the formation of NADPH at 340 nm as a function of time in a coupled aconitase/isocitrate dehydrogenase reaction. The reconstituted aconitase activity was found to be 40 units/mg.
Gel Retardation Assays-Binding reactions containing 5Ј-end-labeled IRE and increasing concentrations of either IRP-1 or of the aconitase form of IRP-1 were incubated at 4°C for 20 min in buffer N. RNA-protein complex formation was analyzed by a 8% non-denaturing polyacrylamide gel as described previously (40).
Protein Footprinting-IRP-1 protein or its aconitase form (4 M) either alone or in the presence of RNA (TRS-G165 or IRE) were subjected to partial proteolytic digestions using different proteinases (Roche Molecular Biochemicals). Recombinant IRP-1 was first treated with 2% 2-mercaptoethanol. Then, RNA binding was assayed in the presence of a 2-fold excess of TRS-1 or ferritin IRE at 4°C for 20 min in 50 mM Hepes-NaOH, pH 7.6, 150 mM potassium acetate, 1.5 mM MgCl 2 , and 5 mM 2-mercaptoethanol. Proteolytic digestions were performed in 10 l with Arg-C (0.02 unit), Asp-N (0.004 g), Glu-C (0.1 g), Lys-C (0.02 U), or trypsin (0.2 g) at 37°C for 30 min. The reactions were stopped by the addition of 0.5 volume of 2ϫ sample buffer containing 50 mM Tris-HCl, pH 6.8, 8% SDS, 25% glycerol, 4% 2-mercaptoethanol, 0.02% bromphenol blue. The samples were denatured at 95°C for 2 min before running on 5% stacking/15% separating polyacrylamide-SDS gels. To get optimal resolution of small peptides, the cleavage products were resolved using 10% stacking/16.5% separation polyacrylamide-Tricine-SDS gel electrophoresis. Prestained rainbow proteins (from 90 kDa to 7.7 kDa) were used as internal markers (Bio-Rad). These experiments were repeated five times with different protein purifications. The main cleavages were reproducibly found in all of the experiments. However, the intensity of Asp-N specific cleavages varied in some of the experiments depending on the batch of protease used.
Analysis of the Proteolytic Cleavages-After electrophoresis, the generated peptides were transferred from SDS gels to polyvinylidene di-fluoride membranes (Hybond-P, Amersham Pharmacia Biotech). The transfer was performed at 1 mA/cm 2 in 25 mM Tris-HCl, pH 8.3, 0.2 M glycine, and 20% methanol for 1 h at 20°C using a semi-dry blotting apparatus. The membrane was then blocked for 15 min at 20°C in 20 mM Tris-HCl, pH 7.6, 137 mM NaCl, 0.1% Nonidet P-40 (TBST buffer) with 10% milk powder. After washing three times in TBST buffer, the membrane was incubated at 4°C for at least 2 h with a 1:3300 dilution of RGS-His mouse antibody (Qiagen) in TBS buffer. After three washes with TBST, the membrane was incubated for 1 h at 20°C with 1:1000 dilution of alkaline phosphatase-conjugated secondary antibody (antimouse FAB/POD, Roche Molecular Biochemicals). After three washing steps with buffer TBST, the detection was done using an ECL Western blotting kit (Amersham Pharmacia Biotech). The bands were then revealed after autoradiography of the membrane (5 s to 1 min).

Assignment of the Proteolytic Cleavages in IRP-1-
We probed the accessibility of recombinant human IRP-1 (37) using a set of endoproteases that cleave after basic and acidic residues. The accessibility of lysine residues was probed with Lys-C and trypsin, arginines with Arg-C and trypsin, aspartic acid with Asp-N and Glu-C, and glutamic acid with Glu-C. These proteinases were used under native conditions that are optimal for the stability of the RNA-IRP-1 complexes. We also verified that none of these proteinases contain RNase activity. Cleavage conditions (time scale and concentration dependences) were adjusted so that less than 50% of the input protein was cleaved and thus that multiple cleavage of the protein was unlikely. All the cuts that were considered appeared with similar kinetics, at low protease concentration, and were reproducibly found. Under these conditions of limited proteolysis, we favor detection of only primary cleavages that reflect the accessibility of amino acids within the native protein structure. We used a strategy derived from Heyduk and Heyduk (41), which allows detection of the cleaved peptide by immunostaining with antibodies specific to the N or C terminus of protein. In our study, a recombinant IRP-1 protein carrying a RGS-His-tag at its N terminus was expressed in E. coli and purified by affinity chromatography (37). It was previously shown that recombinant IRP-1 efficiently binds to the ferritin IRE and can be converted to an aconitase in vitro (37). This implies that the His-tag at the N terminus of IRP-1 does not interfere with IRE binding and Fe-S cluster formation, respectively. After limited proteolysis of IRP-1, the resulting peptides were separated according to size on SDS-polyacrylamide gel electrophoresis (PAGE) and then electroblotted to a polyvinylidene difluoride membrane. The full-length protein and the RGS-His-tag peptides were revealed by using antibodies specific to the RGS-His-tag (Fig. 1A). The mobility of the bands on SDS-PAGE is directly proportional to the distance of the cleavage site from the N terminus of the protein. Thus, by comparing the electrophoretic mobility of appropriate protein standards and the cleavage products generated by proteinases with different amino acid specificities, identification of the cleavage sites can be accomplished. The positions of the cuts were determined by plotting the logarithm of the molecular masses of the assigned fragments against the migration distances in the gel because an almost linear correlation for polypeptide masses above 5 kDa has been described (41,42). An example shown in Fig. 1B revealed that the experimental values are consistent with theoretical predictions and that the gel resolution allows assignments at the level of single amino acids. No information was obtained for peptides shorter than 8 kDa due to the low efficiency of transfer of small peptides onto polyvinylidene difluoride membranes. The experiments have been repeated five times with three different preparations of IRP-1 and showed reproducible proteolytic cleavages. All the cleavage sites in IRP-1 that were identified consistently in all the experiments are listed in Table I.
A large part of IRP-1 is not sensitive to proteolysis, consistent with previous results (25), especially the region encompassing residues 190 to 380. The proteases used here cleave residues that are exposed to the surface of IRP-1 and hence to the solvent because they are sensitive to steric hindrance due to their relatively large molecular sizes (23.5 kDa for trypsin, 27 kDa for Glu-C and Asp-N, 33 kDa for Lys-C, and 59 kDa for Arg-C). Most of the cleavages are located in three defined areas comprising residues 75 to 160, 375 to 400, and 720 to 755. In the N-terminal domain, several fragments with molecular masses around 19 kDa were produced by all the proteases. Taking into account the relative mobility of these bands and specific markers, the cleavage specificity of the respective proteases and the potential target in IRP-1 sequence, it was pos-sible to assign the cleavage sites: two trypsin-and Arg-Cspecific bands were assigned to Arg-149 and Arg-135 (Fig. 1). The strong Asp-N-specific cleavage most likely occurred before Asp-137 because the band migrated at the same position as the trypsin-and Arg-C-specific band at Arg-135. The Lys-C-and trypsin-specific band that migrated between Arg-149 and Arg-135, was attributed to Lys-141. Glu-C induces two main cleavages in this area, one site occurs just below the band corresponding to Arg-149 and was identified as Glu-148. The distance between the two Glu-C-specific bands favors position Glu-155 as the potential second cleavage site (Fig. 1, A and B). were identified after short electrophoretic migrations on SDS-Tricine gels. Conversely, the strong cleavages induced by trypsin, Arg-C, and Glu-C in the C-terminal domain of IRP-1 were assigned on long run SDS-PAGE gels. The major Glu-C-specific cut migrated with a molecular mass around 70 kDa (Fig. 1) and was assigned by sequencing the corresponding large and small fragments using automated Edman degradation (results not shown). The large fragment corresponded to the N terminus of IRP-1, and the N-terminal sequence of the 30-kDa fragment was SWNALA, indicating that the major Glu-C-specific cut occurred at Glu-621. This cut was also previously reported (20). We used the same strategy to identify the three strong Arg-Cspecific cleavages in the N-terminal domain. The resulting fragments migrate slower than fragment Glu-621, suggesting a cleavage at Arg-721, Arg-728, and Arg-732. This was again confirmed by sequencing the two major corresponding C-terminal fragments using Edman degradation, which gave the sequences GTFA and LLNRFL.
Probing the IRE Binding Site on IRP-1-To identify structural changes of IRP-1 induced by IRE binding, we used the footprinting approach to map the IRE binding site in IRP-1. This also provided information on potential amino acids involved in the RNA binding (42,43). The RNA-induced effects on IRP-1 structure were compared using as RNAs the IRE of the ferritin mRNA and the regulatory region of the human TfR mRNA, which contains three of the IRE elements (TRS-G165) (see Fig. 2C). IRP-1 was probed with several proteases (trypsin  Table I). One discrepancy was observed at Arg-149, which was strongly protected by the IRE hairpin toward Arg-C and was still accessible toward trypsin. This result may be explained by the fact that Arg-C has a higher molecular weight and therefore is more sensitive to steric hindrance (Figs. 1A and 2). Other significant protections were observed in the C-terminal domain at Arg-721 and to a lesser extent at Arg-728 and Arg-732 upon IRE binding ( Table  I). Binding of the minimal regulatory region of TfR mRNA (TRS-G165) induced the same changes in IRP-1 as IRE of the ferritin mRNA (Fig. 2B), indicating that binding of TRS-G165 RNA is restricted to the IRE motif. The addition of the same amount of tRNA did not alter the cleavage pattern, demonstrating that the changes in IRP-1 induced by IRE or TRS-G165 binding were specific.
Structural Changes of IRP-1 Induced by Iron-Sulfur Cluster Assembly-It was previously proposed that the aconitase form of IRP-1 may display structural differences compared with apo-IRP-1 because the holoprotein does not efficiently bind to IREs (for a review, see Ref. 34). A recent study indeed showed that the holoprotein appears to be more resistant than the RNA binding form of IRP-1 toward limited proteolysis (25). Therefore, we mapped the structural changes that occur in IRP-1 depending on the iron-sulfur cluster status. It was previously shown that the iron-induced reduction of IRE binding activity in cells could be reconstituted by treatment of IRP-1 with an excess of FeSO 4 in the presence of cysteine (37). First we have verified that reconstitution of the iron-sulfur cluster in IRP-1 conferred aconitase activity to the protein using a coupled aconitase/isocitrate dehydrogenase reaction. Furthermore, we also showed that the holoprotein did not bind to IRE hairpin using non-denaturing gel electrophoresis (data not shown). This is further supported by the fact that ferritin IRE or TRS-  -1 (ACN), and effect of RNA binding (RNA) The intensity of the cleavages in IRP-1 is shown as follows: ϩϩϩ, strong; ϩϩ, medium; and ϩ, low cut. Effect of iron-sulfur cluster or RNA binding: strong protection (ϪϪ), moderate protection (Ϫ), enhancement (ϩ), unchanged (no symbol). The asterisks indicate the positions of the Asp-N cleavages for which their intensities varied in some of the experiments. The amino acids are numbered according to the N terminus of human IRP-1 and do not take into account the His-tag residues. Trypsin G165 RNAs do not affect the proteolytic cleavage pattern of the aconitase form (Fig. 2B). The holoprotein was then subjected to probing by proteases in the presence or absence of the ferritin IRE or TRS-G165 RNA (Fig. 2). These experiments were repeated three times with three different preparations of the aconitase form of IRP-1. Several reproducible protections induced by Fe-S cluster formation were detected in two distant regions, indicating that the holoprotein was more resistant than the IRE binding form of IRP-1 toward proteases, consistent with previous findings (25). Strong protections occurred in the N-terminal domain at Arg-149 toward Arg-C, at Glu-155 toward Glu-C, and at Arg-187 toward trypsin, and in the Cterminal domain at Arg-721, Arg-728, Arg-732 toward trypsin and Arg-C, at Lys-736 toward Lys-C, and at Asp-751 toward Asp-N (Fig. 2, Table I). The protected residues were located in two regions far away from the [Fe-S] cluster coordination sites and most likely reflect conformational changes of IRP-1. Interestingly all these residues were also found protected in IRP-1 upon IRE binding. These data suggest that [Fe-S] cluster formation induces conformational changes of IRP-1 that mask the IRE binding site.

DISCUSSION
For the present work, we used protein footprinting to follow the structural changes of IRP-1 in different functional states, the RNA-binding form (apoprotein), and the aconitase form (holoprotein). Five different proteinases were used to probe the IRP-1 structure under native conditions. The peptide cleavage products were subsequently resolved on SDS-PAGE gels and readily identified by comparing the different protease digestions and the relative mobility of the generated peptides using appropriate internal controls (Fig. 1, A and B). Our data first demonstrated that binding of the minimal regulatory region of TfR mRNA (TRS-G165) or of the ferritin IRE induces the same protections in IRP-1, indicating that binding of TRS-G165 is restricted to the IRE hairpin. The RNA effect was restricted to the N-terminal domain of IRP-1 encompassing residues 80 to 187 and to the region 721 to 732 at several arginine residues. In the N-terminal domain, two sets of protected residues were found separated by several amino acids which were still accessible to proteases (Fig. 1C). The protected residues are located in a region that is highly conserved in IRPs. However an  Fig. 1. B, IRP-1 protein (apoprotein) or the cytoplasmic aconitase form (holoprotein) was digested with several proteinases (trypsin, Arg-C, and Glu-C) in the presence of TRS-1 (ϩ) or an equivalent amount of tRNA (Ϫ). Experimental details are given under "Experimental Procedures." C, secondary structure of the regulatory region of TfR mRNA (TRS-G165) and of the ferritin IRE hairpin. The minimal regulatory region of TfR mRNA is identical to the sequence defined by Casey et al. (10) except that IRE C contains a G165 instead of an adenine, which increased significantly the affinity binding for IRP-1 (35). The three IREs B to D are denoted according to Casey et al. (10). The protection induced by IRP-1 binding are shown in the inset: protected bases are circled and protected riboses are represented by dots. The results are taken from Schlegl et al. (35).
insertion of 73 amino acids in this domain found only in IRP-2 confers iron-mediated degradation (17). Interestingly, a mutant IRP-1 containing this insertion domain of IRP-2 bound RNA but prevented aconitase activity (44). It is interesting to note that this domain has been inserted in a region that is still accessible toward proteases upon RNA binding (Fig. 1C). Some of these cuts were even enhanced upon IRE binding indicating that conformational adjustments of IRP-1 may occur. The Nterminal domain of IRP-1 is also of particular interest because cross-links were identified between IRP-1 and residues 116 and 151 (31)(32)(33), and one UV-cross-linked site was postulated at Ser-127 (33). The protected residue Asp-125 lies exactly in this area. This region is highly conserved among IRPs (Fig. 1C, see Refs. 45 and 52-57) and contained two conserved residues, Asp-125 and His-126, essential for the aconitase activity (29,45). Therefore, it is tempting to propose that the protections reflect direct contacts between the IRE and IRP-1, masking the catalytic site. Interestingly, one of the phosphorylation sites was located at Ser-138 (25) close to the RNA binding site but in a region that remains accessible to proteases (Fig. 1C). This may explain why phosphorylation occurs preferentially in the apoprotein. Other regions have also been shown to contribute to RNA binding. Cross-links were identified in a region encompassing residues 480 to 623 (31), and the deletion of the last 132 amino acids of IRP-1 results in a loss of IRE binding (20). It is of interest that significant protections were detected at Arg-721 and to a lesser extent at Arg-728 and Arg-732. Interestingly, the second phosphorylation site was detected in the same area at Ser-711 (25). These protections are also located in a region where an insertion occurs relative to the sequence of mitochondrial aconitase (23,27) whereas this region is highly conserved in other IRPs (Fig. 1C, see Refs. 45 and 52-57). Also, Arg-699 participates directly in the active center of aconitase because it binds citrate, showing again the close proximity of the two functional sites (29).
Based on sequence alignments between mitochondrial aconitase and IRP-1, we represented all the changes induced either by RNA or by iron-sulfur binding on the tertiary structure of mitochondrial aconitase (Fig. 3). Interestingly, most of these protected residues map in the putative cleft across domains 1 to 3 and in domain 4. We and others previously showed that IRP-1 covers one helical turn of the IRE and makes specific contacts with conserved bases in the hairpin loop, at the bulged cytosine, and also with the ribose-phosphate backbone of the IRE (Fig. 2C and Refs. 35 and 46 -50). Docking the IRE hairpin structure on the mt-aconitase structure suggests that the size of the protected regions in IRE RNA is compatible with its localization deep in the putative cleft of IRP-1 covering residues 80 -135 and 150 -170 in the N-terminal domains and residues 720 -740 in the C-terminal domain, which are protected against protease hydrolysis. However, the cleft as found in the crystal structure of mt-aconitase is not open sufficiently to accommodate the IRE hairpin and therefore conformational changes of IRP-1 are certainly required to accommodate the RNA. It was proposed that the apoprotein may adopt a more flexible conformation and that domains 1-3 and 4 may be separated enough via the hinge linker to accommodate the IRE hairpin (34). The fact that the RNA-induced protections extend in a large area of IRP-1 further suggests that multiple contacts occur between IRP-1 and the IRE. Alternatively, protection at some residues may reflect steric hindrance (as for Arg-149) between the proteinases and the IRE or induced conformational changes that render IRP-1 less sensitive to proteases. Such conformational changes may explain the increased cleavages occurring in region 134 -141. It is interesting to note that this region contains a phosphorylation site at Ser-138 that occurs preferentially in the IRE binding form of IRP-1 (25). We also showed that IRP-1 induces some adjustments in the IRE loop promoting the formation of a metal-ion binding site (35). Therefore the mutual adaptation of both IRP-1 and the IRE hairpin may also contribute to the specific recognition. Despite the similarities between IRP-1 and mt-aconitase, the latter is not able to bind to the IRE (21) suggesting that several crucial amino acids in the cleft are not found in the mt-aconitase. The present data reveal that specific interactions appear to exist between residues in the N-and C-terminal domains of IRP-1 and the IRE. Region 720 -740 contains several conserved residues among related IRPs (Fig. 1C). Furthermore, this area of IRP-1 protein is significantly shorter in pig heart mt-aconitase, and the two conserved arginines 721 and 728 are missing in mt-aconitase (Fig. 1C). Because arginines are frequently hydrogen-bonded with the sugar-phosphate backbone and/or bases of RNA, they might represent potential contact sites. From these data, one may speculate that mutations at those specific conserved residues may alter RNA binding in IRP-1. Conversely, their introduction in the mt-aconitase context might confer a specific IRE binding activity to the enzyme.
The footprinting approach also provides further evidence for significant structural changes of IRP-1 induced by the ironsulfur cluster. As previously shown, the reconstitution of the iron-sulfur cluster in IRP-1 precludes RNA binding and confers aconitase activity. In good agreement with the study of Schalinske et al. (25), the c-aconitase form is more resistant to the action of proteases. It is of interest that the strongest reductions of cleavages were located in the two regions found protected by the IRE, mainly at Asp-125, Arg-149, Glu-155, and Arg-187 in the N-terminal domain and at Arg-721, Arg-728, Arg-732, and Asp-751 in the C-terminal domain. Because these residues are located far away from the three cysteines (437, 503, 506) that have been shown essential for catalysis (29,30,51), the protection may reflect conformational changes resulting in protein structure that is less sensitive to proteinases. All these residues lie in the putative cleft of IRP-1 (Fig. 3) and are also located close to the two phosphorylation sites (25). These  (59). Effect of RNA binding: the protected amino acids are drawn in dark blue and yellow, and enhanced cleavages are shown by red residues. Effect of iron-sulfur insertion: the reduced cleavages as compared with IRP-1 are shown by yellow residues. The residues were assigned based on sequence comparison between human IRP-1 and mitochondrial aconitase (23,27). Residues 122 to 132 shown in light blue have been cross-linked to the ferritin IRE (32,33). data indicate that the enzymatic form of IRP-1 adopts a closed conformation stabilized by the presence of the iron-sulfur cluster. This is consistent with the crystal structure of mt-aconitase, which showed the presence of a narrow cleft stabilized by interactions involving side chain residues of domain 4 and domains 1-3 (28). Conversely, in the absence of the iron-sulfur cluster, residues located in the cleft (in the N-terminal domain) and at the edge of the cleft (in domain 4) become accessible to the solvent, permitting the access of the IRE.
In summary, our data suggest the presence of a bipartite IRE binding site in human IRP-1 and illustrate the potential of the protein to alter its conformation in its different functional states. Obviously, the exact nature of these conformational changes and of the IRE-IRP-1 contact sites will require the crystallographic structure of IRE-IRP-1 complex.