Molecular characterization of a second iron-responsive element binding protein, iron regulatory protein 2. Structure, function, and post-translational regulation.

Several genes critical to the uptake, sequestration, and utilization of iron are regulated at the post-transcriptional level. The mRNAs encoded by these genes contain highly conserved stem-loop structures called iron-responsive elements (IREs). IREs function as the nucleic acid-binding sites for a cytosolic RNA-binding protein called the IRE-binding protein or IRE-BP. Binding of the IRE-BP to IREs is reversibly regulated by the iron status of the cell. The IRE-BP is highly conserved among human, rat, mouse, and rabbit, and it is identical to the cytosolic form of aconitase. In this study, we demonstrate that a distinct human gene encoding a protein which is 57% identical to the initially described IRE-BP, now referred to as iron regulatory protein 1 or IRP1, is also capable of binding to IREs with the same in vitro affinity and specificity the originally identified protein. This second gene product, which we call IRP2, is expressed in many tissues, but its mRNA abundance and tissue distribution are different from IRP1. In most cell lines tested, levels of IRP2 are inversely regulated by iron levels due to iron-dependent regulation of the half-life of the protein. In addition to changes in total amounts of IRP2, we demonstrate that the IRE binding activity of IRP2 can also vary up to 4-fold in the absence of any change in IRP2 protein levels. The possible reasons for the existence of a second IRP are discussed.

1 To whom correspondence should be addressed.

309
1988), also referred to as IRF (Mullner et al., 1989), FRP (Walden et al., 19891, and p90 (Harrell et al., 1991). Recently, the term iron regulatory protein 1 or IRPl has been agreed upon for this protein. The terms cytosolic aconitase or IRE-BP will be used to refer to the two mutually exclusive functional activities of IRP1.
The RNA binding activity of IRPl is reversibly regulated by the iron status of the cell. When cells are deprived of iron, IRPl has high affinity for mRNAs that contain an IRE and, conversely, when cells are iron replete, IRE binding activity is lost. IRPl is now known to be identical to cytosolic aconitase (Kaptain et al., 1991;Kennedy et al., 19921, an enzyme described decades ago that had previously lacked a defined role in cytosolic metabolism. Both mitochondrial and cytosolic aconitases contain a t4Fe-4SI cluster that is required for enzymatic activity  and which plays a crucial role in the reversible regulation of IRPl in response to changes in cellular iron. The RNA binding form of the protein has no aconitase activity, and the non-binding form of the protein has full aconitase activity (Haile et al., 1992a(Haile et al., , 1992bGray et al., 1993). The reversible alternation between forms of the protein that either bind RNA or function as an enzyme appears to involve the complete disassembly and reassembly of the cluster (Haile et al., 1992b;Emery-Goodman et al., 1993).
In this study we characterize the properties of a cDNAclosely related to the cDNA of IRPl which was cloned when the IRPl was initially cloned (Rouault et al., 1990). We subsequently demonstrated that the gene product of this clone binds IREs with high specificity and affinity . In addition, this protein, which we refer to as IRP2, is regulated within cells in response to manipulations of iron availability. In the studies presented here, we characterize IRP2, describe its modes of regulation, tissue distribution, and speculate on its role in iron homeostasis.
MATERIALS AND METHODS Gene cloning, DNA sequencing, and chromosomal localization was performed as described previously (Rouault et al., 1990). Overlapping clones permitted reconstruction of an open reading frame coding for a protein of predicted size of 105 kDa for IRP2.
RNA Analysis-Tissue RNA expression was determined by Northern hybridization of selected radiolabeled fragments of the clones of IRPl and IRP2 to 2 pg of poly(A)' selected RNA from each of the tissues indicated in Fig. 8 (Clontech Laboratories, Inc., Palo Alto, CA). Probes were random-primed (Pharmacia Biotech. Inc.) to comparable specific activities (5 x lo7 cpd50 ng of DNA) and shown not to cross-hybridize with the mRNA for the other clone. Similar sized DNA fragments (bases 1-675 of IRP2 and a 647-base pair EcoRI fragment from nucleotides 1327 to 1974 in the cDNA of IRP1) with similar GC contents were used as probes. Quantitation was performed on a PhosphorImager (Molecular Dynamics, Sunnyvale, CA) Visualization of Endogenous IRP2"Endogenous IRP2 was visualized by Western blot analysis and in gel shift assays. Murine B6 fibroblasts and MEL cells were grown and lysed as described previously (Haile et al., 1992a), and 40 pg of protein of a 1% Triton lysate was resolved on an 8% SDS-PAGE and blotted as described previously ). An antipeptide antibody was raised in rabbits against the peptide QKAGKLSPLKVQSKKLP, a mouse sequence derived from a 79 amino acid insertion in the sequence of IRPB relative to IRP1. This peptide was conjugated to an Fmoc multiple antigenic peptide resin (Tam, 1988), and the multiple antigenic peptide was used to immunize rabbits. Anti-peptide antibody affinity purified with peptide conjugated to Sepharose beads was used at a 1:200 dilution to detect endogenous protein by Western blot. Proteins were transferred to nitrocellulose filters (Schleicher and Schuell), blocked and washed as described previously (Kaptain et al., 1991), probed with antibody, and visualized with '251-labeled donkey anti-rabbit immunoglobulin (Amersham Corp.) followed by autoradiography. To demonstrate specificity, a separate blot was probed with anti-peptide antibody in the presence of excess purified IRPB (200 pg/ml.) In order to detect endogenous IRPS, lysates (10 pg) of B6 murine fibroblasts were added to IRE probe (50,000 cpm), and the resulting complexes were fractionated on a 10% acrylamide gel run for 2.5 h at 170 V. Affinity purified Ab was used to specifically identify IRP2. Antibody (5 pg) was added to lysates and IRE probe (specific activity of 30,000 cpdng) and compared with the lane seen when Es-glycine buffer alone was added.
Construction and lhnsfection of Epitope-tagged IRP2-The stop codon of the coding sequence was replaced with sequences that would encode a thrombin site (Engel et al., 1992) and a myc epitope; the resulting construct was cloned into a mammalian expression vector pCDSLRa using oligonucleotides and polymerase chain reaction methodology similar to tagging methods previously described (Kaptain et al., 1991). Transient transfections into COS cells were performed, and lysates from these transfections were used in Western blots and binding studies. In addition, transiently transfected COS cells were metabolically labeled with [%Imethionine and immunoprecipitated using antimyc antibody (Kaptain et al., 1991).
Overexpression and Purification ofIRP2-The open reading frame of the epitope-tagged IRPS construct was cloned into pVL1393 (Invitropen) and expressed in Hi-5 cells. A Dounce homogenate of the insect cells was purified on a MonoQ column, as was done for IRPl (Basilion et al., 1994a, 1994131, with the purified IRP2 eluting from the column at 250 nm KC1 and resulting in a preparation that was 95% pure. Purified material was used in Scatchard analyses and competitions. Expression and Regulation of Recombinant IRP2 in Stable Cell Lines-The open reading frame of the myc-tagged IRPS was cloned into the p220 episomal expression vector with a glucocorticoid-inducible promoter through ligation of the open reading frame of IRP2 into the pGRE5 vector and subsequent ligation of the XbaI fragment of the IRP2-pGRE5 construct (Mader and White, 1993). HeLa and RD4 cells were transfected with the expression construct, and stable cell lines were selected in hygromycin as described previously for IRPl (Philpott et at., 1994). After 32 h of induction of expression with dexamethasone (20 nM), cells were treated for 16 h with either 100 p~ desferrioxamine, 25 PM hemin, or 100 pg/ml ferric ammonium citrate (Sigma) in the continued presence of dexamethasone. Binding of recombinant protein to IRE was assessed in a supershift assay using anti-myc antibody to the myc-tagged recombinant protein (Philpott et al., 1994).
RNA Gel Mobility Shift and Scatchard Analyses-Gel mobility shift studies were performed using 10 pg of lysate and varying amounts of probe as described previously . For Scatchard analysis, 10 pg of lysate from transiently transfected COS cells was added to labeled IRE (specific activity of 56,000 cpdng) in concentrations ranging from 2 to 50 PM. Free and bound forms of IRE were quantitated on dried gels using the PhosphorImager and affinities derived . In addition, IRPl and IRPB preparations purified on MonoQ were compared for their ability to bind the ferritin IRE, the murine eALAS IRE (Dandekar et al., 1991), and the human TfR IRE B (Casey et al., 1989). Probe concentrations ranged from 1 PM to 1 p~, and binding curves were generated along with estimates of the Kd of each protein for a given IRE. W Cross-linking of Tagged IRP2"Lysates (15 pg) from transiently transfected COS cells were prepared (see above) and incubated with radiolabeled IRE (3 x lo5 cpm of IRE of specific activity 8 x lo5 cpdng) alone or in conjunction with 100-fold molar excess of unlabeled specific (IRE) or nonspecific (tRNA) competitors. UV cross-linking was performed as described previously (Kaptain et al., 1991), and cross-linked products were immunoprecipitated with anti-myc antibody, boiled in sample buffer, and resolved on a 7.5% SDS-PAGE.
Synthesis of IRES-The complements of the TW IRE B and eALAS sequences were incorporated into oligonucleotide templates, and radiolabeled IRES were synthesized as described previously ( The position of the 79 amino acid insertion of IRP2 relative to IRPl is shown along with the peptide that was used to raise specific antibodies. The percentage identity between corresponding domains is shown under each domain. Note that with the exception of the 79 amino acid insertion, the alignment and sequence conservation is high between the two proteins. Half-life Studies of IRP2-RD4 cells stably transformed with the epitope-tagged IRPB construct cloned into the p220 expression vector were plated at a density of 1.5 x 105/60-mm plate, induced with dexamethasone (20 nM) for 32 h, and treated with either desferrioxamine (100 p~) or ferric ammonium citrate (100 pg/ml) for 16 h in the continued presence of dexamethasone. Cells were radiolabeled with 0.2 mCi of Trans label (ICN)/plate in cystine-and methionine-free Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum (dialyzed) for 1 h and in the continued presence of dexamethasone and either desferrioxamine or ferric ammonium citrate. After biosynthetic labeling, plates were washed with phosphate-buffered saline x 3, and Triton lysates were made from harvested cells immediately post-labeling (0 h) and at 3-, 6-, 9-, 12-, and 24-h time points. Immunoprecipitation of each lysate was performed with anti-myc antibody, and immunoprecipitated proteins were resolved on a 8% SDS-PAGE as described previously (Kaptain et al., 1991). IRP2 was quantitated on a Phos-phorImager, and autoradiography was performed.
TATGGAAGGCACTGCTl'CCGATAATTAmCTCCCTATAGTGAGTCG-RESULTS RNA Binding Properties of Recombinant ZRP2"In order to assess the RNA binding properties of IRPB, the full-length 963 codon open reading frame plus a carboxyl-terminal epitope tag, derived from the e-myc protein, was transiently expressed in COS cells (see Fig. 1 and legend). The alignment and comparison between the two human proteins predicted to be encoded by these cDNAs has been previously presented  and is schematized in Fig. 1. The second protein, which we now term IRP2, has a predicted molecular mass 105 kDa, containing 963 amino acids in contrast to the 98 kDa mass and 889 amino acids of IRP1. The two proteins are 57% identical and 79% similar, and the alignment demonstrates that this similarity extends over the entire lengths of the proteins. IRP2 contains a 79 amino acid insertion (at residue 136) relative to IRPl which accounts for the difference in masses. Both IRPl and 2 demonstrate approximately 30% overall sequence identity to mitochondrial aconitases.
Recombinant IRPS protein was detected by immunoblotting with an anti-myc epitope antibody (Fig. 2a), and an IRE gel shift complex specific to the transfections was identified (Fig.   2b). Lysates from untransfected COS cells showed a low level of specific IRE band shift that was not affected by either mock transfection or transfection with vector alone. Lysates from cells transfected with the vector which expressed either IRPl or IRPB showed clear IRE-specific transfection-dependent band shift complexes (denoted by the dashes in Fig. 2b). In these gel retardation assays, several nonspecific bands of varying intensity were seen. Specificity was demonstrated by the ability of unlabeled IRE, but not tRNA, to compete with the formation of a radiolabeled shifted complex.

43-
FIG. 2. a, dctc~tion of rccombinant rpitope-tagged IIIE'1 and IRPZ. Immunoprecipitation with anti-my antisera of IRPl and IRP2 from 4 x 10' cpm of I:"Slmethionine-labeled lysates from transiently transfected of 98 and 105 kDa, respectively. h, detection of IRE binding activity of COS cells resolves epitope-tagged IRPs in their predicted size positions transiently expressed IRPl and IRP2. Lysates from transient transfections of COS cells were incubated with radiolabeled IRE, and a gel mobility shift assay was performed as described under "Materials and Methods." The dashes denote the positions of the specific gel retardation complexes. A single specific complex is derived from the clone for IRPl uersus IRPZ as indicated by competition with excess unlabeled IRE or tRNA. Transfection of vector alone and mock transfections show faint IRE-specific endogenous bands from COS cells. The last three lanes show the gel shift pattern seen when liver lysate is used . The Iar7e on the far right shows probe without lysate.
While the band shifts shown in Fig. 2b suggested that IRPB binds IREs, they did not prove that the expressed recombinant IRP2 directly interacted with the labeled RNA. In order to assess the interaction between the IRE and IRPB using a complementary technique, we performed W cross-linking of radiolabeled IRE with lysates containing the expressed proteins. Lysates from COS cells transfected with IRP1, IRP2, or mock transfected were incubated with "P-IRE-containing RNA either alone or in the presence of an excess of unlabeled IRE or tRNA and irradiated with W light. The lysates were then immunoprecipitated with a n anti-myc monoclonal antibody and precipitates were analyzed by SDS-PAGE and autoradiography. As shown in Fig. 3, lysates from cells transfected with either IRPl or IRPB contained myc-tagged protein products that were specifically radiolabeled with the IRE-containing RNA after irradiation. Thus, the IRPB expressed in these cells interacts directly with the IRE-containing RNA.
To analyze the specificity of the IRP2-IRE interaction, we examined the binding of IREs containing point mutations. Thus, for example, the deletion of the conserved first C of the IRE loop destroys the high affinity binding to IRP1, consistent with the fact that this altered IRE will not serve to mediate translational repression in cells (Rouault et al., 1988;Bettany et al., 1992;Jaffrey et al., 1993). The results shown in Fig. 4 demonstrate that IRP2, like IRP1, clearly distinguishes between the parental wild type IRE and the inactivating point mutation. In addition, we examined the effect of heparin which will inhibit low affinity, nonspecific RNA-protein interactions. Heparin at two different concentrations does not differentially affect the interaction of the RNA with either IRPl or 2 (Fig. 4). We directly quantitated the affinity of the IRP2-IRE interaction by performing formal binding curves and Scatchard analyses. This analysis revealed a single binding constant with a calculated K,, of approximately 60 PM, indistinguishable from that obtained using transiently expressed IRP1. These experiments demonstrated that IRPZ was capable of binding a single naturally occurring IRE derived from the human ferritin H chain with high affinity and specificity. We considered the possibility that the two IRE-binding proteins might be able to distinguish between different naturally occurring IREs derived from different mRNAs. To test this, we quantitated the in vitro binding affinities of purified IRPB versus IRPl for other functional IREs including IRE B of the human TfR IREs and the murine eALAS IRE as described under "Materials and Methods.'' In both cases, IRPB exhibited a binding affinity of approximately 50-100 PM, comparable to the binding affinity of IRPl measured for these ligands in the same experiment (data not shown). Thus, the two proteins are capable of recognizing a t least three distinct IREs with the same apparent specificity and affinity.
Identifying the Endogenous IRP2-In order to identify the endogenous IRP2 protein, we immunized rabbits with peptides  6 and 1 2 ) and resolved on a native gel. The migration positions of the IRPs are distinct, and both demonstrate specific IRE binding that is not competed by excess AC or tRNA. from the predicted human and mouse IRPB sequences, including a peptide which is from a specific 79 amino acid insertion in IRPS relative to IRP1. One particular anti-peptide antiserum, raised in rabbits to the murine IRP2 sequence QKAGKLSP-LKVQSKKLP,' showed high reactivity against both mouse and human IRP2 protein. The specificity of this antiserum was demonstrated by immunoblotting against purified human IRPl (Basilion et al., 1994b3 and IRPB as test antigens (Fig. 5a). Only IRP2 was detected by the anti-peptide antibody. Using this antibody, endogenous IRPB was detected by Western blot analysis in lysates from mouse B6 fibroblasts, MEL cells, and human HeLa cells. The specificity of the band migrating in SDS-PAGE a t approximately 100 kDa a s IRPB (Fig. 56) was revealed by competition with 200 pg/ml of purified recombinant IRP2.
In previous studies, more than one specific IRE band shift complex has been observed from lysates from a variety o f species (Leibold and Munro, 1988;Henderson et al., 1993;Chan et al., 1994). That a second band in such assays represents a distinct protein has recently been suggested by a study in which antibodies directed against IRPl failed to immunodeplete a second IRE-specific RNA.protein complex . When murine B6 fibroblasts were treated with desferrioxamine, two band shift complexes were seen (Fig. 5c). In addition to the major complex previously associated with IRP1, a complex of faster mobility was detected that bound IRES with a specificity and apparent affinity similar to the major complex previously associated with IRPl  Using the IRP2-specific antibody, we sought to identify whether an IRP2.IRE complex could be detected in cell lysates using a radiolabeled IRE. When affinity-purified antipeptide antibody to IRPB was added to the band shift, there was specific loss of the faster migrating band, most likely because the antibody prevented either RNA binding or entry of the IRP2-specific complex into the well, while the slower migrating IRPl complex was unaffected (Fig. 5c).
As with IRP1, little band shift activity was detected in lysates derived from iron replete cells (Fig. 5d ). Quantitation of the autoradiograms revealed a greater than 10-fold induction of IRPB RNA binding activity in desferrioxamine-treated cells. However, in contrast to IRP1, no increase in IRPS band shift activity could be recruited by the inclusion of 2% 2-mercaptoethanol in the assay. As previously shown, and as shown in Fig.  5d, such treatment reveals cryptic IRPl RNA binding activity from cells that are iron replete Barton et K. Iwai, unpublished data. al., 1990;Mullner et al., 1992). Titration of reductant failed to reveal any concentration of 2ME that activated cryptic IRP2 binding activity (data not shown).
Complexity of Iron-dependent Regulation of IRP2-The results shown in Fig. 5d demonstrate that cells can regulate the RNA binding activity of IRP2 in response to iron manipulations. In previous work, we (Tang et al., 1993) have shown that a wide range of regulated RNA binding activity can occur in the absence of any change in IRPl protein levels. The failure to recruit IRPB RNA binding activity with reductants suggested either that IRP2, in contrast to IRP1, is insensitive to these reagents or that there is a loss of IRP2 in iron replete cells. To distinguish between these two possibilities, we examined the total levels of IRP2 in lysates derived from murine B6 cells that had been either iron deprived or made iron replete. Immunoblotting of total lysates revealed a significant increase in the level of IRP2 protein in response to treatment with desferrioxamine (Fig. 6a). Direct lysis of cells in boiling SDS-PAGE sample buffer produced the same results (data not shown). Identical iron regulation of IRP2 levels was seen in MEL cells, COS cells, and RD4 cells (data not shown). In contrast, similar manipulations of iron levels in HeLa cells did not change levels of endogenous IRPB (Fig. 621).
To further assess the nature o f regulation of IRPB, we stably transfected HeLa cell lines with constructs designed to express full-length human IRP2 modified with a carboxyl-terminal epitope tag. These constructs were cloned into an Epstein-Barr virus episomal vector which allowed controlled expression of the recombinant IRP2 via a glucocorticoid-responsive promoter (Mader and White, 1993). Recombinant IRPB was expressed in these cells only after dexamethasone treatment. The induced cells were either treated with iron (given either as ferric ammonium citrate or hemin), desferrioxamine, or left untreated for 16 h. Regulation of RNA binding activity of the recombinant IRPB in these cells was assessed by a gel mobility supershift ( Fig. 7) as described previously using antibody to the myc epitope (Philpott et al., 1994). As previously observed for IRP1, the RNA binding activity of the recombinant IRP2 was regulated in response to iron manipulation, with the greatest binding activity seen in iron-depleted cells and the least observed in iron replete cells. The total amount of recombinant IRP2 protein in these lysates was assayed by immunoblotting with the anti-myc epitope antibody. As shown in Fig. 7, there were no detectable differences in the levels of total IRPS protein, consistent with results from Western blotting of HeLa endogenous IRP2 (Fig. 6h). Again, in contrast to IRP1, the addition of reducing agents did not further activate IRP2 RNA binding activity (data not shown). Thus, the RNA binding activity of IRPB in HeLa cells can be post-translationally modified, like IRP1, in cells in response to iron manipulations in the absence o f changes in the level of protein.
In contrast to observations in HeLa cells, desferrioxamine treatment of a number of other cell lines resulted in a significant increase in IRPB levels. In order to determine the mechanism of regulation of protein levels, we studied stably transformed RD4 cell lines which express the full-length epitopetagged human IRP2. Regulation of endogenous IRPB levels had been demonstrated in RD4 cells (see above). After induction of expression of recombinant IRPB with dexamethasone, the cells were treated with either 100 p~ desferrioxamine or 100 pg/ml ferric ammonium citrate for 16 h before radiolabeling for 1 h with ["Slmethionine and cystine. After labeling, the cells were placed in chase medium containing either desferrioxamine or ferric ammonium citrate. The cells were then lysed, and the myc-tagged recombinant protein was immunoprecipitated a t various times as shown (Fig. 8); resulting gels demonstrated a on an 8% SDS-PAGE and probed with either IRP2 anti-peptide antibody (top panel) as described under "Materials and Methods" or with anti-myc antibody (lowerpanel). b, detection of endogenous IRP2 with antipeptide antibodies. Lysate (40 pg) of murine B6 fibroblasts was electrophoresed on an 8% SDS-PAGE, transferred to nitrocellulose filters, and probed with antibody to IRP2 peptide as described under "Materials and Methods." Purified IRP2 was used to preabsorb antibody (+C) a t a concentration of 200 pg/ml. The arrow denotes the position of the specific IRP2 band. c, presence of a specific IRP2 complex in native gels. Lysates (10 pg) of B6 murine fibroblasts and radiolabeled IRE probe were electrophoresed on a 10% native polyacrylamide gel. The first lane (C) reveals two separable complexes which are not affected by excess nonspecific (lane N ) competitor (1 pg of tRNA) and are competed by specific ( S ) competitor (200 ng of unlabeled IRE). Anti-IRP2 antibody depleted the lower specific gel shift complex. The arrow denotes the position of the specific band which is no longer present when antibody specific for IRPB is added. d, endogenous regulation of IRP2-specific gel-shift activity. Lysates from murine B6 fibroblasts were electrophoresed on a 10% polyacrylamide gel after treatment with desferrioxamine (D), ferric ammonium citrate (F), hemin (HI, or no treatment (C). Lane 1 serves as a marker lane, and the arrow denotes the position of the IRP2-specific complex. Lanes 2 , 4 , 6, and 8 have not been treated with 1% 2Me, and lanes 3 , 5 , 7, and 9 are treated with 1% 2Me. Note that the complex which resolves above the arrow shows significant recruitment of binding activity in the presence of 1% 2Me, whereas the IRPB complex is detectable only in lanes from cells treated with desferrioxamine, and no recruitment is seen with 1% 2Me. 6. a , regulation of total amounts of IRP2 after manipulations of cellular iron status. Lysates from murine B6 fibroblasts (100 pg) which were treated for 16 h with desferrioxamine (100 PM), ferric ammonium citrate (100 pglml), hemin (25 p~) , or no treatment (C) were separated on a 8% SDS-PAGE, transferred to nitrocellulose filters, and probed with anti-IRP2 antibody a t a 1:200 dilution. The arrow denotes the position of a band on Western blot migrating slightly above the 98 kDa marker which was shown in Fig. 5h to be specific for IRPZ. b, total amounts of IRP2 do not change after iron manipulations of HeLa cells. Lysates from HeLa cells treated for 16 h prior to lysis as described for Fig. 5a (except that hemin concentration was 10 pm) were electrophoresed and Western blot analysis performed. The arrow denotes the position of the IRP2-specific band.
clear difference between the rate of decay of the protein in the two iron conditions. In the presence of the iron chelator, no loss of protein was seen over 24 h of chase. In the presence of iron, in contrast, the protein had a half-life of approximately 6 h ( Fig.  7 B ) , a difference that can account for the range of regulation of protein levels observed. Tissue Distribution of IRPl and IRP2 mRNA Levels-We analyzed the tissue distribution of the mRNA to compare the expression of the genes IRPl and IRPB (data not shown). In only one tissue, brain, is IRP2 the predominant mRNA. For all other tissues tested there is between 2 and 10 times more IRPl mRNA than IRPB mRNA. The predominance of IRPB mRNA in brain is even more pronounced in fetal samples (data not shown). The mRNA for IRPB is 7 kilobases in length versus 4 kilobases for the mRNA of IRP1; neither probe cross-reacted with the other mRNA. DISCUSSION !Zbo IRE-binding Proteins-This report clearly demonstrates that the cDNA highly related to the originally cloned IRPl (Rouault et al., 1990 also encodes a protein capable of binding IRES with high affinity and specificity. We have since tried to rescreen libraries for additional related cDNAs and have found none. Over the past several years, there have been several reports showing that lysates derived from either tissues or cells contained two distinctly migrating complexes when a I"" iron. RD4 cells were induced to express recombinant IRPB with dexamethasone. After a 16-h treatment with desferrioxamine (D) or ferric ammonium citrate ( F ) , a 1-h pulse with I3'lS Trans label was followed by harvesting of the radiolabeled cells at the indicated time points, and lysates were immunoprecipitated with anti-myc antibody. Arrows point to the IRP2 band. Note that loss of IRPL with iron treatment ( F ) is not accompanied by appearance of higher molecular weight bands. b, a graphic representation of the quantitated bands from gels in panel a indicates a half-life of approximately 6 h for the protein after treatment with iron (0) and no apparent degradation of the protein under conditions of iron depletion (m). assayed using RNA gel mobility shifts and that both of these band shift complexes had specificity for the IRE stem loop structure (Leibold and Munro, 1988;Barton and Munro, 1990;Henderson et al., 1993;Chan et al., 1994). Recently, data has been presented  which showed that this second complex is produced by a distinct IRE-binding protein.
Criteria for distinguishing this second IRE-binding protein from IRPl included distinct migration of complexes on SDS-PAGE after UV cross-linking, generation of different peptide maps from these UV cross-linked RNA.protein complexes, different mobility on ion-exchange chromatography, and lack of reactivity with an anti-peptide antibody to IRP1. While these differences could have arisen as the result of one or more modifications of IRP1, we believe that all of the properties of what we have now proven to be a distinct gene product, IRP2, are compatible with the properties described for the protein that Henderson et al. (1993) termed IRF,. Indeed, their prediction of a molecular mass of 105 kDa, based upon UV cross-linking to RNA and SDS-PAGE migration is in excellent accord with the predicted mass and SDS-PAGE migration of IRP2.
It is surprising that the multiple independent purifications of a high affinity IRE-binding protein from a variety of species and tissues all led to the purification of IRPl Walden et al., 1989;Neupert et al., 1990;Yu et al., 1992). We suspect that differences in tissue abundance of IRPl and IRPB from the sources used, perhaps coupled with the unique activation of IRPl binding activity by reducing agents, explains the repeated identification of only IRPl by biochemical means. The identical RNA binding properties of IRPl and IRPB imply that there is considerable structural similarity between the RNA-binding sites of the two proteins. The evidence that IRPl is likely to assume a three-dimensional structure quite similar to mitochondrial aconitase is strong (reviewed in see also Rouault et al., 1991;Lauble et al., 1992). When IRPl loses its Fe-S cluster, an RNA-binding site which includes residues of the aconitase active site is revealed. This was demonstrated by: 1) the localization of the predominant site of UV cross-linking between the IRE and IRPl to the surface of the cleft in protein domain 1 (Basilion et al., 1994a), and 2) the involvement of 3 active site arginine residues in RNA binding, determined by site-directed mutagenesis (Philpott et al., 1994). Of the 3 active site arginine residues which contribute to RNA binding within the IRPl active site cleft, 2 are conserved in IRP2. The third arginine residue important in RNA binding in IRPl is a lysine in IRPB, but the arginine to lysine mutation at this position in IRPl has been shown to bind IRES equally well (Philpott et al., 1994). Thus, our inferences relating the structure of IRPl to its interaction with RNA are compatible with the similarity of the IRE binding properties of the closely related protein IRPS.
Regulation of ZRP2 by Iron-Perhaps the most striking aspect of the biology of IRPl is its regulation by iron. In this study we observe a complex pattern of regulation of IRP2 activity in response to changes in iron status. In HeLa cells, there is iron-dependent regulation of the RNA binding activity of IRP2 in the absence of changes in the level of IRPB protein. The nature of the specific post-translational modification responsible for altering RNA binding activity is difficult to assess for IRP2 because of two differences between IRPB and IRP1. First, the binding of IRPB was not increased by 2ME treatment, even when the protein could be shown to be present by Western blot data (Fig. This is consistent with published observations Chan et al., 1994) in which no 2ME recruitable activity of the second IRE-specific band shift complex was seen. Second, because of the absence of a measurable Fe-S cluster-dependent enzyme activity for IRP2, we can only speculate a s to whether loss of RNA binding activity of IRPB results from the assembly of an Fe-S cluster, as in IRP1. However, the conservation of each of the cluster ligating cysteines, coupled with conservation of the IRE-binding site leads us to propose that reversible cluster assembly, and concomitant occlusion of the RNA-binding site occurs in IRPB as well as IRP1.
In cells other than HeLa, we observe an increase in the level of total IRPB in response to iron deprivation which, in part, explains the substantial increase in IRPB RNA binding activity. Iron-mediated regulation of IRPB is a complex phenomenon in which two distinguishable modes of regulation contribute to the overall IRE binding activity. The explanation for the irondependent change in IRP2 levels is that the half-life of the protein is greatly lengthened under conditions of iron deprivation. This is similar t o the type of regulation proposed for IRPl in rabbit cells in culture reported by Thach and colleagues (Goessling et al., 1992(Goessling et al., , 1994. While we have not observed this type of regulation for IRP1, Goessling et al. have suggested that this phenomenon is modulated by both cell growth rate and the cell growth conditions.
The mechanism by which high intracellular iron levels effect the degradation of IRPB can now be addressed. It will be particularly interesting to determine whether the potential to assemble an Fe-S cluster via the conserved cluster ligating cys-

Iron Regulatory Protein 2
teines will affect the half-life regulation. The major sequence difference between the two IRPs is in the first domain 79 amino acid insertion present in the first domain of IRP2. Interestingly, this region contains 5 cysteines that may be involved in binding iron. It will be interesting to determine the role ofthis insertion in the regulation of IRPS by mutagenesis.
Increasing Complexity of the IRP Regulatory System-The presence of two IRE-binding proteins presents a new set of questions about this regulatory system. We can consider a variety of ways in which overall iron homeostasis and control might require the existence of these two proteins. These include distinct patterns of protein expression, distinct RNA targets, different effector functions, and distinct modes of regulation. In order to assign relative roles to these two proteins, it will be necessary to determine the levels of expression of the two proteins in individual tissues and cells and, perhaps more importantly, the relative RNA binding activities of the two proteins. This point has been addressed using gel-retardation assays  in a description of a second IRE-BP the activity of which was highest in intestine and brain. In vitro RNA-binding data do not yet point to any difference in target specificity. Furthermore, whether each protein, when bound to a target RNA, has the same effector function remains an intriguing but untested possibility. Clearly the two proteins can be differentially regulated in that IRP2 binding activity changes dramatically as a result of protein degradation, whereas IRPl binding activity changes in the absence of significant changes in protein levels. These differences in the mode of regulation may allow qualitatively or quantitatively different signals to differentially affect the function of the two proteins. Because IRPl is reciprocally regulated as an IREbinding protein and cytosolic aconitase, perhaps there are conditions under which it would be advantageous to activate IRE binding activity but not lose aconitase activity. One difference between the two proteins that may be relevant to regulation is the role of specific substrates for aconitase such as citrate, cis-aconitate, and isocitrate, which stabilize IRPl against in vitro cluster disassembly and other in vitro manipulations that activate IRPl RNA binding activity (Haile et al., 1992b). The mutation of a substrate binding arginine to a lysine in IRP2, coupled with the data that the identical mutation in mitochondrial aconitase significantly lowers substrate binding affinity (Zheng et al., 19921, suggests that IRPS may not bind citrate and may therefore be more susceptible to oxidative cluster disassembly than IRP1, if IRPS indeed has an FeS cluster.
The identification of a second IRP makes the task of understanding the physiology of iron regulation more complex. Adding to this complexity may be the existence of other target mRNAs for these proteins, perhaps with limited tissue distributions. The combination of distinct regulatory stimuli, different regulatory responses, and differentially regulated tissue distributions of IRPl and IRP2 may allow greater flexibility and sensitivity in the function of this high-affinity post-transcriptional regulatory system.