Inhibition of an iron-responsive element/iron regulatory protein-1 complex by ATP binding and hydrolysis Zvezdana Popovic and Douglas M. Templeton Laboratory Medicine and Pathobiology, University of Toronto, Canada Mechanisms of post-transcriptional regulation of gene expression include control of initiation of translation and regulation of mRNA degradation. Among the best-studied models for these processes is regulation of proteins involved in iron homeostasis. These control mechanisms involve functional iron-responsive ele- ments (IREs) in the 5¢-UTRs or 3¢-UTRs of mRNAs that interact with iron regulatory proteins (IRPs), depending upon the amount of iron present in the cell. Two IRPs have been identified: IRP-1, which contains a 4Fe)4S iron–sulfur cluster [1], and IRP-2, which does not [2,3]. IRP-1 has 30% amino acid identity to mitochondrial aconitase [4], a 4Fe)4S enzyme involved in the tricarboxylic acid cycle. IRP-1 is generally believed to interconvert between an enzymatically inac- tive IRE-binding state and a nonbinding form with aconitase activity, the latter requiring an intact 4Fe-4S cluster. Thus, the simple model for iron sensing by IRP-1 involves direct association of iron with the iron– sulfur center to form a complete 4Fe)4S cluster. A linkage between cellular iron levels and energy metabolism is suggested by the influence of agents that Keywords ATP binding; ATP hydrolysis; energy metabolism; iron regulatory proteins; iron- responsive element Correspondence D. M. Templeton, Laboratory Medicine and Pathobiology, University of Toronto, 1 King’s College Circle, Toronto, Ontario, M5S 1A8, Canada Fax: +1 416 978 5959 Tel: +1 416 978 3972 E-mail: doug.templeton@utoronto.ca (Received 22 February 2007, revised 29 March 2007, accepted 24 April 2007) doi:10.1111/j.1742-4658.2007.05843.x Iron regulatory protein-1 binding to the iron-responsive element of mRNA is sensitive to iron, oxidative stress, NO, and hypoxia. Each of these agents changes the level of intracellular ATP, suggesting a link between iron levels and cellular energy metabolism. Furthermore, restoration of iron regula- tory protein-1 aconitase activity after NO removal has been shown to require mitochondrial ATP. We demonstrate here that the iron-responsive element-binding activity of iron regulatory protein is ATP-dependent in HepG2 cells. Iron cannot decrease iron regulatory protein binding activity in cell extracts if they are simultaneously treated with an uncoupler of oxi- dative phosphorylation. Physiologic concentrations of ATP inhibit iron- responsive element ⁄ iron regulatory protein binding in cell extracts and binding of iron-responsive element to recombinant iron regulatory protein-1. ADP has the same effect, in contrast to the nonhydrolyzable analog adenosine 5¢-(b,c-imido)triphosphate, indicating that in order to inhibit iron regulatory protein-1 binding activity, ATP must be hydrolyzed. Indeed, recombinant iron regulatory protein-1 binds ATP with a K d of 86±17lm in a filter-binding assay, and can be photo-crosslinked to azido-ATP. Upon binding, ATP is hydrolyzed. The kinetic parameters [K m ¼ 5.3 lm, V max ¼ 3.4 nmolÆmin )1 Æ(mg protein) )1 ] are consistent with those of a number of other ATP-hydrolyzing proteins, including the RNA- binding helicases. Although the iron-responsive element does not itself hydrolyze ATP, its presence enhances iron regulatory protein-1’s ATPase activity, and ATP hydrolysis results in loss of the complex in gel shift assays. Abbreviations AMP-PNP, adenosine 5¢-(b,c-imido)triphosphate; ATP-cS, adenosine 5¢-O-(3-thiotriphosphate); CCCP, carbonyl cyanide m-chlorophenylhydrazone; EMSA, electrophoretic mobility shift assay; IRE, iron-responsive element; IRP, iron regulatory protein. 3108 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS affect ATP levels and regulate IRP-1. In addition to iron, NO [5], H 2 O 2 [5] and oxidative stress in general [6] influence the activity of IRP-1. All these agents decrease the level of intracellular ATP, and concomit- antly increase IRP-1 binding activity, suggesting integ- rated regulatory mechanisms. In addition, IRP-1 may be regulated by phosphorylation [7]. Generation of NO results in an increase in IRP-1 activity, and a decrease in cytosolic aconitase activity [8–12]. NO donors also lead to ATP depletion [13], consistent with inhibition by NO of a number of enzymes involved in ATP synthesis through glycolysis [14], electron trans- port, and the tricarboxylic acid cycle [15]. ATP deple- tion with NO parallels an increase in IRP-1 binding activity. Exposure to H 2 O 2 promotes removal of the 4Fe)4S cluster of IRP-1, again increasing IRE-binding activity. It is proposed that the H 2 O 2 -mediated conver- sion from cytosolic aconitase to an IRE-binding protein is a result of a signaling pathway rather than of direct chemical modification of the 4Fe)4S cluster by H 2 O 2 [9,16,17]. Indeed, superoxide, H 2 O 2 , and HO are all capable of damaging components of the elec- tron transport apparatus, and thus can disrupt mitoch- ondrial function and limit ATP production [18], while increasing IRP-1 binding activity. Hypoxia also decrea- ses ATP as cells switch their primary means of energy production from the tricarboxylic acid cycle to glyco- lysis [19], and modulates cellular iron homeostasis in human hepatoma and erythroleukemia cells by enhan- cing IRP-1 binding capacity [20], although in rodent cells it has been found to decrease total IRP binding [21], perhaps due to a higher IRP-2 ⁄ IRP-1 ratio in the rodent [20]. The present study was undertaken to examine the possible interaction of ATP with the IRE–IRP system. We have determined that ATP binds to IRP-1, is hydrolyzed, and disrupts IRE–IRP-1 binding. Results Effect of uncoupling oxidative phosphorylation on IRP binding activity We treated HepG2 cells with iron (20 lgÆmL )1 ) for 3 h in the presence or absence of carbonyl cyanide m-chlo- rophenylhydrazone (CCCP), and RNA binding was subsequently examined by electrophoretic mobility shift assay (EMSA) with human ferritin H-chain IRE. Iron treatment decreased IRP binding activity to 60% of control values, but simultaneous treatment with iron and 10 lm CCCP prevented the iron-dependent inhibition of IRP binding activity (Fig. 1A). In parallel, ATP was measured in samples subjected to each treatment. In the absence of oxidative phosphory- lation, the ATP levels in CCCP-treated cells were about half of those in the control cells. Iron treatment did not significantly affect ATP levels in either CCCP- treated or control cells (Fig. 1B). Western blots of IRP-1 demonstrate that the IRP-1 protein level does not change with iron and ⁄ or CCCP treatment (Fig. 1C). IRP-2 was weakly detected by western blot- ting of extracts from HepG2 cells (data not shown). Purification of IRP-1 Recombinant His-tagged hIRP-1 purified on an Ni 2+ affinity column retained several lower molecular weight bands (Fig. 2A), some of which were also detectable by western blotting with antibody to IRP-1 (Fig. 2B), consistent with lower molecular weight fragments in both native and recombinant preparations reported earlier [22]. However, the possibility that these peptides could present artefactual ATP binding or ATPase activity prompted us to purify the protein further. Recombinant IRP-1 was purified to homogeneity by elution from biotinylated IRE–streptavidin agarose with 1 m KCl (fraction E2, Fig. 2). Fraction E2 is essentially pure on an overexposed silver-stained gel (Fig. 2A), is positive for IRP-1 on western blot (Fig. 2B), and retains IRE-binding properties on EMSA (Fig. 2C). Experiments reported below as using affinity-purified recombinant protein were performed on material purified as fraction E2. ATP modulates IRE–IRP interaction in vitro To determine whether ATP directly influences IRP– IRE complexation, we incubated HepG2 cell extracts with different ATP concentrations prior to the addi- tion of labeled IRE and subsequent EMSA (Fig. 3). Binding activity was substantially decreased by 2.5 mm ATP and was undetectable at 5 mm ATP. When puri- fied recombinant IRP-1 was tested for IRE binding in the presence of ATP, the results were similar (Fig. 3), suggesting that interaction can occur among IRP-1, IRE and ATP without involvement of cellular proteins lacking IRP activity. Inclusion of 2% b-mercaptoetha- nol in the reaction mixture had no effect with either HepG2 extract, Ni 2+ –nitrolotriacetic acid-purified recombinant protein, or affinity-purified recombinant protein (Fig. 3B). ATP is hydrolyzed to inhibit IRE–IRP binding To determine the specificity of ATP’s effect on the IRE–IRP interaction, complex formation was studied Z. Popovic and D. M. Templeton ATP binding to IRP-1 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS 3109 in the presence of other nucleotide triphosphates. CTP, GTP, and UTP had no significant effect on IRP-1 bind- ing in either HepG2 cell extracts (Fig. 4A) or with recombinant protein (Fig. 4B). To determine whether ATP hydrolysis is required to modulate IRP–IRE inter- action, we analyzed IRE–IRP complex formation in the presence of the ATP analogs adenosine 5¢-O -(3-thiotri- phosphate) (ATP-cS) and adenosine 5¢-(b,c-imido)tri- phosphate (AMP-PNP). The imidodiphosphate analog AMP-PNP, which is nonhydrolyzable between the b and c phosphorus atoms, had no effect on IRE binding either in cell extracts or with recombinant protein. In whole cell extracts, the IRE complex was inhibited by only about 30% by ATP-cS(5mm). However, ATP-cS had the same inhibitory effect as 5 mm ATP on the IRE complex formed with recombinant IRP-1. ATP-cS is hydrolyzed by some phosphatases and ATPases [23], and the results may reflect a difference between native and recombinant protein. Furthermore, 5 mm ADP inhibited the IRP–IRE complex to the same extent as 5mm ATP, suggesting that ATP is hydrolyzed to ADP to produce the inhibitory effect. Note that after ATP depletion in Fig. 1, binding is not abolished by cellular ADP, presumably because [ATP] ⁄ [ADP] ratios are nor- mally 50–100 in respiring cells [24]. IRP-1 but not IRE has ATPase activity Having demonstrated that ATP interferes with the IRE–IRP complex, we performed experiments to Fig. 1. Effect of an uncoupler of oxidative phosphorylation on IRP binding activity. HepG2 cells growing in complete medium were left untreated (Control) or treated with 20 lgÆmL )1 of Fe as ferric ammonium citrate, 10 lM CCCP, or 20 lgÆmL )1 of Fe and 10 lM CCCP together, for 3 h. (A) EMSA of IRP–IRE binding activity in total cell extracts using the ferritin IRE probe. Binding activity without b-mercapto- ethanol treatment was quantified by scanning densitometry of the autoradiograms, and expressed as a percentage of control. Values in the histogram are means ± SD from three separate experiments. Representative autoradiograms before and after b-mercaptoethanol treatment are also shown. Values differing from control by one-way ANOVA followed by Dunnett’s test are indicated; *P < 0.05, **P < 0.01. (B) Meas- urement of cellular ATP levels following treatments as in (A). ATP was measured in whole cell lysates with a luciferin ⁄ luciferase assay kit, and expressed relative to the values in untreated control cells. Values are means ± SD from three independent experiments expressed relat- ive to control taken as 100% in each experiment, and statistical differences from control are indicated as in (A). The absolute concentration in control cells was 2.27 ± 0.79 pmol ATPÆlg )1 protein. (C) Level of IRP-1 in whole cell extracts following different treatments as in (A). Equal amounts of protein (80 lg per lane) were subjected to western blot analysis with anti-IRP-1 serum. The position of IRP-1 just above the 100 kDa molecular mass marker (compare Fig. 2) is indicated by the arrow. ATP binding to IRP-1 Z. Popovic and D. M. Templeton 3110 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS establish whether IRE or IRP-1 has ATP hydrolytic activity. ATPase activities were measured with a sensi- tive assay that monitors the production of inorganic [ 32 P]phosphate from [ 32 P]ATP[cP] by TLC. Recombin- ant IRP-1 (200–800 ng; 250 nm to 1 lm) has ATPase activity in the presence of 300 lm ATP (Fig. 5A,B) that correlates with the amount of protein in the reac- tion. A reaction with BSA as a protein control had negligible activity. After boiling of IRP-1 for 5 min, less than 3% activity remained. Size exclusion chroma- tography of recombinant IRP-1 could not separate IRE-binding activity from ATPase activity (data not shown). Purified IRE RNA was devoid of ATPase activity (Fig. 5C). The kinetics of ATP hydrolysis were studied with affinity-purified recombinant IRP-1. Release of inor- ganic phosphate was linear with time (Fig. 6A) and plateaued with increasing protein at about 7 nmol phos- phateÆmin )1 Æmg )1 IRP-1 (Fig. 6B). Dependence on [ATP] yielded an apparent K m value of 5.3 ± 1.7 lm with respect to ATP hydrolysis and V max ¼ 3.4 ± 1.7 nmolÆmin )1 Æmg )1 (Fig. 6C,D). ATP hydrolysis can be used for production of energy to do work or to phosphorylate proteins. Although we detected free [ 32 P]phosphate on TLC, we still checked for protein phosphorylation by separation of an ATP hydrolysis mixture containing IRP-1 by SDS ⁄ PAGE and subsequent autoradiogra- phy. We could not detect any autophosphorylation of IRP-1 (data not shown). In a similar experiment, Eisenstein et al. [25] demonstrated that purified rat liver IRP-1 is a substrate for different protein kinases but does not autophosphorylate in the presence of [ 32 P]ATP[cP]. IRP-1 binds ATP More direct proof that ATP binds directly to IRP-1 was sought in a photolabeling experiment. Recombin- ant IRP-1 was incubated with 8-azido-[ 32 P]ATP[aP] for 2 min at 4 °C, and this was followed by UV- induced covalent crosslinking of bound [a- 32 P]nucleo- tide. The proteins were separated by SDS ⁄ PAGE, and AB C Fig. 3. ATP inhibits IRP–IRE binding activity. IRE binding activity was measured by EMSA using either cytosolic extract from HepG2 cells or recombinant human IRP-1 after addition of the indicated concentration of ATP. A representative gel is shown in (A), and the values in the histograms (C) are means ± SD from three independ- ent experiments. Values marked *** differ from control ([ATP] ¼ 0) at P < 0.001. (B) EMSA of HepG2 cell extracts, Ni 2+ –nitrilotriacetic acid-purified ecombinant IRP-1 (Ni-IRP-1), and affinity-purified IRP-1 (A-IRP-1) without ATP, with 5 m M ATP, or with 5 mM ATP plus 2% b-mercaptoethanol (ME). Fig. 2. Purification of IRP-1. Recombinant IRP-1 was purified from E. coli by Ni 2+ –nitrilotriacetic acid agarose chromatography and IRE affinity chromatography. (A) Silver-stained gels of molecular mass markers (lane M; sizes in kDa indicated by arrows), crude E. coli lysate (lane L), 200 ng (N2) and 100 ng (N1) of Ni 2+ –nitrilotriacetic acid agarose-purified IRP-1, and 10 lL aliquots of successive 50 lL elutions from an IRE affinity column with 1 M KCl-containing elution buffer (E1, E2). (B) Western blot analysis with IRP-1 antibody. Lanes represent 500 ng of Ni 2+ –nitrilotriacetic acid agarose-purified IRP-1 (N5), and IRE affinity chromatography flow through (FT), wash (W), and successive 1 M KCl elutions (E1–E4). (C) IRE binding activity by EMSA of 500 ng of Ni 2+ –nitrilotriacetic acid agarose-purified IRP-1 (N5), and IRE affinity chroma- tography flow through (FT), wash (W), and 1 M KCl elutions (E1, E2). Z. Popovic and D. M. Templeton ATP binding to IRP-1 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS 3111 proteins that had bound to 8-azido-[ 32 P]ATP[aP] were visualized by autoradiography of dried gels. Specificity of binding to IRP-1 was demonstrated by competition with unlabeled ATP, ATP-cS, and unlabeled 8-azido- ATP, and by lack of competition with the related pur- ine nucleotide, GTP (Fig. 7A,B). Binding of increasing amounts of 8-azido-[ 32 P]ATP[aP] (Fig. 7C) indicated a K d >30lm. To characterize the binding further, a filter-binding assay was performed with up to 500 lm [ 32 P]ATP[aP] and gave a K d ¼ 86 ± 17 lm (Fig. 7D). ATPase activity is enhanced in the presence of IRE Binding to RNA and ATPase activity are characteris- tics of RNA helicases, and RNA binding increases the ATPase activity of these proteins [26]. To test whether the ATPase activity of IRP-1 is also enhanced in the presence of IRE, we measured hydrolysis of [ 32 P]ATP[cP] in the presence of 0–400 ng of IRE (Fig. 8). Compared to the amount of ATP hydrolyzed with 400 ng of IRP alone, addition of IRE increased the release of inorganic [ 32 P]phosphate by about 50%. IRE alone at 400 ng had no hydrolytic activity (Fig. 5C). Discussion We have demonstrated here that treatment of HepG2 cells with an uncoupler of oxidative phosphorylation depleted them of ATP and prevented suppression of Fig. 4. Requirement for ATP hydrolysis to inhibit IRP–IRE binding activity. EMSA was performed for IRE binding to HepG2 cytosolic extract (upper panel) or recombinant IRP-1 (lower panel). The first lane in each panel is without added nucleotide. In subsequent lanes, the indicated nucleotide analog was added prior to protein, to a final nucleotide concentration of 5 m M. Representative auto- radiograms are shown, and the bars show means ± SD of the intensities from separate experiments (HepG2 extract, n ¼ 4; recombinant IRP-1, n ¼ 3), expressed relative to the nucleotide-free lane, taken as 100%. Values marked ** are significantly lower than the nucleotide-free lane; P < 0.01. A B C Fig. 5. Recombinant IRP-1 has ATPase activity. (A) TLC autoradio- gram of [ 32 P]ATP[cP] reaction mixtures containing 300 lM ATP in the presence of increasing amounts of recombinant IRP-1. The position of free inorganic phosphate is indicated by the arrow. In the last lane, 800 lg of BSA is included as a nonspecific protein control. (B) Combined results (means ± SD) of three repetitions of experiments as in (A), with inorganic phosphate quantitated by scraping and scintillation counting. Significant increases in the pres- ence of IRP-1 are indicated; **P < 0.01, ***P < 0.001. (C) TLC autoradiogram of [ 32 P]ATP[cP] reaction mixtures in the presence of increasing amounts of IRE mRNA. ATP binding to IRP-1 Z. Popovic and D. M. Templeton 3112 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS IRP activity by iron. A similar effect of CCCP-induced ATP depletion was observed on the reconstitution of cytosolic and mitochondrial aconitase activities after removal of NO-generating agents [27]. That is, the high IRE-binding activity of IRP-1 after NO exposure remained elevated in CCCP-treated cells after cessation of NO flux. Moreover, Bouton et al. [27] also observed that IRP-1 cannot dissociate from IRE if mitochondria are unable to produce ATP. They suggested that IRP- 1–IRE complex dissociation, an obligate step upstream of 4Fe)4S cluster repair, utilizes an ATP-dependent mechanism. Although this is a plausible explanation for our results, we have not attempted to rule out other possibilities. For instance, interference with mito- chondrial function and dissipation of the mitochond- rial transmembrane potential by CCCP could prevent reconstitution by interfering with mitochondrial 4Fe)4S cluster synthesis, although Li et al. have dem- onstrated that reconstitution of mammalian cytosolic aconitase probably involves cytosolic forms of enzymes of cluster synthesis [28]. Although iron treatment has a dramatic impact on iron–sulfur proteins and the bioenergetic function of mitochondria [29], 3 h of iron treatment of our HepG2 cultures did not significantly increase the ATP level compared to that in control cells (Fig. 1). Oexle et al. [30] found a significant increase in ATP after 24 h in iron-treated, differentiating K562 cells as compared to control or deferoxamine-treated cells, and aconitase activity was also increased in iron-treated cells, as expected. Thus, one may speculate that iron regulates IRP binding activity in part through changes in ATP concentration (Fig. 1) and ⁄ or that ATP is required for iron-mediated regulation of IRP-1 binding activity. On the other hand, a K d of 86 lm argues against direct regulation of IRP-1 binding by ATP. Total cellular ATP levels are c.5mm [32], and this would suggest that IRP-1 would be saturated with ATP under normal circumstances, preventing IRE binding. Indeed, this might account for the observation, derived from IRP-1 gene-ablated mice, that IRP-2 rather than IRP-1 is responsible for the post-transcriptional regulation of iron homeostasis, and that IRP-1 functions as a cyto- solic aconitase, rather than as an RNA-binding pro- tein, in most cells [33]. However, the lower K m values of many ATPases in the micromolar range suggest that they experience lower local concentrations of ATP, e.g. due to compartmentalization. Furthermore, at least 50% of cytosolic ADP is protein-bound [34]. Thus, whether IRP-1 experiences subsaturating cytosolic ATP concentrations in the cell remains an open ques- tion. Certainly, our estimated K m value of 5.3 lm is quite in keeping with the values reported for a number of other ATP-hydrolyzing proteins, e.g. Escherichia coli DnaK (20 lm) [35], Hsc70 (1.4 lm) [36], and F1- ATPase (15 lm) [37]. Furthermore, these proteins have V max values in the range 1.1–3.5 nmolÆmin )1 Æmg )1 , comparable to the value of 3.4 nmolÆmin )1 Æmg )1 measured here, supporting a physiologic relevance of the observed ATP binding. The requirement for more than 1 mm ATP to abol- ish binding in EMSA (Fig. 3) is at first sight inconsis- tent with a K d of 86 lm. The K m and K d values are in reasonable agreement, and it may be that factors other AB CD Fig. 6. Kinetics of ATP hydrolysis by affinity- purified recombinant IRP-1. (A) Time dependence of ATP hydrolysis. IRP-1 (20 ng) was incubated in a final volume of 20 lL with 10 n M [ 32 P]ATP[cP] in the pres- ence of 10 l M ATP for 1 h, and [ 32 P]phos- phate release was measured by TLC. (B) Dependence of ATP hydrolysis on IRP-1 concentration. The indicated amount of IRP was included in the reaction with conditions as in (A). (C) A representative thin layer chromatogram of ATP hydrolysis with 20 ng of affinity-purified recombinant IRP-1, 10 n M [ 32 P]ATP[cP] and the indicated amount of unlabeled ATP for 1 h. Bk is the reaction mixture with no added IRP-1 or cold ATP. (D) Rate data calculated from (C). Nonlinear regression (r 2 ¼ 0.67) of the rate data gives K m ¼ 5.3 ± 1.7 lM and V max ¼ 3.4 ± 0.3 nmolÆmin )1 Æmg )1 . Z. Popovic and D. M. Templeton ATP binding to IRP-1 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS 3113 than binding and hydrolysis, e.g. a conformational change induced by high ATP concentrations, are necessary for disruption of RNA binding. However, other factors in the EMSA protocol, including changes in Mg 2+ and ATP concentrations during electropho- retic separation, may influence binding, and we think this the more likely explanation; direct comparison of the ATP dependence of EMSA and binding of ATP to purified protein under optimized conditions may not be appropriate. Nor can the concentration of ATP neces- sary to disrupt IRP–IRE binding on EMSA be readily compared to effective concentrations in vivo, where additional interactions may be involved. Disruption of the 4Fe)4S cluster of IRP-1 with 2% b-mercaptoethanol is widely used to uncover total IRP-1 binding activity. However, in our experiments addition of 2% b-mercaptoethanol did not reconstitute IRP-1 binding inhibited by ATP, and nor was the pro- tein degraded. Similarly, Gonzalez et al. [38] reported that IRP binding activity inhibited by interferon-c ⁄ lipopolysaccharide-dependent NO production and phorbol ester treatment was not recovered by exposure of IRP-1 to 2% b-mercaptoethanol, and nor was this due to protein degradation. Furthermore, we previously found that b-mercaptoethanol somewhat diminished the binding activity achieved by in vitro treatment with deferoxamine [39]. Thus, while b-mercaptoethanol can disrupt the 4Fe)4S cluster and facilitate binding of IRP-1, it may also decrease bind- ing by opening up the protein to internal disulfide bond formation [40]. These data suggest a novel mech- anism for ATP-dependent inhibition of IRP-1 binding activity that cannot be recovered by reductive treat- ment with b-mercaptoethanol. We demonstrate a requirement for ATP hydrolysis in order to inhibit recombinant IRP-1 binding activity, and both ATP and ADP suppress binding (Fig. 3). The mechanism of activation of IRP-1 by oxidative stress might well involve ATP. Thus, IRP-1 activation with H 2 O 2 requires Mg 2+ , and is sensitive to treatment with alkaline phosphatase that results in a 90% reduc- tion of ATP levels [41]. Furthermore, ATP-cS and GTP-cS inhibited IRP-1 activation by H 2 O 2 signifi- cantly. Pantopoulos & Hentze [41] conclude that IRP-1 activation by H 2 O 2 in permeabilized cells appears to require ATP and GTP, indicating an energy dependence of the process and ⁄ or the involvement of a phosphorylation–dephosphorylation cycle. We found that ATP-cS causes only about 30% inhibition of binding in whole cell extracts but 80% inhibition with recombinant protein, and this difference may arise from different characteristics of native HepG2 protein and recombinant protein, such as the presence of a His-tag on the latter. Although ATP-cS is often con- sidered a nonhydrolyzable ATP analog, it undergoes hydrolysis in some circumstances, and is a substrate for RNA-dependent nucleotide hydrolysis by helicases [23]. Because hydrolysis of a phosphorothioate group is actually predicted to be faster than that of the corresponding phosphate on the basis of chemical considerations, it has been suggested that the rate- determining step may be a conformational change that takes place after substrate binding [23]. Native and recombinant proteins may have different conforma- A B D C Fig. 7. Photolabeling of IRP-1 with 8-azido-ATP. (A) Two hundred nanograms of affinity-purified recombinant IRP-1 was reacted with 5 l M 8-azido-[ 32 P]ATP[aP], and then subjected to gradient SDS ⁄ PAGE; and the gel was then silver stained (Ag), and autoradio- graphed ( 32 P). The arrow indicates the position of IRP-1 at 105 kDa. (B) Autoradiograph after photo-crosslinking of 4 lg portions of Ni 2+ –nitrilotriacetic acid-purified recombinant IRP-1 with 5 lM 8-azi- do-[ 32 P]ATP[aP] alone in control lanes (C), or in the presence of competitors: 20 m M ATP (A2), 20 mM GTP (G2), 100 mM ATP (A10), 100 m M GTP (G10), 20 mM ATP-cS(c), and 20 mM unlabeled 8-azido-ATP (Z). (C) Photo-crosslinking of 4 lg portions of Ni 2+ –nitril- otriacetic acid-purified recombinant IRP-1 with increasing concentra- tions of 8-azido-[ 32 P]ATP[aP] as indicated. The reaction mixture was electrophoresed, silver stained, and autoradiographed. (D) Filter- binding assay of 0.62 lg of IRP-1 incubated with 100 n M [ 32 P]ATP[aP] and the indicated concentration of unlabeled ATP. The reaction mixture was collected on nitrocellulose membranes as described in Experimental procedures. The saturation plot was fitted by nonlinear regression (r 2 ¼ 0.98) to give values of K d ¼ 86 ± 17 lM and B max ¼ 4.6 ± 0.3, typical of three such experiments. ATP binding to IRP-1 Z. Popovic and D. M. Templeton 3114 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS tional flexibilities. Alternatively, lower inhibition by ATP-cS in cell extracts may be due to the presence of additional cellular proteins from HepG2 cytosol that are involved in IRP–IRE complex formation. Further- more, ATP-cS, but not AMP-PNP, inhibited IRP–IRE binding in extracts from fibroblasts treated with the analogs in culture [41]. Therefore, the 30% inhibition of binding in our cell extracts still suggests that ATP should be hydrolyzed in order to interact with the IRP-1–IRE complex. This is confirmed by the lack of any effect of the nonhydrolyzable imidodiphosphate analog AMP-PNP. Direct binding of ATP to IRP-1 was demonstrated by photolabeling with radioactive photosensitive 8-azi- do-[ 32 P]ATP[aP] (Fig. 7), a method that has been used to identify ATP-binding sites on proteins, e.g. on histi- dine permease [42] and chaperonin GroEL [43]. Labe- ling with 8-azido-[ 32 P]ATP[aP], together with evidence of binding of [ 32 P]ATP[aP] but not [ 32 P]ATP[cP] in the filter assay (data not shown), is a strong indication of an ATPase activity of IRP-1. Competition by ATP and analogs, but not GTP (Fig. 7) or CTP (data not shown), supports specificity of the binding site(s). A large excess of unlabeled ATP was necessary to elimin- ate the signal. The reason for this is unknown, but it has been observed previously in similar ATP photo- labeling experiments [44–47]. A simple model, then, is that IRP-1 is bound to RNA at lower ATP concentrations, but at higher concentrations it hydrolyzes ATP and dissociates from RNA. This means that a low local ATP level main- tains the IRP-1–IRE complex at high iron concentra- tion, and when the ATP level increases, the complex dissociates. Increased ATP hydrolysis in the presence of IRE is reminiscent of an important ATPase protein family ) the helicases [48]. They unwind duplex RNAs in concert with the hydrolysis of nucleoside. For example, the RNA-binding helicase eIF4A undergoes cyclic conformational changes upon ATP binding and hydrolysis such that the eIF4A–ADP complex has a greatly decreased affinity for ssRNA [26]. Our data do not show whether ATP inhibits the binding of IRP-1 to IRE, or instead facilitates dissociation of the com- plex. Enhancement of ATP hydrolysis by IRE in the presence of IRP-1, but absence of hydrolysis in the presence of IRE alone, suggests that ATP may interact with the complex. However, the enhancement is not dramatic, and whether ATP inhibits formation or increases dissociation of the IRE–IRP-1 complex can- not be definitively determined from the present data. If formation of the IRP-1–IRE complex depends on iron concentration, but its dissociation depends on ATP concentration, then the expression of IRE-con- taining genes would actually depend on both iron and ATP levels. At low ATP levels, transferrin receptor mRNA would be stable, and transferrin receptor on the membrane would continue to take up iron. In par- allel, ferritin mRNA (and other 5¢-IRE mRNAs) would still have IRP-1 bound, and ferritin synthesis would be blocked, so iron would not be stored in ferr- itin but relocated to iron-binding proteins, many of them located in mitochondria and involved in energy metabolism. The present observations suggest an expanded role for the IRE–IRP system in regulating cellular energy metabolism. Experimental procedures Cell culture and iron loading HepG2 cells (no. HB 8065) were obtained from the Ameri- can Type Culture Collection (Rockville, MD) and grown in a-MEM supplemented with 10% fetal bovine serum, peni- cillin (100 UÆmL )1 ) and streptomycin (100 lgÆmL )1 ). Cells AB Fig. 8. IRE stimulates ATPase activity of IRP-1. (A) Hydrolysis reaction of [ 32 P]ATP[cP]. The blank contains neither pro- tein nor RNA. Other lanes contain 400 ng of recombinant IRP-1 and the indicated amount of IRE RNA. Values are means ± SD from three separate experiments. *Significantly greater than IRP-1 alone (P<0.05). (B) Autoradiogram of a thin layer chromatogram showing the release of [ 32 P]phosphate. Lanes are as in (A). Z. Popovic and D. M. Templeton ATP binding to IRP-1 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS 3115 were plated at a density of 4 · 10 6 cells per 60 mm dish, allowed to attach and grow overnight, and then treated with iron (20 lgÆmL )1 ) as ferric ammonium citrate, 10 lm CCCP, or both. After 3 h, cells were harvested and protein was extracted (see below). Preparation of cytoplasmic extracts Monolayers of HepG2 cells were scraped in NaCl ⁄ P i and centrifuged at 11 000 g for 1 min with an Eppendorf 5415R centrifuge. The pellet was resuspended in extraction buffer (EB: 10 mm Hepes, pH 7.6, 3 mm MgCl 2 ,40mm KCl, 1mm dithiothreitol, 0.2% Nonidet P-40). After sonication for 5 s, the suspension was centrifuged for 1 min at 11 000 g with an Eppendorf 5415R centrifuge. The super- natant was recentrifuged under the same conditions for use in the IRE-binding assay [49]. The protein content of the extracts was determined with a Bradford-based protein determination kit (Bio-Rad, Mississauga, ON, Canada). The protein extract was aliquoted and stored at ) 80 °C. Preparation of RNA transcripts Transcription was performed in vitro with 1 lgofBamH1 linearized plasmid pSPTfer [50], coding the human ferritin H-chain IRE (provided by L. C. Ku ¨ hn, ISREC, Epalinges, Switzerland) in the presence of 50 lCi of [ 32 P]CTP[aP] (800 CiÆmmol )1 ; ICN, Costa Mesa, CA) and T7 RNA Polymerase using a Promega in vitro transcription system (Promega, Madison, WI). Full-length transcripts were puri- fied on QuickSpin columns (Roche Molecular Biochemicals, Laval, PQ, Canada) for radiolabeled RNA purification according to the manufacturer’s procedure. Purification of recombinant IRP-1 Human recombinant IRP-1 was purified from E. coli trans- formed with pT7-his-IRP-1 plasmid (a gift from K Pantop- oulos, McGill University) as previously described [51]. Bacterial pellets were resuspended in 20 mm Tris ⁄ HCl (pH 8.0), with 250 mm NaCl and 0.5% Nonidet P-40, fro- zen, thawed, and sonicated. The lysates were clarified by centrifugation (4300 g, Sorvall centrifuge with SS34 rotor). Ni 2+ –nitrilotriacetic acid agarose beads (Qiagen, Mississ- auga, ON, Canada) were washed twice with buffer N (24 mm Hepes, pH 7.6, with 150 mm potassium acetate, 1.5 mm MgCl 2 , and 5% glycerol). KCl was added to the lysate to a final concentration of 0.4 m, and the lysate was tumbled with Ni 2 –nitrilotriacetic acid agarose beads for 1 h at 4 °C. This mixture was poured into a column and washed sequentially at 4 °C with buffer N plus 0.4 m KCl, buffer N alone, and buffer N with 5 mm imidazole. IRP-1 was eluted with buffer N containing 50 mm imidazole. Recombinant protein was further purified with the use of streptavidin-conjugated paramagnetic spheres (Invitrogen Canada Inc., Burlington, ON, Canada) and biotinylated IRE RNA [52]. For biotinylation, 100 lL of a transcription reaction mixture was prepared containing 5 lLof10mm GTP, UTP and ATP, 2.5 lLof10mm CTP, and 2.5 lLof 10 mm biotin-14-CTP (Invitrogen Canada Inc.) according to the manufacturer’s instruction. Biotin-14-CTP is a CTP analog that contains biotin attached to the N4 position via a 14-atom linker. Full-length transcripts were purified on QuickSpin columns and used for affinity chromatography. Streptavidin beads (100 lL) were washed two times with 200 lL of binding buffer (20 mm Hepes, pH 7.5, 150 mm KCl, 1 mm phenylmethanesulfonyl fluoride), resuspended in 100 lL of binding buffer, and added to biotinylated IRE (10 lg in 150 lL of binding buffer). Biotinylated IRE was bound to streptavidin beads for 10 min at 4 °C, and unbound IRE was washed with 100 lL of binding buffer. Fifty micrograms of fresh IRP-1 was then added to mag- netic beads in 100 mm KCl in 200 lL of binding buffer, and tumbled for 90 min at 4 °C. Beads were washed three times with 400 lL of washing buffer (20 mm Hepes, pH 7.5, 150 mm KCl, 2 mm MgCl 2 , 0.5 mm dithiothreitol, 0.5 mm phenylmethanesulfonyl fluoride, 5% glycerol, and 0.1% Nonidet P-40), and eluted five times with 50 lLof elution buffer (50 mm Hepes, pH 7.5, 0.5 mm EDTA, 1 m KCl, 1 mm phenylmethanesulfonyl fluoride, 5% glycerol). EMSA RNA–protein interactions were analyzed in cytoplasmic extracts containing 15–30 lg of total protein, or with 400 ng of Ni 2+ –nitrilotriacetic acid-isolated recombinant IRP-1 or 20 ng of affinity-purified IRP-1. Extracts were mixed with a molar excess of labeled ferritin H-chain IRE (10 ng) in EB buffer without Nonidet P-40 and with or without ATP (final volume 20 lL). In some experiments, various nucleotide analogs were added at 5 mm final con- centration to the EB buffer prior to adding protein and IRE. RNAse T1 was added after 10 min of incubation, and 2 lL of 2.5 mgÆmL )1 heparin was added after a further 10 min for an additional 10 min period. IRE–IRP com- plexes were resolved in 6% nondenaturing polyacrylamide gels. In some experiments, parallel samples were treated with 2% b-mercaptoethanol before addition of the labeled RNA probe to unmask total IRP binding activity. Western blot analysis Proteins were resolved by 8% SDS ⁄ PAGE, transferred to nitrocellulose membranes, incubated with antibodies to IRP-1 or IRP-2 (Alpha Diagnostic International, San Anto- nio, TX) (1 : 5000 dilution), and detected using ECL west- ern blot detection reagents (Amersham Bioscience, Baie d’Urfe ´ , PQ, Canada). ATP binding to IRP-1 Z. Popovic and D. M. Templeton 3116 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS 8-Azido-[ 32 P]nucleotide binding experiments Recombinant human IRP-1 (0.2–4 lg) was combined with 8-azido-[ 32 P]ATP[aP] (Affinity Labeling Technologies, Lex- ington, KY; specific activity 10–15 CiÆmmol )1 ) at a final concentration of 20 lm and assay buffer (25 mm Tris, pH 7.6, 100 mm KCl, 5 mm MgCl 2 and 10% glycerol) in a total volume of 20 lL according to the manufacturer’s pro- tocol. After 2 min of incubation on ice, the samples were crosslinked immediately with a 254 nm UV lamp for 2 min. The samples were separated on either a 4–15% gradient or 8% SDS ⁄ PAGE gels (Bio-Rad), silver stained, dried, and exposed to X-ray film. Filter-binding assay ATP-binding activity was determined by a filter-binding assay as previously described [53]. IRP-1 (0.5–1 lg) was incubated with 100 nm [ 32 P]ATP[aP] (Perkin Elmer Can- ada, Woodridge, ON, Canada; 3000 CiÆmmol )1 ) at room temperature for 30 min in 20 lL of binding buffer (25 mm Tris, pH 7.6, 100 mm KCl, 5 mm MgCl 2 ,1mm dithiothrei- tol, and 10% glycerol) containing different concentrations of unlabeled ATP. Nitrocellulose membranes were soaked briefly in 0.4 m KOH, rinsed thoroughly, and mounted on a dot blot apparatus. The binding reaction was stopped by addition of 70 lL of ice-cold buffer, and the reaction mix- ture was loaded on the dot blot appratus in triplicate 30 lL aliquots, and washed twice with 200 lL of cold binding buffer. Nitrocellulose membranes were exposed to film, and each spot was cut out and counted by liquid scintillation. ATP hydrolysis and TLC ATP hydrolysis was determined by measuring the release of [ 32 P]phosphate from [ 32 P]ATP[cP] (Perkin Elmer Canada; 3000 CiÆmmol )1 ) [36]. IRP-1 was added to 10 nm [ 32 P]ATP [cP] in 25 mm Tris ⁄ HCl (pH 7.6), 100 mm KCl, 5 mm MgCl 2 , 10% glycerol and varying concentration of ATP in a final volume of 20 lL. After incubation for 60 min at room temperature, 2 lL was spotted on polyethyleneimine cellulose TLC plates (Sigma, St Louis, MO, USA). IRE (100–800 ng) was analyzed under the same conditions. The spotted samples were resolved using 0.5 m lithium chloride in 0.5 m formic acid and visualized by autoradiography. Spots containing the released [ 32 P]phosphate were cut and measured by scintillation counting. ATP measurement To extract ATP, cells were harvested in cold 0.6 m HClO 4 and centrifuged at 9000 g for 1 min with an Eppendorf 5415R centrifuge. The pellet was saved for protein deter- mination by Lowry’s method [54,55]. The supernatant was neutralized with 5 m KOH and 0.4 m imidazole, and centri- fuged again. Aliquots of ATP-containing extract were saved at ) 80 °C. ATP was measured with a commercial lucifer- in ⁄ luciferase (ENLITEN) kit (Promega) according to the manufacturer’s protocol. Briefly, 100 l L of a 1 : 500 diluted sample was added to a microplate, and 50 lL of ENLITEN rL ⁄ L reagent was added immediately before measurement and mixed by repeated pipetting. Measurements were taken in a microplate luminometer (MicroLumat Plus; ED & G Berthold, Gaithersburg, MD, USA). Three measurements were made for each sample and were compared to a stan- dard curve of ATP solutions ranging from 10 to 750 nm. Statistical methods Values from multiple experiments are expressed as mean ± SD. Differences between means in multiple com- parisons were analyzed by one-way anova. Comparisons of multiple values against a control value were performed using Dunnett’s post hoc test. Kinetic data and binding plots were fitted by nonlinear regression using the program prism (GraphPad Software, San Diego, CA). Acknowledgements This work was supported by grants from the Heart and Stroke Foundation of Canada (grant T4134) and the Canadian Institutes of Health Research (grant MT11270). We thank Dr Tania Christova for helpful discussions. References 1 Kennedy MC, Mende-Mueller L, Blondin GA & Beinert H (1992) Purification and characterization of cytosolic aconitase from beef liver and its relationship to the iron-responsive element binding protein. Proc Natl Acad Sci USA 89, 11730–11734 (erratum appears in Proc Natl Acad Sci USA 90, 2556). 2 Samaniego F, Chin J, Iwai K, Rouault TA & Klausner RD (1994) Molecular characterization of a second iron- responsive element binding protein, iron regulatory protein 2. Structure, function, and post-translational regulation. J Biol Chem 269, 30904–30910. 3 Guo B, Yu Y & Leibold EA (1994) Iron regulates cyto- plasmic levels of a novel iron-responsive element-bind- ing protein without aconitase activity. J Biol Chem 269, 24252–24260. 4 Rouault TA, Stout CD, Kaptain S, Harford JB & Klausner RD (1991) Structural relationship between an iron-regulated RNA-binding protein (IRE-BP) and aco- nitase: functional implications. Cell 64, 881–883. 5 Martins EL, Robalinho RL & Meneghini R (1995) Oxi- dative stress induces activation of a cytosolic protein Z. Popovic and D. M. Templeton ATP binding to IRP-1 FEBS Journal 274 (2007) 3108–3119 ª 2007 The Authors Journal compilation ª 2007 FEBS 3117 [...]... Recombinant ¨ iron -regulatory factor functions as an iron-responsiveelement -binding protein, a translational repressor and an aconitase A functional assay for translation repression and direct demonstration of the iron switch Eur J Biochem 218, 657–667 Scaturro M, Sala A, Cutrona G, Raimondi L, Cannino G, Fontana S, Pucci-Minafra I & Liegro ID (2003) Purification by affinity chromatography of H10 RNAbinding... mutational analysis for the ATP binding of DnaA protein Functions of two conserved amino acids (Lys-178 and Asp-235) located in the ATP- binding domain of DnaA protein in vitro and in vivo J Biol Chem 273, 20847–20851 Lowry OH, Rosebrough NJ, Farr AL & Randall RJ (1951) Protein measurement with the Folin phenol reagent J Biol Chem 193, 265–275 Walker JM (2002) The Protein Protocols Handbook, 2nd edn Humana... Billiar TR, Curran RD, Stuehr DJ, Ochoa JB & Simmons RL (1991) Effect of exogenous and endogenous nitric oxide on mitochondrial respiration of rat hepatocytes Am J Physiol 260, C910–C916 Mueller S & Pantopoulos K (2002) Activation of iron regulatory protein-1 by oxidative stress Methods Enzymol 348, 324–337 Pantopoulos K (2004) Iron metabolism and the IRE ⁄ IRP regulatory system: an update Ann NY Acad.. .ATP binding to IRP-1 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Z Popovic and D M Templeton responsible for control of iron uptake Arch Biochem Biophys 316, 128–134 Pantopoulos K & Hentze MW (1995) Rapid responses to oxidative stress mediated by iron regulatory protein EMBO J 14, 2917–2924 Schalinske KL & Eisenstein RS (1996) Phosphorylation and activation of both iron regulatory proteins 1 and 2... J, Geromanos S, Tempst P & Hartl FU (1993) Identification of nucleotide -binding regions in the chaperonin proteins GroEL and GroES Nature 366, 279–282 Chirico WJ, Waters MG & Blobel G (1988) 70K heat shock related proteins stimulate protein translocation into microsomes Nature 332, 805–810 Zimmerman DL & Walter P (1991) An ATP- binding membrane protein is required for protein translocation ATP binding. .. 42 43 44 45 regulatory protein 2 dominates iron homeostasis EMBO J 23, 386–395 Morikofer-Zwez S & Walter P (1989) Binding of ADP ¨ to rat liver cytosolic proteins and its influence on the ratio of free ATP ⁄ free ADP Biochem J 259, 117–124 Liberek K, Marszalek J, Ang D, Georgopoulos C & Zylicz M (1991) Escherichia coli DnaJ and GrpE heat shock proteins jointly stimulate ATPase activity of DnaK Proc... insight into the mechanism of nitration in vivo and its impact on IRP-1 functions J Biol Chem 279, 43345– 43351 Popovic Z & Templeton DM (2004) Iron accumulation and iron -regulatory protein activity in human hepatoma (HepG2) cells Mol Cell Biochem 265, 37–45 Kuhn LC (2003) Regulation of mRNA translation and ¨ stability in iron metabolism: is there a redox switch? In Cellular Implications of Redox Signaling... C & Danon A, eds), pp 327–360 Imperial College Press, London Pantopoulos K & Hentze MW (1998) Activation of iron regulatory protein-1 by oxidative stress in vitro Proc Natl Acad Sci USA 95, 10559–10563 Mimura CS, Admon A, Hurt KA & Ames GF (1990) The nucleotide -binding site of HisP, a membrane protein of the histidine permease Identification of amino acid residues photoaffinity labeled by 8-azido -ATP. .. J Biol Chem 271, 7168–7176 Pantopoulos K & Hentze MW (1995) Nitric oxide signaling to iron -regulatory protein: direct control of ferritin mRNA translation and transferrin receptor mRNA stability in transfected fibroblasts Proc Natl Acad Sci USA 92, 1267–1271 Pantopoulos K, Weiss G & Hentze MW (1996) Nitric oxide and oxidative stress (H2O2) control mammalian iron metabolism by different pathways Mol Cell... 1–13 Heck DE, Kagan VE, Shvedova AA & Laskin JD (2005) An epigrammatic (abridged) recounting of the myriad tales of astonishing deeds and dire consequences pertaining to nitric oxide and reactive oxygen species in mitochondria with an ancillary missive concerning the origins of apoptosis Toxicology 208, 259–271 Harris A (2002) Hypoxia ) a key regulatory factor in tumour growth Nat Rev Cancer 2, 38–47 . Inhibition of an iron-responsive element/iron regulatory protein-1 complex by ATP binding and hydrolysis Zvezdana Popovic and Douglas M supported by grants from the Heart and Stroke Foundation of Canada (grant T4134) and the Canadian Institutes of Health Research (grant MT11270). We thank Dr Tania