In its aconitase-inactive form, IRP-1 (iron regulatory protein-1)/cytosolic aconitase binds to the IRE (iron-responsive element) of several mRNAs to effect post-transcriptional regulation. We have shown previously that IRP-1 has ATPase activity and that binding of ATP suppresses the IRP-1/IRE interaction. In the present study, we characterize the binding activity further. Binding is observed with both [α-32P]ATP and [α-32P]ADP, but not with [γ-32P]ATP. Recombinant IRP-1 binds approximately two molecules of ATP, and positive co-operativity is observed with a Hill coefficient of 1.67±0.36 (EC50=44 μM) commencing at 1 μM ATP. Similar characteristics are observed with both apoprotein and the aconitase form. On binding, ATP is hydrolysed to ADP, and similar binding parameters and co-operativity are seen with ADP, suggesting that ATP hydrolysis is not rate limiting in product formation. The non-hydrolysable analogue AMP-PNP (adenosine 5′-[β,γ-imido]triphosphate) does not induce co-operativity. Upon incubation of IRP-1 with increasing concentrations of ATP or ADP, the protein migrates more slowly on agarose gel electrophoresis, and there is a shift in the CD spectrum. In this new state, adenosine nucleotide binding is competed for by other nucleotides (CTP, GTP and AMP-PNP), although ATP and ADP, but not the other nucleotides, partially stabilize the protein against spontaneous loss of aconitase activity when incubated at 37 °C. A mutant IRP-1(C437S) lacking aconitase activity shows only one ATP-binding site and lacks co-operativity. It has increased IRE-binding capacity and lower ATPase activity (Km=75±17 nmol/min per mg of protein) compared with the wild-type protein (Km=147±48 nmol/min per mg of protein). Under normal cellular conditions, it is predicted that ATP/ADP will maintain IRP-1 in a non-IRE-binding state.
- ATP hydrolysis
- co-operative binding
- iron-regulatory protein
- iron-responsive element
- nucleotide binding
IRPs (iron-regulatory proteins) exert post-transcriptional regulatory control of expression of proteins involved in iron homoeostasis. This control involves interaction of IRPs with functional IREs (iron-responsive elements) in the 5′ or 3′ untranslated regions of mRNAs. Two IRPs have been identified: IRP-1, which contains a [4Fe–4S] iron–sulfur cluster, and IRP-2, which does not [1,2]. IRP-1 is generally believed to interconvert between an enzymatically inactive IRE-binding state on the one hand, and a non-binding form with a [4Fe–4S] cluster and enzymatic aconitase activity on the other. A simple model for the mechanism by which IRP-1 can sense iron involves direct association of iron with an iron-depleted form of the protein to form a complete [4Fe–4S] cluster. However, it is well-established that the mitochondrion also plays an important role in switching IRP-1 from IRE-binding activity to active aconitase . One reason is that the mitochondrion is the site of cytosolic Fe–S cluster assembly. More than a dozen genes have been implicated in Fe–S cluster biogenesis, and homologues of some of them have been characterized in mammalian systems [4,5]. Disruption of cluster assembly in mitochondria switches aconitase activity of IRP-1 into IRE binding [6,7]. Evidence from yeast and mammalian cells suggests that the cell does not respond to total iron levels, but to the size of a regulatory iron pool or perhaps to the flux of iron through the Fe–S assembly pathway [3,8,9]. Furthermore, depleting ATP by uncoupling of oxidative phosphorylation in mitochondria prevents the switch of IRP-1 from the IRE-binding form to active enzyme with a [4Fe–4S] cluster [10,11]. In vitro, physiological concentrations of ATP inhibit IRE/IRP-1 binding both in cell extracts and with recombinant IRP-1. ADP has the same effect, in contrast with the non-hydrolysable analogue AMP-PNP (adenosine 5′-[β,γ-imido]triphosphate), indicating that in order to inhibit IRP-1-binding activity, ATP must be hydrolysed .
In addition to regulation of IRP-1 by iron levels, its IRE-binding activity is also affected by hypoxia, H2O2 and oxidative stress in general [12,13]. Each of these agents decreases the level of intracellular ATP, while increasing IRP-1-binding activity, suggesting a possible link between cellular iron levels and energy metabolism . Iron also modulates ATP levels, possibly through IRP-1-dependent translational control of mitochondrial aconitase . Oexle et al.  demonstrated increased mitochondrial oxygen consumption and ATP formation via oxidative phosphorylation in iron-supplemented K562 cells, and a subsequent reduction after iron deprivation. Furthermore, IRP-1 may be regulated by phosphorylation/dephosphorylation . In the present study we further investigate the nature of the interaction of ATP with recombinant IRP-1 in vitro. We demonstrate co-operativity of adenosine nucleotide binding, and present evidence for differential binding by different structural states of the protein.
Purification of recombinant IRP-1
Plasmids pT7-his-IRP-1  and pSG5-human IRP-1(C437S)  (gifts from Dr Kostas Pantopoulos, Lady Davis Institute for Medical Research, McGill University, Montréal, QC, Canada) were cut with Pst1/BstEII restriction enzymes. A 1178 bp Pst1/BstEII fragment from mutant pSG5-human IRP-1, encompassing the cysteine-to-serine mutation, was purified and inserted into a 4 kb Pst1/BstEII fragment of pT7-his-IRP-1 to give pT7-his-IRP-1(C437S). Both pT7 plasmids were grown in Escherichia coli BL21 cells. Human recombinant IRP-1 (rIRP-1) and rIRP-1(C437S) were purified from transformed E. coli as described previously  and eluted from Ni-NTA (Ni2+-nitrilotriacetate) agarose beads with 50 mM imidazole. Purified native and rIRP-1 appear to exist as an equilibrium of holo- and apoprotein forms with intermediate (e.g. disulfide-bonded) species also present [11,19,20]. In order to isolate holo- and apoprotein for binding studies, several protocols of chemical modification were evaluated. For the studies reported in the present paper, active aconitase (holoprotein) was produced by treating 50 μg of rIRP-1 in 100 mM Hepes (pH 7.4), 100 mM DTT (dithiothreitol), with 1 mM ferrous ethylene ammonium sulfate and 1 mM sodium sulfide for 1 h at room temperature under argon . The [4Fe–4S] cluster-free apoprotein was produced by treating 50 μg of rIRP-1 with 100 mM DTT under alkaline conditions (100 mM Tris, pH 8.8). Proteins were desalted on protein desalting spin columns (Pierce) and equilibrated with 24 mM Hepes (pH 7.6), containing 150 mM potassium acetate, 1.5 mM MgCl2 and 5% glycerol. IRE-binding activity was tested with an EMSA (electrophoretic mobility-shift assay) as described previously , using 0.5 μg of protein incubated with 30 ng of labelled IRE, and resolved on a 6% non-denaturing polyacrylamide gel.
Preparation of RNA transcripts
Transcription was performed in vitro with 1 μg of BamH1-linearized plasmid pSPTfer , coding the human ferritin H-chain IRE as described previously , in the presence of 50 μCi of [α-32P]CTP (800 Ci/mmol; ICN) and T7 RNA polymerase, using a Promega in vitro transcription system.
ATP-binding activity was determined by a filter-binding assay  with slight modifications. IRP-1 protein (1 μg) was incubated with [α-32P]ATP at room temperature (unless otherwise specified) for 30 min in 20 μl of binding buffer [25 mM Tris (pH 7.6), with 100 mM KCl, 5 mM MgCl2, 1 mM DTT and 10% glycerol] with different concentrations of ATP/Mg2+. Competition experiments were performed in the presence of different concentrations of unlabelled ATP, ADP, GTP, CTP or AMP-PNP. The nitrocellulose membrane was soaked briefly in 0.4 M KOH, rinsed thoroughly, incubated in binding buffer and mounted on a dot-blot apparatus. The binding reaction was stopped by addition of 75 μl of ice-cold binding buffer and loaded on to the dot blot (30 μl in triplicate wells). Each well was washed twice with 200 μl of ice-cold binding buffer. The membrane was dried, exposed to film, and each spot cut out and counted in a liquid scintillation counter.
Ni-NTA agarose-binding assay
In another set of experiments, nucleotide binding to recombinant IRP-1 was followed by Ni-NTA agarose binding using a method described by Obermann et al. . Ni-NTA agarose beads were pre-incubated (25 μl per sample) with 5 mM ATP/5 mM MgCl2 in 24 mM Hepes buffer (pH 7.6), with 150 mM potassium acetate, 5% glycerol and 0.4 M KCl, for 30 min at room temperature and washed twice with 24 mM Hepes (pH 7.6), 150 mM potassium acetate and 5% glycerol. IRP-1 (5 μg) was incubated with 100 nM [α-32P]ATP at room temperature for 30 min in 80 μl of binding buffer [25 mM Tris (pH 7.6), with 100 mM KCl, 1 mM DTT and 10% glycerol], with different concentrations of unlabelled ATP/Mg2+. The reaction mixture was mixed with ATP-saturated aliquots of Ni-NTA and rotated for 30 min at room temperature. Each tube was washed twice with 400 μl of 24 mM Hepes (pH 7.6), containing 150 mM potassium acetate and 5% glycerol, and eluted with 400 μl of 24 mM Hepes (pH 7.6), with 150 mM potassium acetate, 5% glycerol and 100 mM imidazole. A blank tube contained no protein. Eluates were counted by liquid scintillation.
IRE filter-binding assay
IRP-1 or IRP-1(C437S) (0.5 μg) was incubated with the indicated amount of radiolabelled probe in EB buffer [10 mM Hepes (pH 7.6), with 3 mM MgCl2, 40 mM KCl and 1 mM DTT] for 30 min, and applied to a nitrocellulose membrane following the ATP-binding protocol described above.
Preparation of [α-32P]ADP
[α-32P]ADP was prepared by hexokinase treatment of [α-32P]ATP (3000 Ci/mmol) (PerkinElmer). In total, 3 μl (10 pmol) of [α-32P]ATP was incubated at room temperature with 4 m-units/4 μl hexokinase (1 m-unit/μl; Sigma), with 0.11 M glucose, 10 mM MgCl2 and 30 mM Tris buffer (pH 8), for 60 min. Purity of [α-32P]ADP was confirmed by TLC. Aliquots (2 μl) were spotted on to polyethyleneimine cellulose TLC plates (Sigma), resolved using 0.5 M lithium chloride in 0.5 M formic acid, and visualized by autoradiography. Conversion into ADP was quantitative as determined by TLC.
Western blot analysis
For PAGE, treated IRP-1 is mixed with 50 mM Tris/HCl (pH 6.8), with 100 mM DTT, 2% SDS, 0.1% Bromphenol Blue and 10% glycerol, and boiled for 5 min. Proteins were resolved on acrylamide gels (4–15% gradient or 8%) run in 25 mM Tris, 192 mM glycine, with or without SDS. Separated proteins were transferred on to nitrocellulose membranes, incubated with an anti-IRP-1 antibody (Alpha Diagnostic International; 1:5000 dilution), and detected using ECL (enhanced chemiluminescence) Western blot detection reagents (Amersham Bioscience).
Non-denaturing gel electrophoresis
Agarose gel (0.8%) was made in 25 mM Tris/192 mM glycine and electrophoresis was carried out in the same buffer. Samples were loaded in 62.5 mM Tris/HCl (pH 6.8), with 40% glycerol and 0.01% Bromophenol Blue, and run at 80 V. After electrophoresis, gels were dried and exposed to X-ray film.
ATPase assays were performed by measuring the release of inorganic phosphate with Malachite Green amplification . Protein (1.25 μg) was incubated in 100 μl of 25 mM Tris/HCl (pH 7.6), containing 5 mM MgCl2, 0.02% Triton X-100, 1 mM DTT and 0–2 mM ATP, for 30 min at 37 °C. Then 600 μl of Malachite Green solution (0.17% Malachite Green and 1.05% ammonium molybdate in 8.2% HCl) was added for 1 min and the reaction stopped with 75 μl of 34% citric acid. Absorbance was measured at 660 nm and compared with standard phosphate solutions.
Aconitase activity was measured in a direct assay with 20 mM cis-aconitate as the substrate. The disappearance of aconitate was monitored spectrophotometrically at 240 nm, using a molar absorption coefficient ε=3600 M−1·cm−1 [26,27]. Aconitase activity is reported as the amount of substrate converted in μmol/min per mg of protein at pH 7.4 and 25 °C.
Recombinant protein in imidazole/Tris elution buffer was transferred to 10 mM potassium phosphate (pH 8.1) by passage through a PD10 Sepharose G-25 column (Amersham Biosciences) and then concentrated in Microcon TM10 tubes (Millipore) to approx. 3 μM. Triplicate spectra were recorded at 0.5 nm intervals in a J-810 spectropolarimeter (Jasco) thermostatically controlled at 20 °C, in a cuvette with a 0.2 cm path length.
Kinetic data and binding plots were fitted by non-linear regression using GraphPad Prism (GraphPad Software). Significance of Hill coefficients was determined using a one-sample Student's t test (GraphPad InStat).
ATP is hydrolysed after binding to rIRP-1
The presence of ATP at physiological concentrations (>1 mM) inhibits IRP-1–IRE complex formation observed by EMSA, and we demonstrated previously that IRP-1 binds ATP and has ATPase activity . To learn more about the IRP-1–ATP interaction, we performed nucleotide-binding experiments with rIRP-1 using a nitrocellulose filter-binding assay. At a concentration of 10 μM ATP, binding of [α-32P]ATP is readily detected, whereas binding of label from [γ-32P]ATP is negligible (Figure 1A). At 25 °C, binding of [α-32P]ATP reaches a plateau over 1 h, whereas [γ-32P]ATP is still not observed (Figure 1B). This suggests that bound ATP is hydrolysed rapidly upon binding.
ATP binding increased with IRP-1 concentration, saturating at approx. 1 μM protein (results not shown). The binding is effectively pH-independent across a physiological range, at both 10 μM and 100 μM ATP (Supplementary Figure S1A at http://www.BiochemJ.org/bj/430/bj4300315add.htm). It requires the presence of Mg2+, as substitution of MgCl2 in the binding buffer by 10 mM EDTA eliminates binding, and addition of Mg2+ restores it (results not shown). Further studies were carried out at pH 7.6 in stoichiometric solutions of ATP/Mg2+.
To rule out filter binding as an artefact, e.g. due to precipitation of IRP-1 and non-specific binding of ATP to the precipitate on the filter membrane, binding was measured in an independent assay. IRP-1 was incubated with [α-32P]ATP and various concentrations of unlabelled ATP up to 500 μM. IRP-1 was then bound through its His tag to Ni-NTA agarose. Subsequent elution with 100 mM imidazole detached His-tagged IRP-1 with bound [α-32P]ATP (Supplementary Figure S1B). IRP-1 bound to Ni-NTA agarose beads binds ATP with a Bmax of 3.74±0.78 and a Kd of 74.4±49.8 μM (R2=0.83), comparable with Bmax=4.6±0.3 and Kd=86±17 μM reported previously . Additional evidence of specificity comes from the absence of binding when BSA was substituted for IRP-1 in the filter-binding assay (results not shown).
ATP-binding characteristics change with ATP concentration
We further characterized the ATP-binding interaction by increasing the concentration of unlabelled ATP while maintaining the concentration of [α-32P]ATP at 50 nM (Figures 2A and 2B). A 2–10-fold excess of unlabelled ATP partially competed with binding of labelled ATP. However, at 1.25 μM, ATP binding to IRP-1 changed dramatically (Figure 2A), increasing approx. 50-fold. Analysis of IRP-1 binding to ATP in the 0.5–200 μM range gave Bmax=1.88±0.16 and Kd=54.8±12.5 μM (R2=0.73). The corresponding Hill slope was 1.32±0.26 (R2=0.86) (Figures 2C and 2D), suggestive of co-operative binding.
IRP-1 binds to ADP and ATP with similar characteristics
Previously  we demonstrated that ADP has the same inhibitory effect on the IRP-1–IRE complex as ATP. Therefore we studied the effect of ADP on [α-32P]ATP binding in filter-binding assays (see Supplementary Figures S2A–S2C at http://www.BiochemJ.org/bj/430/bj4300315add.htm). Again, we detected competition of radioligand binding at low concentrations of unlabelled ADP, and at approx. 1 μM ADP binding again increased. Non-linear regression analysis of [α-32P]ATP binding in the presence of ADP gave an apparent Bmax of 2.37±0.30 with a Kd of 121±30 μM (R2=0.89). Again, a Hill slope of 1.39±0.35 (R2=0.88) was consistent with co-operative binding.
Detailed examination of Hill plots
Representative experiments reported above give Hill slopes of 1.32 for ATP and 1.39 for ADP, corresponding to the illustrated binding curves (Figure 2 and Supplementary Figure S2). These indicate co-operative binding and are consistent with the sharp increase in binding above 1 μM ATP. In fact, analysis of multiple data sets over several months strengthened this conclusion. Seven data sets for ATP binding were analysed and two failed to achieve a plateau. Of the remaining five, S.D.s on the Hill coefficient ranged from 6 to 14%, and the mean value of 1.60±0.37 was significantly greater than 1.0 (P=0.02). With ADP, six data sets were re-analysed. One showed no plateau and one showed a S.D. of 54%. Eliminating these two, the remaining four data sets gave a Hill coefficient of 1.67±0.36, greater than 1.0 (P=0.03). These composite Hill coefficients were associated with EC50 values of 44±22 and 40±15 μM for ATP and ADP respectively. These values indicate that adenine nucleotide binding to IRP-1 shows strong positive co-operativity above 1 μM, and the binding does not discriminate between ATP and ADP.
Binding to chemically prepared holo- and apo-protein
Holo- and apo-protein were prepared as described in the Experimental section and were characterized by aconitase activity, EMSA and Western blotting. Both holoprotein and rIRP-1 showed low IRE binding that was substantially increased by treatment with 2-mercaptoethanol (Figure 3A), which is thought to disrupt the [4Fe–4S] cluster and convert aconitase into the binding form. In contrast, the apoprotein preparation shows full binding activity. On the other hand, production of the apoprotein completely eliminates aconitase activity, whereas reconstitution leads to an approximate 3-fold increase (Figure 3C).
Two independent preparations from a separate isolation of rIRP-1 were used to study ATP and ADP binding (Table 1). The new preparation of rIRP-1 showed Bmax values and Hill slopes consistent with the above analyses, again indicating two binding sites for both ATP and ADP, with co-operative binding. These properties were retained for both nucleotides in the holoprotein, with an indication of a possible additional site and even stronger co-operativity in the apoprotein. However, higher Kd values suggested weaker binding, particularly with the apoprotein (315±49 μM and 422±87 μM with the apoprotein for ATP and ADP respectively; and 115±3 μM and 153±37 μM with the holoprotein). These results indicate that, whereas both aconitase and RNA-binding activity are present in a relatively low portion of rIRP-1 as isolated, both properties can be restored. Furthermore, both holo- and apo-protein appear to contribute to the co-operative binding of nucleotides at two sites in the concentration range 0–200 μM. Subsequent experiments with unmodified rIRP-1 are justified in order to minimize possible conformational changes caused by the chemical treatments used to isolate the holo- and apo-protein species, as suggested by the higher Kd values.
Hydrolysis is required for ATP-induced co-operative binding
Binding of label from [α-32P]ATP, but not [γ-32P]ATP, to IRP-1 suggests that hydrolysis is necessary for binding. To confirm this, we attempted to suppress hydrolysis by performing the binding reaction at 4 °C. However, neither label was bound under these circumstances (Supplementary Figure S3A at http://www.BiochemJ.org/bj/430/bj4300315add.htm). This could be because of a loss of hydrolytic activity, if hydrolysis is indeed a requirement for binding, or because of a change in conformation that disrupts the binding site. Therefore we measured binding of [α-32P]ADP (prepared by the reaction of [α-32P]ATP with hexokinase) directly. [α-32P]ADP binding saturated in a time similar to [α-32P]ATP (compare Figure 1B and Supplementary Figure S3B), but with a lower Bmax=1.04 and a Kd=42.4 μM (Supplementary Figure S3C). [α-32P]ADP also fails to bind at 4 °C (Supplementary Figure S3A). Taken together, these data suggest that the primary binding site is indeed occupied by ADP, but that ATP hydrolysis allows occupation of this site. Subsequent experiments reported below are performed with [α-32P]ATP.
The ATP analogue AMP-PNP, which is non-hydrolysable between the β and γ phosphorus atoms, had no effect on IRE binding either in cell extracts or with recombinant protein . This further implies that ATP must be hydrolysed in order to modulate the IRP–IRE interaction, but it remains unclear whether the energy derived from hydrolysis is important, or rather the product, ADP, is sufficient to inhibit binding of IRP-1 to IRE. To address this, we incubated rIRP-1 with a non-hydrolysable analogue, AMP-PNP, and measured binding of [α-32P]ATP in the filter-binding assay. In contrast with ATP and ADP, AMP-PNP does not produce co-operative radioligand binding to IRP-1 (Figure 4A). This further suggests that the product ADP is required to effect a change in the IRP-1 conformation that is unfavourable for IRP-1–IRE complex formation. CTP and GTP also fail to show co-operativity with ATP binding (Figure 4A). Absolute binding (c.p.m.) of [α-32P]ATP is increased approx. 10-fold by 1 μM unlabelled ATP or ADP, despite the decreasing radiospecific activity with increasing addition of non-radiolabelled nucleotide (Figure 4B). AMP-PNP, CTP and GTP do not increase counts above background. However, when [α-32P]ATP was bound in the presence of 5 μM unlabelled ATP (i.e. above the concentration effecting enhanced nucleotide binding), and then excess unlabelled nucleotide was added subsequently to compete with the label, the nucleotides ATP, ADP, AMP-PNP and CTP were equally effective in competition. We conclude that a conformational change brought about by ADP renders the protein susceptible to competitive binding by the other nucleotides. Surprisingly, GTP was highly competitive under these circumstances, competing with ADP-induced binding at 5 μM (Figure 4C).
ATP and ADP binding change the IRP-1 structure
We previously demonstrated the binding of ATP to IRP-1 by photolabelling with 8-azido-[α-32P]ATP . In the present study we confirmed nucleotide binding after incubation of IRP-1 with [α-32P]ATP in the presence of unlabelled ATP and separation by SDS/PAGE. Radioligand binds to the major band of rIRP-1 on a silver-stained gel, as seen on autoradiography, and this band is further identified as IRP-1 by Western blotting with an anti-IRP-1 antibody (Figure 5A).
The above radionucleotide-binding experiments indicate the allosteric nature of ATP–IRP-1 interactions, and furthermore indicate that ADP binding produces a state of IRP-1 that does not bind IRE. The change in the binding state of IRP-1 occurs at approx. 1 μM AT(D)P when binding of radionucleotide sharply increases. We attempted to visualize the ATP-modified bound form(s) by means of non-denaturing agarose gel electrophoresis. Native electrophoresis indeed suggests changes of IRP-1 structure (Figure 5B), in that an increase of non-radiolabelled ATP concentration results in species that move slower into the agarose gel, opposite to the expectation if the mobility differences were due to increased negative charge on binding AT(D)P. As in the filter-binding assay, bound [α-32P]ATP is not competed out by addition of unlabelled ATP up to 4–6 μM, consistent with co-operatively enhanced binding in the filter-binding assay. The shift in electrophoretic mobility was already apparent at 0.5 μM ATP, somewhat below the threshold concentration for enhanced ATP binding, and plateaus at approx. 10 μM. No 32P remained on the gel when radioligand and non-radiolabelled ATP were incubated with BSA, IRE or without protein (results not shown), confirming specific binding to IRP-1.
Addition of SDS to the loading buffer prevented the apparent conformational change and, irrespective of the concentration of ATP, the IRP-1–ATP complex migrated at the same position in the presence of SDS (results not shown). No complex of radioligand with IRP-1 was detected when IRP-1 was incubated with [γ-32P]ATP, or with [α-32P]ATP in the presence of excess unlabelled AMP-PNP or CTP (results not shown). From the previous experiments one would predict that ADP, too, would induce a structural change of IRP-1, and this is indeed the case (see Figure 8B, discussed below).
To determine that these shifts in agarose electrophoretic mobility truly arose from a conformational change, we recorded CD spectra of the protein in the absence and presence of ATP (Figure 6A). The native protein shows a minimum at approx. 208 nm, consistent with an estimated α-helical content of approx. 25% . On titration with ATP, a progressive loss in optical rotation, θ, between 208 and 230 nm is seen from 5 to 50 μM ATP. In this concentration range, ATP does not contribute significantly to the spectrum (Figure 6A, inset). Tracings of parallel samples of IRP-1 treated with either ATP or ADP (each 50 μM) show identical spectral shifts in the range 208–230 nm (Figure 6B).
Cytosolic aconitase activity requires an intact [4Fe–4S] cluster, and is conformation- and temperature-sensitive, losing up to 30% of enzyme activity after 16 h at 0 °C . We measured the aconitase activity of IRP-1 held on ice compared with that incubated at 37 °C, a process that leads to loss of the rather unstable aconitase activity. We then determined whether inclusion of various nucleotides could stabilize aconitase activity. In a typical experiment where 70% of the control activity was lost after 30 min of incubation, both ATP and ADP offered partial protection (approx. 50–60% of full activity, P<0.01) (Figure 7). No protection was achieved with CTP, GTP or the non-hydrolysable AMP-PNP.
IRP-1(C437S) shows diminished ATP binding and hydrolysis
Mutation of Cys437 eliminates a ligand of the [4Fe–4S] cluster and destabilizes cluster formation [30,31]. Wild-type IRP-1 and IRP-1(C437S) purified on Ni-NTA show a similar pattern on SDS/PAGE and Western blots (Figure 8A). As expected [30,31], the mutant shows significantly greater IRE binding (Figure 9A) and absent aconitase activity (results not shown). However, its ATP binding is significantly diminished (Figure 9B); whereas IRP-1 binds up to 4 mol of ATP (Bmax=3.97, Kd=66 μM), the mutant binds only 1 mol of ATP (Bmax=1.12, Kd=123 μM). Furthermore, the mutant fails to show co-operative adenosine nucleotide binding (results not shown). ATP hydrolysis proceeds at a lower rate with the mutant (Vmax=75±17 nmol/min per mg of protein compared with 147±48 nmol/min per mg of protein for wild-type), and at a Km of 0.18±0.09 mM for mutant compared with 0.30±17 mM for wild-type (Figure 9C); however, 5 mM ATP, but not GTP, is able to suppress IRE binding of both wild-type and mutant proteins (Figure 9D). Whereas wild-type protein is shifted on non-denaturing agarose gels in the presence of both ATP and ADP, such a structural change is not evident with the mutant protein (Figure 8B).
In the present study we confirm that rIRP-1 binds ATP and ADP, and upon binding undergoes structural changes that influence its electrophoretic mobility and CD spectrum. At ATP concentrations in the nanomolar range, IRP-1 has a low affinity for ATP that remains up to approx. 1 μM ATP. Bound ATP is hydrolysed, as demonstrated by failure to detect any significant [γ-32P]ATP binding. This can be explained by the ATPase activity of IRP-1 . However, increasing ATP concentration to the micromolar range and above has a co-operative effect on IRP-1 binding. Mg2+ is required for ATP binding, in agreement with previous findings that Mg2+ is important for IRP-1 activation . In the micromolar ATP range, binding is strongly dependent on protein concentration, a characteristic of allosteric proteins. Co-operativity is usually the result of conformational change, and indeed addition of ATP/ADP results in an altered structure visible as discrete forms of the IRP-1–ATP complex with decreased mobility on native agarose gels, and a shift in the CD spectrum in the 5–50 μM range. Co-operative binding permits a much more sensitive response to nucleotide concentration.
Previously, we demonstrated that the IRP-1–IRE complex is absent in the presence of ATP . Inhibition of complex formation can be achieved with either ATP or ADP, suggesting that the nucleotides stabilize the protein in a non-IRE-binding form. This effect was not observed with non-hydrolysable analogues , and again in the present study the non-hydrolysable analogue AMP-PNP does not bind to IRP-1; and it does not induce a co-operative effect, does not compete with ATP/ADP (Figure 4B) and does not form shifted complexes on native agarose gels (results not shown). Nucleotides may stabilize non-IRE-binding form(s), and ATP and ADP provide at least temporary protection against loss of aconitase activity upon incubation at 37 °C. However, co-operative binding of the hydroysable nucleotides is conserved in both the chemically prepared holo- and apo- forms of the protein, and in the apoprotein there is a tendency for increased binding and co-operativity. Thus nucleotide binding may stabilize both states of the protein independent of a switch between them.
It has been estimated that only 1–5% of IRP-1 is in the RNA-binding form in tissues [6,33,34]. Consistent with this, we found that purification of IRP-1 on an IRE-affinity column gave a relatively low yield of protein with a Vmax for ATP hydrolysis of only 3.4 nmol/min per mg of protein . In contrast, total rIRP-1 purified on Ni-NTA is found in the present study to have a Vmax 40-fold higher (147 nmol/min per mg of protein). This indicates that the IRE-binding form is a relatively minor component, and is consistent with a large increase in binding upon preparation of the apo-form (Figure 3A). The C437S mutant strengthens the argument that IRE binding is associated with an aconitase-deficient state, whereas ATP binding and hydrolytic activity are greater in the non-IRE-binding state. Increased IRE-binding activity of the mutant is associated with a Vmax for ATP hydrolysis only half that of wild-type protein (75 nmol/min per mg of protein). Loss of one nucleotide-binding site in the mutant protein, but not the apoprotein preparation, indicates additional structural changes in the mutant beyond those due to the absence of the [4Fe–4S] cluster. Despite retention of only a single site, ATP binding is still able to block interaction with IRE (Figure 9D).
Strong competition by GTP with ATP binding occurs only in the protein structure that shows enhanced binding at higher ADP concentration. In this context, it is interesting to note that ATP and GTP are required for efficient biogenesis of the [4Fe–4S] cluster of mitochondrial aconitase [35,36], and a possible role of the structurally altered form of cytosolic aconitase in responding to GTP levels is intriguing.
Values of Kd found for ATP by filter binding (54.8±12.5 μM) and to protein bound on Ni-NTA agarose (74.4±49.8 μM) are consistent with that reported previously for IRE-purified protein (86±17 μM) , and found in the present study for ADP (121±30 μM). Total cellular ATP levels of approx. 5 mM  imply that IRP-1 would be saturated with ADP under normal circumstances, and stabilized in a non-IRE-binding aconitase-active form. If this is true, it could account for the studies noted above showing that only a small portion of cellular IRP-1 is in the IRE-binding form [6,33,34], and the evidence from IRP-1 gene-ablated mice that IRP-2, and not IRP-1, is responsible for post-transcriptional regulation of iron homoeostasis .
IRP-1 lacking the [4Fe–4S] cluster is degraded through ubiquitin ligase by the same mechanism as IRP-2 [38,39]; in iron-replete cells the [4Fe–4S] form should be stable. If a pool of apoIRP-1 exists, it should be protected (e.g. by oligomerization or compartmentalization), possibly through ATP-induced conformational changes. Cytosolic aconitases from both bacteria and mammals have a tendency to oligomerize. E. coli aconitase AcnB forms a homodimer both in vitro and in vivo, and the monomer–dimer transition is dependent on the availability of iron . Only the monomer can act as a post-transcriptional regulator  binding IRE. Yikilmaz et al.  found that purified human recombinant apo-IRP1 exists in a slowly reversible monomer–dimer self-association equilibrium. They showed that binding of IRE drives a conformational change toward a complex in which IRP1 is entirely monomeric.
Interactions of IRP-1 with ATP/ADP occur in two nucleotide concentration regimes. On the one hand, concentrations in the 1–10 mM range block RNA binding  and stabilize aconitase activity (Figure 7). On the other hand, concentrations of 1–10 μM produce a shift in conformation dectectable by CD and native agarose gel electrophoresis, and encompass a transition reflecting positive co-operativity in adenosine nucleotide binding. Typical cellular ATP levels at approx. 5 mM [37,41] imply a role in regulation of the IRE-binding/aconitase transition. However, the binding constants of AT(D)P ( and the present study) imply saturation of binding sites at this concentration, which may explain why very little IRP-1 is in the IRE-binding form and may function preferentially as a true cytosolic aconitase.
In the lower concentration range, it is unclear how micromolar nucleotide levels could effect functional changes in the comparatively ATP-rich environment of the cell. Nevertheless, it is true that many biological phenomena are regulated in this concentration region. Documented effects of ATP on actin polymerization are reported to have a half-maximal effective concentration (EC50) of 200 nM in U937 cells, and plateau at 100 μM . The EC50 for stimulation of ERK (extracellular-signal-regulated kinase) phosphorylation leading to proliferation in cultured smooth muscle cells is 2–3 μM ATP . The Km values of many ATPases are in the micromolar range, suggesting again that regulation at this level is relevant. Lower local concentrations of ATP/ADP may occur, e.g. due to compartmentalization. At least 50% of cytosolic ADP is protein-bound . Thus whether IRP-1 experiences sub-saturating cytosolic ATP concentrations in the cell remains an open question. Certainly our estimated Km value of 5.3 μM  is in keeping with Km values reported for other ATP-hydrolysing proteins, e.g. E. coli DnaK (20 μM) , Hsc70 (heat-shock cognate 70 stress protein) (1.4 μM)  and F1-ATPase (15 μM) . Furthermore, these proteins have Vmax values in the range 1.1–3.5 nmol/min per mg of protein, comparable with the value of 3.4 nmol/min per mg of protein reported for IRP-1 , supporting a physiological relevance to the observed high-affinity ATP binding.
Finally, high ATP concentrations that suppress the IRE–IRP-1 complex would favour ferritin translation. Concomitantly, high ATP inhibits oxidative phosphorylation , and the newly synthesized ferritin would be available to shelter excess O2 and iron . In a period of high cytosolic ATP and attenuated oxidative phosphorylation, surplus citrate from the tricarboxylic acid cycle is exported from mitochondria and utilized for lipid, glutamate and NADPH production. It has been suggested that cytosolic aconitase is involved in these metabolic processes [5,49,50].
Zvezdana Popovic and Douglas Templeton contributed to the experimental design, data interpretation and to the writing of the paper.
This work was supported by the Canadian Institutes of Health Research [grant number MT11270 (to D. M. T.)].
Abbreviations: AMP-PNP, adenosine 5′-[β,γ-imido]triphosphate; DTT, dithiothreitol; EMSA, electrophoretic mobility-shift assay; IRE, iron-responsive element; IRP, iron regulatory protein; Ni-NTA, Ni2+-nitrilotriacetate; rIRP-1, recombinant IRP-1
- © The Authors Journal compilation © 2010 Biochemical Society