Abstract
The first step of iron biomineralization mediated by ferritin is the oxidation at the ferroxidase active site of two ferrous ions to a diferric oxo/hydroxo species. Metal-loaded ferritin crystals obtained by soaking crystals of frog ferritin in FeSO4 and CuSO4 solutions followed by flash freezing provided X-ray crystal structures of the tripositive iron and bipositive copper adducts at 2.7 Å and 2.8 Å resolution, respectively. At variance with the already available structures, the crystal form used in this study contains 24 independent subunits in the asymmetric unit permitting comparison between them. For the first time the diferric species at the ferroxidase site is identified in eukaryotic ferritins. Anomalous difference Fourier maps for crystals (iron crystal 1) obtained after long soaking times in FeSO4 solution invariantly showed diferric species with a Fe-Fe average distance of 3.1 ± 0.1 Å, strongly indicative of the presence of a μ-oxo/hydroxo bridge between the irons; protein ligands for each iron ion (Fe1 and Fe2) were also unequivocally identified and found to be the same in all subunits. For copper bound ferritin, dicopper(II) centers are also observed. While copper at site 1 is essentially in the same position and has the same coordination environment as Fe1, copper at site 2 is displaced towards His54, now acting as a ligand; this results in an increased intermetal distance (4.3 ± 0.4 Å). His54 coordination and longer metal-metal distances might represent peculiar features of divalent cations at the ferroxidase site. This oxidation-dependent structural information may provide key features for the mechanistic pathway in high eukaryotic ferritins that drives uptake of bivalent cation and release of ferric products at the catalytic site. This mechanism is supported by the X-ray picture obtained after only 1 minute of soaking in FeSO4 solutions (iron crystal 2) which reasonably contains the metal at different oxidation states. Here two different diiron species are trapped in the active site, with intermetal distances corresponding to those of the ferric dimer in crystal 1 and of the dicopper centers and corresponding rearrangement of the His54 side chain.
Keywords: ferritin, X-ray, iron, copper, reaction mechanism
INTRODUCTION
Ferritins are intracellular proteins capable of storing thousands of iron(III) ions as a solid mineral.1 The ferritin molecule is a multimeric system formed by 24 subunits that assemble in such a way to form an external cage around the biomineral storage cavity (Fig. 1).2 The subunits are characterized by a four helix bundle structure. Iron(II) enters the protein shell through ion channels and reaches the active site, known as the ferroxidase site, constituted by a dinuclear metal ion binding site. Here, the two substrates, iron(II) and O2, are coupled to obtain iron(III) mineral precursors.2-5 The ferroxidase site functions as a gated site for the transfer of ferric products into the cavity following oxidation.6;7 Here a structural characterization is reported for frog ferritin, a particular system constituted by 24 identical subunits, called M, subunits and compared with the already available structural information on similar systems.
Figure 1.

Two different views of the nanocage structure of ferritin obtained for Fe crystal 2: (a) centered on the C3 pore and (b) centered on the C4 pore. In c, the structure of a single 4-helix bundle subunit is shown together with helix numbering. In all panels, orange spheres represent iron atoms.
Studying iron in ferritins is difficult because of the transient nature of the metal binding at this low affinity site (Kd of the order of 10-5 M)8 which has prevented up to now a full understanding of the iron binding by X-ray crystallography. Most of the structural information about iron-binding has been inferred either from spectroscopy9 or through metal substitution.2-4 X-ray crystal structures of iron-bound ferritins available for some organisms reflect the transient nature of iron-protein interactions.
In anaerobes, ferritin has been proposed to protect the organisms from traces of oxygen and its partially reduced derivatives.10 The structure of ferritin from the anaerobe Pyrococcus furiosus11 is paradigmatic, with three iron ions detected under aerobic conditions (Fig. S1): an iron site A appears to be the highest affinity site since it is found occupied in the isolated protein, while site B and site C are found occupied only after soaking under aerobic conditions. The closest iron-iron distances have been found between sites A and B and vary between 2.6 and 3.6 Å in the different subunits. Among eukaryotic proteins, iron has been observed only at the ferroxidase site of marine pennate diatoms Pseudo-nitzschia multiseries.12 This ferritin is largely divergent from those of other eukaryotes. A putative iron site similar to the A site P. furiosus is found occupied by a water molecule (Fig. S1). An iron atom is observed bound to four glutamate residues in a site that is close, but not coincident, with site B of P. furiosus. Other two iron ions are positioned towards the protein core; neither of them corresponds to site C of P. furiosus. This variability of binding sites possibly reflects the transient iron binding and gives the perception that further efforts are needed to attempt an understanding of the system.
No iron-bound crystal structure is available for ferritin from higher eukaryotes. The only structural data available on the ferroxidase site have been obtained from X-ray absorption spectroscopy on the present system.9 In that study three different states have been trapped by rapid freeze quench of iron(II) ferritin reacting with O2.9 An initial diferrous, high-spin state is described with iron being in an average 6-coordinated environment with a Fe-Fe distance of 3.44 Å. After 25 ms, a peroxodiferric intermediate has been described with the very short Fe-Fe distance of 2.53 Å, which rapidly (in about 1 min) evolves towards a diferric μ-oxo/hydroxo species with a Fe-Fe distance of 3.00 Å.9
A better characterization of the iron site is available for the ferritin-like proteins called bacterioferritins. They are 24-mer oligomeric proteins containing both binuclear iron centers and b-type hemes. Their primary functions have been proposed to be detoxification of iron or protection against O2 and its radical products.13 At variance with ferritins, it is generally accepted that in bacterioferritins the iron at the catalytic center is stably bound,2;14 acts as a cofactor and displays an enzymatic role in the mineralization process.2;14 As a consequence of this stability, dinuclear iron clusters have been more easily observed by X-ray crystallography at the catalytic site of bacterioferritins. The di-iron species obtained by soaking (< 15 min) apobacterionferritin in iron(II) solutions under aerobic conditions, followed by flash freezing, was proposed to be in the reduced state. The structure of the metal center has a symmetric architecture, with one His and one Glu residue per iron and two bridging Glu residues.14;15 The metal-metal distance is about 4 Å in P. aeruginosa bacterioferritin15 and 3.7 Å in E. coli bacterioferritin.14 In the latter system, longer soaking times provided a metal center proposed to be a μ-oxo/hydroxo di-iron(III) species, with intact coordination sphere as far as the protein residues are concerned,14 and an inter-metal distance of 3.6 Å, which is similar to that of other di-iron enzymes. 11;12;16-19 In P. aeruginosa bacterioferritin the ferroxidase site was found devoid of iron upon oxidation.15
We report here crystal structures of iron loaded frog M ferritin determined by flash freezing crystals soaked for different times in iron(II) solutions under aerobic conditions. These structures provide the first X-ray picture of iron(III) products at the ferroxidase site in higher eukaryotes ferritins.
With very short soaking times (1 min) two different iron-protein binding modes at the catalytic site have been obtained. With the same approach the structure of the copper(II)-bound form has been solved, which may represent a model for the short-lived iron(II)-ferritin, i.e. the reactant species.
All together these structures provide a picture of the mechanism for the reaction at the ferroxidase site: here iron oxidation by dioxygen is accompanied by a decrease in the intermetal distance and in the number of protein residues involved in metal coordination that facilitate release of the products as di-iron clusters.
EXPERIMENTAL PROCEDURES
Crystallization
Frog M ferritin was expressed and purified as previously described.3 Crystals suitable for X-ray crystallography were grown using different procedures. One type of Fe-ferritin crystal (Fe crystal 1) was prepared from an iron pre-loaded protein solution using the sitting drop vapor diffusion technique at 20 °C. Iron loading was performed in aerobic conditions by adding a freshly prepared aqueous solution of FeSO4 · 7 H2O to a 14 mg/mL demineralized ferritin solution in 20 mM TrisHCl at pH 7.5, in the approximate ratio of 2 irons/subunit. Drops were prepared by mixing equal volumes (2 μL) of iron-loaded ferritin solution and a precipitant solution of 2 M sodium formate. Crystals of the iron-loaded frog M ferritin grew in 4-7 days as rust-orange prisms. Further iron loading of the crystals was done by adding 1 μL of a freshly prepared 60 mM FeSO4 · 7 H2O solution to the crystallization drop (iron concentration into the drop: ≈ 10 mM or stoichiometric ratio: ≈ 45 irons/subunit). After 90 min from the iron addition the crystals were transferred to a cryoprotectant solution (made of 20% ethylene glycol and 80% precipitant solution) and flash frozen in liquid nitrogen.
A second type of iron-loaded crystal (Fe crystal 2) was obtained from demineralized frog M ferritin (14 mg/mL solution in 20 mM TrisHCl at pH 7.5) using 2.1 M sodium formate as precipitant in a sitting drop setup (drops made of 4 μL protein + 2 μL precipitant). 1 μL of a 64 mM FeSO4 · 7 H2O solution was added to the 6 μL drop containing the demineralized crystal, obtaining an approximate Fe/subunit ratio of 24. After 1 min the crystal (Fe crystal 2) was fished in a cryoloop, transferred to the cryo-solution (see above) and frozen in liquid nitrogen. At the moment of freezing the crystal had a much paler orange color with respect to Fe crystal 1.
The copper loaded crystals were prepared by crystallizing a 7 mg/mL demineralized ferritin solution in 20 mM TrisHCl at pH 7.5 with 2.5 M sodium formate as precipitant in the same crystallization setup as the iron-loaded ferritin. Drops were prepared by mixing 4 μL of the above protein solution and 2 μL of precipitant. The drops containing the demineralized ferritin crystals were then treated with 1 μL of a 20 mM CuSO4 · 5 H2O solution (approximate Cu/subunit ratio of 16). After 24 hours the crystals have been transferred to a cryoprotectant solution (made of 20% ethylene glycol and 80% 3M sodium formate) and flash frozen in liquid nitrogen.
Data Collection, Structure Solution and Refinement
Data collection and refinement statistics for the three structures are reported in Table S1.
A complete anomalous dataset of the iron-loaded frog M ferritin was collected from crystal type 1 at the ESRF (European Synchrotron Radiation Facility, Grenoble, France) on beamline ID14-4 at the Zn-K edge (λ = 1.2915 Å, the longest available wavelength on that beamline). Images were collected at 100 K on the CCD detector ADSC Quantum Q315r. The data were integrated using the program MOSFLM 7.0.420 and scaled with SCALA21from the CCP4 suite.22 Fe crystal 1 belonged to the trigonal space group P3121 with cell parameters a = 210.72 Å, b = 210.72 Å, c = 324.41 Å. The structure was solved by molecular replacement, using the software MOLREP.23 The unit cell size of the trigonal crystals strongly indicated that an entire ferritin molecule (24 subunits) was present in the asymmetric unit. A rough packing model was obtained by treating each single molecule as a featureless spherical object with a radius of ≈130 Å. An entire 24mer of frog M ferritin was used as searching model (PDB code 1MFR) for the rotation and translation functions24 after excluding non-protein atoms and water molecules. The positions of the iron ions were determined from anomalous difference Fourier maps calculated using the program FFT from the CCP4 suite.22 The anomalous signals corresponding to Fe were between 3.5 σ and 10.0 σ in the anomalous difference maps.
The structure was refined with the program REFMAC.22;25 The refinement protocol involved a sequence of iterative manual rebuilding of the model and maximum-likelihood refinement.
The molecular graphics software Coot26 was used to visualize the structures and for manual rebuilding and modeling of the missing atoms into the electron density. Water molecules were added using the program ARP/wARP.27 The final model has been inspected manually and checked with the programs Coot26 and PROCHECK.28 The final model has good stereochemistry with 97.2% of residues in the favored regions and 2.8% in the additionally allowed and none in the disallowed regions of the Ramachandran plot.
A complete anomalous diffraction data set was collected on the iron-loaded frog M ferritin crystal type 2 at the ESRF beamline ID29 at the Fe-K edge (λ = 1. 7389 Å, T = 100 K). Images were collected on a PILATUS 6M detector. Data reduction and structure solution were performed using the same procedures described above. The crystal belongs to the trigonal space group P3121 with cell parameters a = 210.74 Å, b = 210.74 Å, c = 322.04 Å and with a whole frog M ferritin 24mer in the asymmetric unit. The positions of the iron ions were determined from anomalous difference Fourier maps and the anomalous signals corresponding to Fe were between 5.0 σ and 10.5 σ. The structure was refined with the program PHENIX29 using a protocol similar to that described above. The final model has good stereochemistry with 97.2% of residues in the favored regions and 2.8% in the additionally allowed and none in the disallowed regions of the Ramachandran plot.
A complete anomalous data set on the Cu-loaded frog M ferritin was collected at the ESRF (European Synchrotron Radiation Facility, Grenoble, France) on beamline ID29 at the Cu-K edge wavelength (λ = 1.378 Å, T = 100 K). Images were collected on a PILATUS 6M detector. Data reduction has been performed as above. Copper-loaded frog M ferritin belongs to the trigonal space group P3121 with cell parameters a = 210.21 Å, b = 210.21 Å, c = 322.80 Å and with a whole frog M ferritin 24mer in the asymmetric unit. The structure was solved by molecular replacement, similarly to the previous structures. The positions of the copper ions were determined from anomalous difference Fourier maps. The anomalous signals corresponding to Cu ions were between 5.5 σ and 11.0 σ in the anomalous maps. The structure was refined with the program REFMAC22;25 and validated by the same protocol used for Fe crystal 1. The final model has good stereochemistry with 96.9% of residues in the favored regions and 3.1% in the additionally allowed and none in the disallowed regions of the Ramachandran plot.
All the figures were generated using CCP4MG.30
The crystal form used for diffraction experiments contains one whole frog M ferritin molecule per asymmetric unit giving 24 independent copies of each subunit. This allowed us to obtain the standard deviations estimates (esd) of the differences in distances within the 24 subunits. The esd on the Fe-Fe (Fe crystal 1) and on Cu-Cu distances obtained in this way is 0.1 Å and 0.4 Å, respectively. The estimated average error on the coordinates from crystallographic refinement is 0.2 Å for the iron adduct (Fe crystal 1), 0.4 Å for the second iron adduct (Fe crystal 2) and 0.2 Å for the copper adduct.29;31 However, the actual error on the atomic positions of the metal ions is lower, although difficult to estimate, due to the larger electron density present on such centers.
The occupancies of almost all metal sites are fractional. In the case of Fe crystal 1 and of the Cu crystal, occupancies have been adjusted to obtain ADPs comparable with the surrounding protein atoms. The metal occupancies in the ferroxidase site have been fixed between 0.6 and 0.9 in Fe crystal 1 (average 0.8) and between 0.5 and 0.9 for the Cu crystal (average 0.6). For the second iron crystal, occupancies have been fixed to unity and the ADP has been left to refine freely with values being at maximum 90 % higher than the nearby atoms.
PDB Codes
Final coordinates and structure factors have been deposited in PDB under the accession code 3RBC for Fe crystal 1, 3RGD for Fe crystal 2, 3RE7 for the Cu crystal.
Electronic absorption and circular dichroism spectra
Absorption and circular dichroism spectra (300 – 850 nm) of frog M ferritin solutions were recorded under aerobic condition on a Varian Cary 50 Scan. The demineralized protein was titrated at first with 4 equivalents of copper(II) ions (solution of CuSO4 · 5 H2O). The solution became light green with an electronic absorption spectrum having a very broad band centered at 740 nm. Then, 2 equivalents of iron(II) (FeSO4 · 7 H2O in HCl 0.1%) were added to this solution that immediately became dark gray and, after 15 h, turned into orange, showing slow iron oxidation in the presence of copper.
Visible absorption spectra (300-850 nm) of the iron loaded crystals (Fe crystal 1) and copper loaded crystals were recorded on a GE Healthcare NanoVuePlus Spectrophotometer and showed the same bands observed for the corresponding solutions.
RESULTS AND DISCUSSION
Iron(III) at the ferroxidase site
At variance with the already available structures, where the asymmetric unit was constituted by a single subunit and the 24-mer was generated by symmetry transformations, in all our crystals the asymmetric unit is constituted by the 24-mer and data are independently obtained for the 24 subunits. They are identical within the experimental error (pairwise rmsd values in the 0.21-0.26 Å). The structure of the polypeptide part is superimposable with that of the already available frog M ferritins (rmds values in the range 0.12-0.22 Å).
Crystals soaked for 90 min in iron(II) were rust-orange and showed electronic absorption spectra typical for iron(III) ferritin in solution (see Experimental Procedures, Fe crystal 1). This shows that iron(II) undergoes oxidation under these conditions. Invariably two iron ions in the ferroxidase site were detected with the same coordination environment and intermetal distances. Iron(III) at site 1 (Fe1) is coordinated to three protein residues (monodentate Glu23 on helix H1, Glu58, which bridges the two iron ions, and His61 on helix H2) whereas iron(III) at site 2 (Fe2) binds only two protein residues (bridging Glu58 on helix H2, and Glu103 on helix H3 which acts as a bidentate ligand for Fe2) (Scheme 1, right panel and Fig. 2a). This coordination environment is consistent with previous NMR data, where paramagnetic induced line-broadening settled the position of dimeric iron(III) products away from helix H4, towards helices H1 and H2.3 The B factors are essentially the same for Fe1 and Fe2. The refined occupancies for Fe1 and Fe2 are also the same within the experimental error and of the order of 80%. All coordination bonds, determined with a precision of 0.3 Å, are relatively long (Table S2 and Scheme 1, right panel). This effect may be partially attributable to the partial occupancy of the metal sites. Ligand-metal-ligand bond angles involving protein residues are provided in Table S3. Fe1 and Fe2 structurally correspond to site A and B of P. furiosus (Fig. S1C), respectively, but the Fe-Fe distance observed in the 24 structurally independent subunits shows a much narrow spread with respect to the P. furiosus ferritin, being 3.1 ± 0.1 Å (Table S2). Such a distance suggests the presence of an oxo/hydroxo bridge,32;33 although not visible at the present resolution. Indeed, the expected short bonds for Fe(III)-oxo (≈1.80 Å)32;34 hide the oxygen atom within the electron density of the di-iron cluster. The same inter-metal distance, within experimental error, has been reported for the diferric product of ferritin catalysis from rapid freeze quench EXAFS measurements.9 In the present structure, a further bridging water or hydroxo group is detectable at about 2.6 Å from each iron ion. Other interactions with terminal water molecules at longer distances are also operative (Table S2); a caveat should be given that the position and number of water molecules around iron ions could be severely biased by the resolution limit and Fourier series truncation effects. As a whole, the Fe-Fe moiety with its non-protein ligands is loosely anchored to the protein: Fe2 forms coordination bonds with two amino acids while Fe1 with three. For comparison, in other diiron proteins the number of iron protein monodentate ligands for each iron ion ranges between 3 and 5.35-37 In E.coli bacterioferritin, 4 ligands per iron are found with a total of 5 donor atoms as one Glu acts as bidentate.14 As a consequence of its less restricted coordination, in ferritins the iron(III) dimers may be pushed easily out by incoming iron(II).
Scheme 1.

Proposed structural changes at the ferroxidase site upon iron oxidation. The structural model for the iron(II) form (left panel) is based on the structure of the copper(II) derivative and supported by iron in crystal 2. Observed bonds are represented as broken lines and their average lengths are provided in Å. The proposed oxo-bridge in the diferric center (right panel) is indicated by continuous lines.
Figure 2.

Stereo view of the coordination environment of iron(III) in Fe crystal 1 (a) and copper(II) ions (b) at the ferroxidase site; the models are superimposed to the 2Fo-Fc electron density map contoured at 1.0 σ and 1.5 σ, respectively, (blue wire) and to the anomalous difference electron density map contoured at 3.5 σ (yellow thick wire panel a) and 5.5 σ (copper thick wire panel b). (c) Least-squares superimposition of the Fe bound (yellow sticks) and Cu bound (green sticks) ferritin. (d) Least squares superimposition of all 24 independent subunits of frog M ferritin Fe crystal 2. Iron ions at site 2 occupy tow two distinct positions, one similar to that of copper and one similar to that of iron in crystal 1. The same similarities hold for the side chain of His54. These findings suggest the rearrangement of the di-iron cluster upon oxidation.
Bipositive cations at the ferroxidase site
Unraveling the mechanism by which iron(III) is substituted by iron(II) is another challenge, as the former is expected to bind electron donors with higher affinity than the latter.38 The choice of copper(II) as a model for iron(II) was guided by the consideration that it is a bivalent transition metal ion not prone to oxidation. Furthermore, copper(II) has coordination properties that a posteriori resulted in a distribution inside the protein that is more easily interpretable than that found for other divalent cations.4;24 Consequently the structure of the copper(II) adduct, obtained with the same experimental approach as for the iron form (Fig. 2b) is enlightening. Two copper ions are present at site 1 and site 2 of the ferroxidase site. The Cu-Cu distance in the catalytic site is 4.3 ± 0.4 Å. While the coordination environment of copper at site 1 (Cu1) is similar to that of Fe1, copper at site 2 (Cu2) is bound to a further protein residue, i.e. His54 on helix H2 (Fig. 2b), that changes its conformation with respect to iron crystal 1, going from the distal position adopted in the iron crystal 1 and in the apoprotein (PDB id: 4DAS) to a proximal position. Glu58 side chain reorients but still bridges the two metal ions, whereas no electron density attributable to other bridging atoms is observed. The metal-donor distances, that have a precision of 0.3 Å, are still somewhat larger than usual coordination distances (Table S2). The coordination bond lengths and angles (Table S3) for all the other protein ligands remain the same, within the experimental error, as in iron crystal one. When considering the atoms of the side chains of metal ligands in iron crystal 1 and copper crystal, RMSD values larger than 1 Å are found only for the Oε2 of Glu58 (1.6 Å) and for the Cδ2, Nδ1 and Nε2 of His54 (1.4, 1.8 and 1.7 Å, respectively).
His54 binding to cobalt(II) has been observed in crystals of ferritin obtained by co-crystallization.4 It can be hypothesized that the coordination of iron(II) at the catalytic site is similar to that of copper(II). The presence of a bridging carboxylate and the absence of bridging water/hydroxo are consistent with the weak antiferromagnetic coupling derived from MCD data on the iron(II) species.5
The picture resulting from the copper-bound structure suggests that oxidation requires changes in the coordination environment of iron(II) at site 2 and parallel reduction in the metal-metal distance. This is supported by the structure obtained from the ferritin crystal soaked in FeSO4 solution for a short time (1 min; see Experimental Procedures, Fe crystal 2). As shown in Fig. 2d, the 24 independent ferritin subunits present in the crystal asymmetric unit show different structures of ferric dimers at the ferroxidase site, suggesting the presence of iron trapped in the initial docking site of the ferrous state and in the state corresponding to the final oxidation product. Although no direct evidence of the iron oxidation state in this crystal is available, beside the paler orange color with respect to Fe crystal 1, some structural data support this hypothesis. The shortest iron-iron distances correspond to that already observed in Fe crystal 1; the longest distances are suggestive of iron(II)-iron(II) sites 33;34 and compare very well with the copper(II)-copper(II) distance (Fig. S2). The same holds for Fe2-His54Nε2 distances. The shortest Fe2-His54 distances are found when the Fe1-Fe2 distances are the longest and vice-versa, suggesting the involvement of His54 in iron(II) binding when it enters the ferroxidase site and its detachment from iron binding upon metal oxidation. At variance with crystal 1, the metal ion at site 2 has significantly larger B-factor than that at site 1, supporting the idea that the latter corresponds to a higher affinity preformed metal binding site. At site 2 affinity is modulated by the metal ion oxidation state.
The detachment of His54 in passing from iron(II) to iron(III) is due to the establishment of a short Fe(III)-Fe(III) distance characteristic of a μ-oxo-bound moiety. When iron oxidation occurs via dioxygen, the bridging oxygen atom forces iron at site 2 to move towards Fe1, breaking the Fe2-His54 bond; as a consequence the His54 side chain reorients (Fig. 2c) going from the proximal to the distal conformation. The new position of iron is outside the binding range for His54 (Fig. 2c). His54 may assist an incoming iron(II) to replace iron(III). The structure of iron(III) (Scheme 1, right panel) therefore describes the situation of the transient iron(III) product at the catalytic site, while that of copper(II) may represent the pre-reactive conformation of iron(II) (Scheme 1, left panel). Copper(II), which does not react with oxygen, remains trapped in the pre-reactive conformation. Furthermore, the presence of copper(II) significantly slows down the iron oxidation reaction in competition binding experiments. This was demonstrated by adding 2 equivalents of iron(II) at ferritin containing 4 equivalents of copper(II) ions. Fifteen hours were needed before the solution turned into orange, i.e. before iron is oxidized (see Experimental Procedures).
Other metal binding sites
The complete picture of ferritin provided by crystallography (Fe crystal 2 data) includes iron occupying the so-called C3 pore (Fig. 3a). Distinct anomalous signals corresponding to iron ions are clearly detected, above noise, in the channel and attributed to monomeric iron species caught at different steps along the passage, at varying distances and occupancies from Glu130 side chains. The crystallographic data cannot reveal the iron oxidation state at the C3 pore, however, it is possible that the signals at the ternary axes arise from not yet reacted iron(II), which is attracted by internal negative charges in the channel, consistently with previous reports on other bivalent metal ions.4 This observation supports the proposal that the C3 channel may represent the entry pore.
Figure 3.

Stereo views of the structures for Fe crystal 2 (a) and copper (b) at the C3 pore. For iron, distinct anomalous signals (gold thick wire contoured at 3.5 σ) corresponding to different positions of iron ions are seen in the channel. The iron positions are represented as spheres whose diameter is proportional to the observed anomalous signal intensity. In the case of copper a completely new setting is observed with three pairs of copper(II) ions bound to the three symmetry related subunits(copper thick wire contoured at 5.5 σ).
A weak anomalous signal and a blob of electron density indicate the presence of iron ions also in the pores corresponding to the four-fold axes (C4 pores), at a distance (3.0-3.2 Å) from the Nε2 of His169 that suggest weak or no coordination (Fig. 4a). This single metal ion observed at the end of the C4 pore may diffuse there from the bulk solution rather than through the ferroxidase site and therefore may represent a non physiological situation. Indeed, the path from the catalytic site to the cavity was traced by NMR3 that, together with other spectroscopies,9;39-41 has shown that product migration proceeds via iron clusters.
Figure 4.

Top-down view (stereo mode) of the C4 pore: (a) an anomalous signal between is observed in all 4-fold sites and can be interpreted as the presence of a Fe ion (gold thick wire). (b) In the case of copper (copper thick wire) the four His169 side chains move closer to the metal ion reaching proper coordination distances (2.2-2.5 Å); a water molecule (not shown) binds copper from the inner part of the channel, completing a square pyramidal coordination.
Copper(II) at the C3 and C4 pores (Fig. 3b and 4b) has arrangements different from iron and, presumably, not functionally relevant. For instance, the ternary axes hosts beautiful exanuclear copper clusters bound to His114, Glu130 and by the Cys126 sulfur (of each subunit) that acts as bridging ligand to two adjacent copper ions. The Cys126 side chain rotates about 180° with respect to the unbound structure in order to bind copper (Fig. 3b). At the quaternary axis the four His169, at variance with Fe crystal 2, extend about 1.0 Å into the pore to bind the copper ion (Fig. 4b).
CONCLUSIONS
The structure determinations of M ferritin crystals, obtained under aerobic conditions for two different soaking times in iron(II) solutions, reveal the binding of iron to a higher eukaryotic ferritin. The crystal form used in this study contains 24 independent subunits in the asymmetric unit, providing highly redundant comparisons between subunits. This situation, for long soaking times (Fe crystal 1), results in the observation of invariant diferric sites where the Fe-Fe average distance of 3.1 ± 0.1 Å is, within experimental error, the same found by XAS spectroscopy9 and strongly indicates the presence of a μ-oxo/hydroxo bridge between the irons. Two copper(II) ions also bind at the ferroxidase site, giving rise to a dinuclear center with longer intermetal distances and with His54 as an extra ligand for the metal at site 2. This arrangement of the ligands is comparable, although not identical, to that reported for cobalt(II), where three ions were detected at the ferroxidase site:4 while at site 1 a single cobalt ion is detected, at site 2 two alternative positions for cobalt ions are observed, which correspond to His54 in two conformations (which we have called proximal and distal) (Fig. S2), but always acting as a metal ligand.
The redundant information for the 24 independent subunits allows the detection of two different types of di-iron centers in the crystal soaked for the shorter time (1 min, Fe crystal 2). The metal positions are, within the experimental error, the same as that of iron(III) in the Fe crystal 1 and of copper(II)-containing crystals, as depicted in Scheme 1. The findings on Fe crystal 2, therefore, complete and support the mechanistic proposal of Scheme 1 and may provide a snapshot of the structural changes occurring upon iron oxidation.
The ferroxidase site of M frog ferritin, and site 2 in particular, appears to be quite flexible, depending on the oxidation state of the hosted metal ion. The frog M ferritin His54 is substituted by a Gln in P. furiosus (which shares an overall sequence identity of 32% with M frog ferritin) and by a Glu in Pseudo-nitzschia multiseries (overall identity of 27%). These differences in sequence translate into different metal binding sites, as observed in the respective crystal structures (Fig. S1). A comparison of frog M ferritin with the sequences of other hundred ferritins from eukaryotes having ≥66% sequence identity with it, shows that the position equivalent to His54 is generally occupied by Asn or Gln, which may play the same game to provide flexible ligands to incoming divalent cations and guide the proposed shuttling mechanism.
The path from the catalytic site to the biomineralization cavity is traced only by NMR3 because transient coordination sites along the path are poorly populated, as expected on functional grounds. The observation of iron at different positions within the 3-fold pores, suggests a possible pathway for iron entrance in the protein.
The comparison between the iron(III) and metal(II) derivatives suggests a detailed mechanism for the oxidation reaction in eukaryotic ferritins. Iron(II) is presumably attracted by electrostatic forces to the active site, where its binding (in a pre-reactive conformation) to site 2 is favored by His54. Oxidation and formation of diferric oxo/hydroxo bridges stabilizes the coordination cluster, imposing a shorter inter-metal distance, and destabilizes the binding to the protein. New incoming iron(II) may push the di-iron(III) moiety towards the cavity.
Supplementary Material
Acknowledgments
We thank Marta Argenti for preliminary spectroscopic characterization of copper ferritin and Manuela Benvenuti for help in protein crystallization experiments. S.M and C.P. acknowledge the financial support from CIRMMP. We acknowledge ESRF (Grenoble) for having provided access to the ID14-4 and ID29 beamlines. The research was funded by Italian MIUR (contract FIRB PROTEOMICA MIUR -RBRN07BMCT) and by the NIH (Grant DK20251).
ABBREVIATIONS
- NMR
Nuclear Magnetic Resonance
- EXAFS
Extended X-ray Absorption Fine Structure
- MCD
Magnetic Circular Dichroism
- ESRF
European Synchrotron Radiation Facility
- CCD
Charge-Coupled Device
- ADSC
Area Detector Systems Corporation
- PDB
Protein Data Bank
- FFT
Fast Fourier Transform
- CD
Circular Dichroism
- XAS
X-ray Absorption Spectroscopy
Footnotes
Supporting Information. Tables reporting data collection and refinement statistics.(Table S1) and main structural features of the active site (Table S2). A figure comparing metal binding sites within the 4-helix bundle channel of ferritin from different organisms (Fig. S1). A figure (Fig. S2) comparing the coordination environment of various metal ions bound to M frog ferritin. This material is available free of charge via the Internet at http://pubs.acs.org.
Author Contributions
I.B., S.M. and P.T. designed research and wrote the paper; D.L., C.P. and C.R. prepared protein samples and crystals; C.P. and S.M. acquired data and refined the structure; E.C.T.inspired research and contributed to the discussion of the results.
All authors have given approval to the final version of the manuscript.
References
- 1.Liu X, Theil EC. Acc Chem Res. 2005;38:167–175. doi: 10.1021/ar0302336. [DOI] [PubMed] [Google Scholar]
- 2.Theil EC. Current Opinion in Chemical Biology. 2011;15:1–8. doi: 10.1016/j.cbpa.2011.01.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Turano P, Lalli D, Felli IC, Theil EC, Bertini I. Proc Natl Acad Sci USA. 2010;107:545–550. doi: 10.1073/pnas.0908082106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Tosha T, Ng H-L, Bhattasali O, Alber T, Theil EC. J Am Chem Soc. 2010;132:14562–14569. doi: 10.1021/ja105583d. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Schwartz JK, Liu XS, Tosha T, Theil EC, Solomon EI. J Am Chem Soc. 2008;130:9441–9450. doi: 10.1021/ja801251q. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bou-Abdallah F, Zhao G, Mayne HR, Arosio P, Chasteen ND. J Am Chem Soc. 2005;127:3885–3893. doi: 10.1021/ja044355k. [DOI] [PubMed] [Google Scholar]
- 7.Zhao G, Bou-Abdallah F, Arosio P, Levi S, Janus Chandler C, Chasteen ND. Biochemistry. 2003;42:3142–3150. doi: 10.1021/bi027357v. [DOI] [PubMed] [Google Scholar]
- 8.Bou-Abdallah F, Arosio P, Santambrogio P, Yang X, Janus Chandler C, Chasteen ND. Biochemistry. 2002;41:11184–11191. doi: 10.1021/bi020215g. [DOI] [PubMed] [Google Scholar]
- 9.Hwang J, Krebs C, Huynh BH, Edmondson DE, Theil EC, Penner-Hahn JE. Science. 2000;287:122–125. doi: 10.1126/science.287.5450.122. [DOI] [PubMed] [Google Scholar]
- 10.Tatur J, Hagedoorn PL, Overeijnder ML, Hagen WR. Extremophiles. 2006;10:139–148. doi: 10.1007/s00792-005-0484-x. [DOI] [PubMed] [Google Scholar]
- 11.Tatur J, Hagen WR, Matias PM. J Biol Inorg Chem. 2007;12:615–630. doi: 10.1007/s00775-007-0212-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Marchetti A, Parker MS, Moccia LP, Lin EO, Arrieta AL, Ribalet F, Murphy ME, Maldonado MT, Armbrust EV. Nature. 2009 Jan 22;457(7228):467–70. doi: 10.1038/nature07539. Epub.2008.Nov.26. 2009, 457, 467-470. [DOI] [PubMed] [Google Scholar]
- 13.Carrondo MA. EMBO J. 2003;22:1959–1968. doi: 10.1093/emboj/cdg215. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Crow A, Lawson TL, Lewin A, Moore GR, Le Brun NE. J Am Chem Soc. 2009;131:6808–6813. doi: 10.1021/ja8093444. [DOI] [PubMed] [Google Scholar]
- 15.Weeratunga SK, Lovell S, Yao H, Battaille KP, Fischer CJ, Casey EG, Rivera M. Biochemistry. 2010;49:1160–1175. doi: 10.1021/bi9015204. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Nordlund P, Sjöberg B-M, Ecklund H. Nature. 1990;345:593–598. doi: 10.1038/345593a0. [DOI] [PubMed] [Google Scholar]
- 17.Rosenzweig AC, Brandstetter H, Whittington DA, Nordlund P, Lippard SJ, Frederick CA. Proteins. 1997;29:141–152. [PubMed] [Google Scholar]
- 18.Cho KJ, Shin HJ, Lee JH, Kim KJ, Park SS, Lee Y, Lee C, Park SS, Kim KH. J Mol Biol. 2009;390:83–98. doi: 10.1016/j.jmb.2009.04.078. [DOI] [PubMed] [Google Scholar]
- 19.Hamburger AE, West AP, Hamburger ZA, Hamburger P, Bjorkman PJ. J Mol Biol. 2005;10:558–569. doi: 10.1016/j.jmb.2005.03.074. [DOI] [PubMed] [Google Scholar]
- 20.Leslie AGW. Acta Crystallogr D Biol Crystallogr. 2006;62:48–57. doi: 10.1107/S0907444905039107. [DOI] [PubMed] [Google Scholar]
- 21.Evans PR. Joint CCP4 and ESF-EACBM Newsletter. 1997;33:22–24. [Google Scholar]
- 22.Collaborative Computational Project. N Acta Cryst. 1994;D50:760–763. [Google Scholar]
- 23.Vagin AA, Teplyakov A. J Appl Crystallogr. 1997;30:1022–1025. [Google Scholar]
- 24.Ha Y, Shi D, Small GW, Theil EC, Allewell NM. J Biol Inorg Chem. 1999;4:243–256. doi: 10.1007/s007750050310. [DOI] [PubMed] [Google Scholar]
- 25.Murshudov GN, Vagin AA, Dodson EJ. Acta Crystallogr D Biol Crystallogr. 1997;53:240–255. doi: 10.1107/S0907444996012255. [DOI] [PubMed] [Google Scholar]
- 26.Emsley P, Cowtan K. Acta Crystallogr D Biol Crystallogr. 2004;60:2126–2132. doi: 10.1107/S0907444904019158. [DOI] [PubMed] [Google Scholar]
- 27.Perrakis A, Morris RJH, Lamzin VS. Nat Struct Biol. 1999;6:458–463. doi: 10.1038/8263. [DOI] [PubMed] [Google Scholar]
- 28.Laskowski RA, MacArthur MW, Moss DS, Thornton J. J Appl Crystallogr. 1993;26:283–291. [Google Scholar]
- 29.Adams PD, Afonine PV, Bunkòczi G, Chen VB, Davis IW, Echols N, Headd JJ, Hung L-W, Kapral GJ, Grosse-Kunstleve RW, McCoy AJ, Moriarty NW, Oeffner R, Read RJ, Richardson DC, Richardson JS, Terwilliger TC, Zwart PH. Acta Crystallogr D Biol Crystallogr. 2010;66:213–221. doi: 10.1107/S0907444909052925. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Potterton L, McNicholas S, Krissinel E, Gruber J, Cowtan K, Emsley P, Murshudov GN, Cohen S, Perrakis A, Noble M. Acta Crystallogr D Biol Crystallogr. 2004;60:2288–2294. doi: 10.1107/S0907444904023716. [DOI] [PubMed] [Google Scholar]
- 31.Murshudov GN, Vagin AA, Dodson EJ. Acta Crystallogr D Biol Crystallogr. 1997;1:240–255. doi: 10.1107/S0907444996012255. [DOI] [PubMed] [Google Scholar]
- 32.Vincent JB, Olivier-Lilley GL, Averill BA. Chem Rev. 1990;90:1447–1467. [Google Scholar]
- 33.Tshuva EY, Lippard S. J Chem Rev. 2004;104:987–1012. doi: 10.1021/cr020622y. [DOI] [PubMed] [Google Scholar]
- 34.Kurtz DM., Jr Chem Rev. 1990;90:585–606. [Google Scholar]
- 35.Strand KR, Karlsen S, Kolberg M, Rohr AK, Gorbitz CH, Andersson KK. J Biol Chem. 2004;279:46794–46801. doi: 10.1074/jbc.M407346200. [DOI] [PubMed] [Google Scholar]
- 36.Rosenzweig AC, Frederick CA, Lippard SJ, Nordlund P. Nature. 1993;366:537–543. doi: 10.1038/366537a0. [DOI] [PubMed] [Google Scholar]
- 37.Holmes MA, Trong IL, Turley S, Sieker LC, Stenkamp RE. J Mol Biol. 1991;218:583–593. doi: 10.1016/0022-2836(91)90703-9. [DOI] [PubMed] [Google Scholar]
- 38.Do LH, Lippard SJ. J Am Chem Soc. 2011;133:10568–10581. doi: 10.1021/ja2021312. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Bou-Abdallah F, Papaefthymiou GC, Scheswohl DM, Stanga SD, Arosio P, Chasteen ND. Biochem J. 2002;364:57–63. doi: 10.1042/bj3640057. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Treffry A, Bauminger ER, Hechel D, Hodson NW, Nowik I, Yewdall SJ, Harrison PM. Biochem J. 1993;296:721–728. doi: 10.1042/bj2960721. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Jameson GN, Jin W, Krebs C, Perreira AS, Tavares P, Liu X, Theil EC, Huynh BH. Biochemistry. 2002;41:13435–13443. doi: 10.1021/bi026478s. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
