Iron Regulatory Element and Internal Loop/Bulge Structure for Ferritin mRNA Studied by Cobalt(III) Hexammine Binding, Molecular Modeling, and NMR Spectroscopy

The ferritin IRE, a highly conserved (96 -99% in vertebrates) mRNA translation regulatory element in animal mRNA, was studied by molecular modeling (using MC-SYM and DOCKING) and by NMR spectroscopy. Cobalt(III) hexammine was used to model hydrated Mg 2+. IRE isoforms in other mRNAs regulate mRNA translation or stability; all IREs bind IRPs (iron regulatory proteins). A G ‚C base pair, conserved in ferritin IREs, spans an internal loop/bulge in the middle of an A-helix and, combined with a dynamic G‚U base pair, formed a pocket suitable for Co(III) hexammine binding. On the basis of the effects of Co(III) hexammine on the 1H NMR spectrum and results of automatic docking into the IRE model, the IRE bound Co(III) hexammine at the pocket in the major groove; Mg 2+ may bind to the IRE at the same site on the basis of an analogy to Co(III) hexammine and on the Mg 2+ inhibition of Cu(phen)2 cleavage at the site. Distortion of the IRE helix by the internal loop/bulge near a conserved unpaired C required for IRP binding and adjacent to an IRP cross-linking site suggests a role for the pocket in ferritin IRE/IRP interactions. RNA sequences in the noncoding region of mRNAs can regulate mRNA function. The predicted secondary structure is a hairpin distorted by a bulge, bulge loop, or internal loop. Specificity of the three-dimensional structure of RNA regulatory elements is recognized by proteins as in the Tat/ TAR and Rev/RRE interactions of the HIV virus ( 1-3). Bulge loops and internal loops in RNA induce bends or distortions in helixes, creating specific three-dimensional structures and, often, metal binding sites ( 4). Little is known about the three-dimensional structure of natural regulatory elements in eukaryotic mRNAs. The IRE (iron responsive element) family of isoelements is a particularly well characterized control element in normal animal mRNAs encoding proteins of iron metabolism. IREs are hairpins of 9 or 10 base pairs, interrupted by a bulge loop of 1-4 nucleotides with a conserved C residue and with a terminal hexaloop, CAGUGX (reviewed most recently in refs 5-8). The metal complex Cu(phen) 2 binds at the internal bulge/loop ( 9, 10). All IREs recognize a family of RNA binding proteins, the IRPs (iron regulatory proteins); some IREs recognize other proteins as well, such as initiation factors (11, 12, 14). Single-copy IREs in the 5 ′-untranslated regions of mRNAs regulate ribosome binding, while pentuple-copy IREs in the 3 ′-untranslated regions are part of a rapid turnover element regulating mRNA stability; each type of IRE is highly conserved (96 -99%) which contrasts with the lower sequence conservation (35 -45%) between translation and rapid turnover IREs ( 8). The ferritin IRE is the best characterized IRE in terms of structure and function. Assurance of the biological relevance of IRE studies with synthetic RNA, used here and in other types of experiments, has been uniquely provided by earlier investigations using natural ferritin mRNA [poly(A +) RNA]1 to study IRE structure, the IRP binding site and IRE function in regulating protein synthesis ( 9-11, 14-16); ferritin poly(A+) RNA showed function and/or chemical and enzymatic reactivity similar to those of the synthetic RNAs. The ferritin IRE is the most efficient of the translational regulatory IREs (13), possibly because of a conserved internal loop/bulge involving UGC/C rather than the bulge C of other IREs. Previous NMR studies have focused on the role of the ferritin IRE terminal hexaloop ( 17, 18). In this study, a model of the complete IRE 30-mer is developed, assisted by NMR data from15Nand13C-labeled RNA and cobalt(III) hexammine/RNA complexes; the model is consistent with previous chemical and enzymatic studies. Co(III) hexammine significantly shifted proton NMR resonances of G7 and G27 in the internal loop/bulge region and docked in a pocket caused by distortion of the major groove in the middle of the IRE. The same region is also hypersensitive to cleavage by hydroxyl radical ( 16) and displays Mg† The work was supported in part by NIH Grant DK-20251. * Corresponding author at Department of Biochemistry, North Carolina State University, Raleigh, NC 27695-7622. Phone: 919-5155805. Fax: 919-515-5805. E-mail: Theil@bchserver.bch.ncsu.edu. ‡ Polish Academy of Sciences. § Department of Biochemistry, North Carolina State University. | Department of Chemistry, North Carolina State University. 1 Poly(A+) RNA from a natural cell rich in ferritin mRNA [the embryonic red cell in which∼10% of the mRNA is ferritin mRNA (11)] was used with immunoprecipitation to examine control of ferritin synthesis ( 11, 14, 15) or with specific primers to examine IRE structure in the RNA after reaction with structure probes or IRP binding ( 9, 16). 1505 Biochemistry1998,37, 1505-1512 S0006-2960(97)01981-8 CCC: $15.00 © 1998 American Chemical Society Published on Web 01/23/1998 sensitive changes in cleavage by Cu(phen) 2 (9 , indicating solvent accessibility and suggesting that hydrated Mg 2+ binds at the site. In the IRE model, G7/U6 in the internal loop/ bulge and G18/U19 which cross-link to the IRP ( 21) are 22 Å apart, in contrast to only 18 Å in an IRE model without the interhelical pocket, which may relate to correct positioning in the IRP binding site. MATERIALS AND METHODS RNA Synthesis . The 30-mer representing the frog ferritin IRE (Figure 1) was synthesized using the double-stranded T7 polymerase site and the complement of the 30-mer as a template, as previously described ( 17); vertebrate ferritin IREs are highly conserved ( 5-8) (96-99%), but the frog ferritin IRE is the only one which has been studied in natural [poly(A+)] mRNA as well as in synthetic mRNAs and RNA oligomers. The use of full-length double-stranded template increased the yield∼1.5-2-fold; reaction volumes were 24 90 mL. Cloned T7 polymerase was isolated as described by Studier et al. ( 22). RNA was purified by electrophoresis in urea/acrylamide gels as before ( 17), electroreluted, and concentrated by alcohol precipitation. To study the effect of pH on the detection of the G ‚U base pair (Figure 2C), commercially prepared (Cybersyn) RNA was used, but was purified by gel electrophoresis and dialyzed extensively against water before use. RNA enriched in13C and15N was prepared using 13C/15N nucleotide triphosphates (NTPs) as described for the synthesis of RNA with natural abundance levels of the isotopes ( 19, 20). The13Cand15N-enriched NTPs were prepared using crude rRNA fromMethylophilus methyltrophus provided by the NIH Research Resource for Heavy Atoms at Los Alamos National Laboratory. The crude rRNA was digested with DNAse, extracted with phenol and chloroform/isoamyl alcohol (24:1), and precipitated with alcohol followed by digestion to nucleotides with nuclease P1 ( 23) and conversion to NTPs using nucleoside monophosphate kinase, guanylate kinase, pyruvate kinase, myokinase, phosphoenolpyruvate, and ATP as described by Nikonowicz et al. ( 24). After concentration, lyophilization, and alcohol precipitation, the crude NTPs were dissolved in col d 1 M triethylamine/borate buffer (TEAB) at pH 9.5 and desalted on an Affigel 601 (Biorad) column equilibrated i n 1 M TEAB buffer at 5°C (25); the NTPs were eluted with cold distilled water acidified to pH 4-5 with CO2, lyophilized, dissolved in water, filtered through a washed Centricon 10 filter, and stored at pH 8.1 and-20 °C until use. NMR Spectroscopy.RNA (0.5-1.0 mM in 10 mM sodium phosphate buffer and 0.1 mM EDTA at pH 6.8) was heated at 85°C and slowly cooled in the NMR tube. Spectra were acquired on a Bruker DRX 500 MHz spectrometer. Spectra in H2O were obtained either by the Watergate method (26) or by presaturation of the HDO signal fo r 2 s prior to applying an observation pulse or by using the jump -return water suppression and excitation maximum set to the imino resonances ( 27). Data for the two-dimensional (2D) NOESY experiment in 10% D2O/90% H2O were acquired at 12 °C using Watergate-water suppression [a 3 -9-19 pulse sequence with the gradients for water suppression with excitation maximum set to the imino resonances ( 26)]. The spectrum was 2048 × 256 complex data points with a sweep width of 12 000 Hz, a mixing time of 250 ms, a recycle delay of 1.7 s, and 256 scans per slice. Spectra were processed with FELIX 95.0 software (Biosym/Molecular Simulations, Inc.) using an exponential weighing function or shifted sinebell function to resolve overlapped imino protons. NOESY, DQF-COSY, and TOCSY experiments were recorded in 99.996% D 2O on a 500 MHz GE Omega spectrometer or a Bruker 500 MHz spectrometer. Data sets with 2048 complex points int2 and 512 complex points in t1 were acquired with 5000 Hz sweep widths in both dimensions and 128 scans per slice. NOESY spectra were acquired with mixing times of 120, 200, and 400 ms and a recycle delay o f 2 s at 12 and 20°C. The TOCSY spectrum was recorded with a 75 ms MLEV spin lock pulse and a recycle delay of 1.5 s. The DQF-COSY spectra were recorded with WALTZ decoupling of 31P during acquisition and a recycle delay of 1.6 s. The diagonal and cross-peaks of DQF-COSY spectra were phased with antiphase absorption line shape in both directions. All spectra were processed with combinations of exponential and sine-skewed functions and zero-filled to 2K× 2K data points using XWINNMR Bruker or Felix 95.0 software. Spectra with Co(III) hexammine and with varyious pHs were acquired on a Bruker DRX 500 MHz system. Imino proton spectra were obtained by the Watergate method ( 26). Typically, 2048 scans were collected. 1H spectra in 10% D2O/90% H2O were collected at 12°C in 16K point data sets consisting of 1024 scans each. Spectra of double-labeled RNA were obtained on a Varian Unity Plus 600 MHz NMR spectrometer at the University of Chicago, Biological Sciences Division NMR Facility, used in consultation with Dr. Klaas Hallenga. The 2D ( 1H-15N) HSQC experiments were carried out using gradie