Loops and Bulge/Loops in Iron-responsive Element Isoforms Influence Iron Regulatory Protein Binding

A family of noncoding mRNA sequences, iron-responsive elements (IREs), coordinately regulate several mRNAs through binding a family of mRNA-specific proteins, iron regulatory proteins (IRPs). IREs are hairpins with a constant terminal loop and base-paired stems interrupted by an internal loop/bulge (in ferritin mRNA) or a C-bulge (in m-aconitase, erythroid aminolevulinate synthase, and transferrin receptor mRNAs). IRP2 binding requires the conserved C-G base pair in the terminal loop, whereas IRP1 binding occurs with the C-G or engineered U-A. Here we show the contribution of the IRE internal loop/bulge to IRP2 binding by comparing natural and engineered IRE variants. Conversion of the internal loop/bulge in the ferritin-IRE to a C-bulge, by deletion of U, decreased IRP2 binding by >95%, whereas IRP1 binding changed only 13%. Moreover, IRP2 binding to natural IREs with the C-bulge was similar to the ΔU6 ferritin-IRE: >90% lower than the ferritin-IRE. The results predict mRNA-specific variation in IRE-dependent regulation in vivo and may relate to previously observed differences in iron-induced ferritin and m-aconitase synthesis in liver and cultured cells. Variations in IRE structure and cellular IRP1/IRP2 ratios can provide a range of finely tuned, mRNA-specific responses to the same (iron) signal.

3Ј-noncoding regions of animal mRNAs encoding proteins of iron and oxidative metabolism, regulates synthesis of the encoded proteins posttranscriptionally. Iron regulatory proteins (IRPs) bind to the IREs to inhibit ribosome binding or protect mRNA from ribonuclease cleavage (1)(2)(3)(4)(5). The predicted secondary structures of the IRE family are hairpins with a six-nucleotide terminal loop (CAGUGN*, N* ϭ A, C, or U), interrupted by an internal loop/bulge (UGC/C) (ferritin-IRE) or a C-bulge (TfR, eALAS, and m-aconitase IREs), that is generally supported by enzymatic cleavage and chemical probing (6 -8); NMR spectroscopy shows a G-C base pair in the hairpin loop and in the internal loop/bulge (9 -12).
The significance of two IRPs, apparently equivalent in terms of RNA binding and posttranscriptional regulation, is a puzzle, since exclusivity of IRP1 or IRP2 binding for one or another natural IRE sequence has not yet been observed (25)(26)(27)(28). IRP binding specificity for the internal loop/bulge and C-bulge of IREs examined in this study, showed that conversion of the ferritin-IRE internal loop/bulge to a C-bulge, by deletion of a single base U 6 , decreased IRP2 binding 20-fold, with only a small effect on IRP1 binding. Similarly, a C-bulge in the natural IREs (m-aconitase, erythroid ALAS (eALAS), and the transferrin receptor (TfR)), decreased IRP2 binding 10-fold, compared with the ferritin-IRE. Natural IRP1 and IRP2 in a cell extract produced results similar to those observed with recombinant IRPs. The results coincide with structural differences observed by NMR spectroscopy (11,12) and Cu(phen) 2 probing (6). 2 The differential sensitivity of IRP1 and IRP2 binding to natural variations in IREs at the junction of the two helices (internal loop/bulge or C-bulge) suggests that the presence of two IRPs broadens the regulatory range of IREs and emphasizes the importance of the internal loop/bulge region in RNA-protein interactions.

EXPERIMENTAL PROCEDURES
RNA Preparation-RNA, transcribed using T7 RNA polymerase and a chemically synthesized DNA template (9,29), was purified on 12% polyacrylamide/urea gels; concentrated by ethanol precipitation, resuspended in water and stored at Ϫ80°C until use. 5Ј-32 P labeling of RNA was carried out as described previously (6,30), with purification through NENSORB TM columns (DuPont).
Band-shift and Supershift Assays-5Ј-32 P-Labeled RNAs were heated to 85°C for 5 min in 100 mM KCl, 40 mM Hepes, pH 7.2 and annealed to 25°C before each use. In competition experiments, unlabeled RNAs were heated and annealed as described for labeled RNA before adding to the binding reaction. If RNA was only heated to 65°C, the percentage that bound IRP1 was greatly decreased (50 -60%).

RESULTS
Previous studies that compared IRP1 and IRP2 binding had shown that differential IRP binding occurred only with mutations in the hairpin loop (25-28), but not in natural IREs (20,(38)(39)(40). The hairpin loop is the most conserved part of the IREs; evolutionary divergence occurs in the stem and internal loop regions (1)(2)(3)(4)(5). Recent studies of IREs by NMR and other approaches, which showed significant structural differences in the internal loop/bulge and C-bulge IREs (11,12,41), stimulated reexamination of whether IRP1 and IRP2 differentially bind to the internal loop/bulge and C-bulge IREs. To enhance RNA conformational homogeneity, we synthesized RNA of comparable size (28 -30 nucleotides), purified the RNA using denaturing gel electrophoresis, heated the purified RNA to 85°C, and annealed before each use (see "Experimental Procedures").
The influence of the internal loop/bulge characteristic of the ferritin IRE was investigated with recombinant IRPs, by examining the effect of the deletion of U6, which converted the internal loop/bulge into the C-bulge (Fig. 1, a and e). IRP2 recognizes the ferritin-IRE ⌬U6 much more poorly than the ferritin-IRE ( Fig. 2A and Table I) in contrast to IRP1. Mutated ferritin-IRE HL1, HL2, and C8A ( Fig. 1, f-h and Fig. 2A) were controls, to show that results under the conditions used were comparable to those previously observed (25,26,28,42).
Natural IREs all have the same C-G base pair and the pentameric sequence, CAGUG in the terminal hexaloop, but vary in structure at the interhelix junction (internal loop/bulge or C-bulge). Conversion of the ferritin internal loop/bulge to a C-bulge by deletion (⌬U6) differentially altered IRP recognition ( Fig. 2A and Table I). Thus, IRP1 and IRP2 should also have different interactions with the natural C-bulge IREs (m-aconitase, TfR, and eALAS IREs) compared with the natural internal loop/bulge IRE (ferritin-IRE). The results proved our prediction. All natural C-bulge IREs bind IRP2 much more poorly than IRP1 (10 -12%) when compared with ferritin-IRE binding ( Fig. 2A and Table I) and are similar to the ferritin-IRE ⌬U6.
The observation that the internal loop/bulge enhanced IRP2 binding was confirmed by competition experiments with unlabeled RNAs. A 10-fold molar excess of unlabeled ferritin-IRE (internal loop/bulge IRE) prevented binding of the labeled ferritin-IRE to IRP2, whereas Ͼ50-fold molar excess of unlabeled TfR, m-aconitase, or eALAS IREs (C-bulge IREs) were required ( Fig. 2B and Table I).
To ensure that the differential binding of recombinant IRP2 to IREs with an internal loop/bulge or a C-bulge ( Fig. 2 and Table I) was a property of IRPs, and independent of possible differences between natural and recombinant proteins, IREprotein binding was examined with IRPs in rabbit reticulocyte lysates comparing of the ferritin-IRE to the TfR-IRE. IRPs in such cell extracts can regulate translation of IRE-containing mRNAs (43)(44)(45)(46). The ferritin-IRE formed two RNA-protein complexes in the red cell extracts, in relatively equal amounts (Fig. 3A, lane 3), whereas the TfR-IRE formed only one RNAprotein complex (Fig. 3A, lane 4). The complex in the lower band formed with the ferritin IRE was identified as an IRP2⅐RNA complex with IRP2 antibody (compare lanes 5, 6, and 3, 4 in Fig. 3A). The binding of TfR-IRE to IRP1 is 0.52 Ϯ 0.05 that of the ferritin-IRE (Fig. 3A, lanes 5 and 6,  . For ferritin IRE, the predicted structure is similar to that in solution except for the G-C base pairs in the loops (10 -12, 30). m-aconitase mRNA has alternative predicted IRE structure when longer IRE sequences were used (52), emphasizing the importance of flanking regions (53). Conserved residues in the IRE family are shown in bold type. The residues underlined, G, U, C, are ferritin-IRE-specific (5) and form the internal loop/bulge. Arrows denote mutation sites. Fer ϭ ferritin; IL/B ϭ internal loop/bulge; B ϭ C-bulge. The type of distortion at the junctions of the two helices in the IRE is shown in parentheses (IL/B or B). endogenous IRP2 in red cell extracts was 0.10 Ϯ 0.03 that of the ferritin-IRE (Fig. 3A), which was comparable with binding to recombinant IRP2 (Fig. 2A). DISCUSSION IRP isoforms IRP1 and IRP2 showed quantitative differences in binding to IREs from different mRNAs (Figs. 2 and 3 and Table I). The ferritin IRE is recognized best by both IRP1 and IRP2, compared with the m-aconitase, TfR, and eALAS IREs. Accordingly, a larger fraction of the ferritin IRE is likely to be complexed with IRPs than other IREs, which explains the observation that IRE-dependent regulation in vivo and in vitro has the greatest range for the ferritin IRE (38, 40, 47-49).
Among IRE isoforms, variations in the IRE/IRP interaction were greatest for IRP2, suggesting that IRP2 will make the major contribution to differential IRE-dependent regulation in vivo. IRP1/IRP2 ratios vary considerably in different cell types, exemplified by liver, kidney, and intestine: IRP1 Ͼ IRP2 (50), in RRL: IRP1ϳIRP2 (Fig. 3A) and in a pro-B-lymphocyte cell line, which appears to have only IRP2 (51).
IRP2 is sensitive to engineered changes in the IRE hairpin loop and internal loop/bulge. Because the hairpin loop structure is conserved in all natural IREs, its contribution to IRP2 binding will be constant. However, the variation in structure of natural IREs, with C-bulge or the internal loop/bulge, will differentially influence IRP2 binding to natural IREs (Figs. 2 and 3 and Table I). The ferritin-IRE ⌬U6 with the C-bulge was an even poorer competitor for IRP2 binding than natural IRE isoforms with a C-bulge (Fig. 2B), suggesting context effects even within the group of IREs with a C-bulge. NMR studies suggest more conformational flexibility at the internal loop/ bulge than at the C-bulge in IREs (11,12). The IRE consensus sequence designed for NMR studies (11), contained a C-bulge, ⌬U6, and 3 G-C base pairs next to the bulge in the lower stem,  Fig. 1. A, 32 P-RNA ϩ IRP1 or IRP2; B, 32 P-RNA ϩ unlabeled competitor RNA ϩ IRP2. a ϭ IRP/RNA complex.   Fig. 1, a and b) were incubated at 10°C, in a binding buffer that included tRNA and cell extracts (rabbit reticulocyte lysates (RRL), 30 g of proteins), for 20 min followed by incubation with IRP2 antiserum (18) and then incubation with heparin at 25°C. Protein-RNA complexes were resolved by electrophoresis (see "Experimental Procedures"). A PhosphorImager (Molecular Dynamics) with Image Quant software was used for quantitation. The data are representative of three experiments. * denotes the position of the IRE⅐IRP2 complexes, deduced from mobility changes caused by anti-IRP2 antibody. Unbound IREs are not shown on the figure. A, mobility of RNA complexed to binding proteins in RRL; B, quantitation of IRE⅐IRP2 complexes with ferritin/IRP2 ϭ 1.00. Fer ϭ ferritin.
creating an analogue of the ⌬U6 ferritin-IRE, which likely behaves similarly in IRP2 binding. Note that in addition to the C-bulge or internal loop/bulge, effects of IRE flanking regions have been observed on the predicted structure (m-aconitase IRE) (52) and on both solution structure and translation regulation (ferritin-IRE) (53).
RNA conformational flexibility which is matched to differences in the binding proteins, as recently emphasized for BIV-TAT/tar interactions (54), may also explain the differential binding of IRP1 and IRP2 to the IRE isoforms (Figs. 2 and 3 and Table I). For example, the appropriate conformation around the C residue required to bind IRP2 may be more readily achieved by an IRE with an internal loop/bulge, whereas IRP1 may be able to lock onto the C in any IRE. The C residue was disordered in both C-bulge and internal loop/ bulge IREs (11,12), but the internal loop/bulge forms a flexible pocket in the major groove near the conserved C residue (12). IRP contacts the RNA surface on the minor groove (55), and the flexibility of the internal loop/bulge in the major groove may allow an "induced fit" needed for IRP2 binding on the minor groove surfaces.
The question of IRE-dependent coordination of ferritin, maconitase, or TfR synthesis is raised by the tissue-and cell type-specific distribution of IRP1 and IRP2 and the differential binding of IRP2 to ferritin, m-aconitase, and TfR IREs in vitro (Figs. 2 and 3 and Table I). When IRP2 predominates in cells, IRE-dependent repression may be greater for ferritin than for other mRNAs, which can explain differential iron regulation of ferritin and m-aconitase mRNAs in rat liver and cultured cells (38,49,52). On the other hand, the apparently equal regulation of TfR and ferritin by iron in B-lymphocyte cells lacking IRP1 (51) could be attributed to other factors such as multiple IRE copies and/or to alternate structures (56,57). Since both IRP2 and IRP1 can be phosphorylated by protein kinase C (24), phosphorylation of IRP2 may change IRE binding and could allow coordinate regulation of m-aconitase and TfR mRNAs with ferritin mRNA in IRP2-dominant cell types. Controlled coordination of IRE-dependent regulation through protein kinases would create a regulatory interface between the IRE-dependent regulatory pathways and other metabolic pathways. Differential binding among IRE isoforms, coupled with IRP responses to iron (17,18,20), and the potential for regulated phosphorylation and modulation of RNA/protein recognition, indicate the potential for high precision and fine-tuning of IRE-dependent mRNA regulation.