The ferritin Fe2 site at the diiron catalytic center controls the reaction with O₂ in the rapid mineralization pathway

Abstract

Oxidoreduction in ferritin protein nanocages occurs at sites that bind two Fe(II) substrate ions and O₂, releasing Fe(III)₂–O products, the biomineral precursors. Diferric peroxo intermediates form in ferritins and in the related diiron cofactor oxygenases. Cofactor iron is retained at diiron sites throughout catalysis, contrasting with ferritin. Four of the 6 active site residues are the same in ferritins and diiron oxygenases; ferritin-specific Gln¹³⁷ and variable Asp/Ser/Ala¹⁴⁰ substitute for Glu and His, respectively, in diiron cofactor active sites. To understand the selective functions of diiron substrate and diiron cofactor active site residues, we compared oxidoreductase activity in ferritin with diiron cofactor residues, Gln¹³⁷ → Glu and Asp¹⁴⁰ → His, to ferritin with natural diiron substrate site variations, Asp¹⁴⁰, Ser¹⁴⁰, or Ala¹⁴⁰. In Gln¹³⁷ → Glu ferritin, diferric peroxo intermediates were undetectable; an altered Fe(III)–O product formed, ΔA₃₅₀ = 50% of wild type. In Asp¹⁴⁰ → His ferritin, diferric peroxo intermediates were also undetectable, and Fe(II) oxidation rates decreased 40-fold. Ferritin with Asp¹⁴⁰, Ser¹⁴⁰, or Ala¹⁴⁰ formed diferric peroxo intermediates with variable kinetic stabilities and rates: t_1/2 varied 1- to 10-fold; k_cat varied approximately 2- to 3-fold. Thus, relatively small differences in diiron protein catalytic sites determine whether, and for how long, diferric peroxo intermediates form, and whether the Fe–active site bonds persist throughout the reaction cycle (diiron cofactors) or break to release Fe(III)₂–O products (diiron substrates). The results and the coding similarities for cofactor and substrate site residues—e.g., Glu/Gln and His/Asp pairs share 2 of 3 nucleotides—illustrate the potential simplicity of evolving active sites for diiron cofactors or diiron substrates.

Ferritins are protein nanocages with multiple sites that catalyze the oxidoreduction of Fe(II) and O₂ to produce diferric mineral precursors; the mineral precursors migrate into a central cavity 8 nm in diameter and form Fe(III) hydrated oxo mineral. Iron minerals in ferritins serve to concentrate iron to the high levels needed for the synthesis of iron-containing cofactors. The Fe(II) oxidation reactions in ferritins consume dioxygen or hydrogen peroxide, thereby minimizing dangerous Fenton chemistry (1). Ferritin protein nanocages self-assemble from 24 or 12, 4-α-helix-bundle, subunits in maxiferritin or miniferritin, respectively; miniferritins were named, historically, Dps (DNA protection during starvation) proteins and, to date, are known only in bacteria and archaea; heme-containing ferritins in bacteria are called bacterioferritins (Bfr) (2). The biological importance of ferritin is reflected by wide distribution in most cells of humans, other animals, bacteria, and archaea, by the lethality of gene deletion in mice (3), by mutation effects on the central nervous system (4), and by protection during oxidant stress (5). In addition, the unusual physical stability of ferritin nanocages to temperature and chaotropes has application to materials and protein engineering (6, 7).

Oxidoreductase/ferroxidase (F_ox) catalytic centers with diferrous binding sites, Fe1 and Fe2, occur in each ferritin subunit except in animals, where a second gene encodes a catalytically inactive subunit (ferritin L) that co-assembles with the active subunits in various ratios that depend on the tissue. The ferritin oxidoreductase sites share properties with diiron cofactor catalysts that include the first detectable reaction intermediate in ferritin, a diferric peroxo (DFP) complex, characterized by UV/visible (UV/vis), resonance Raman, Mössbauer, and extended X-ray absorption fine structure (EXAFS) spectroscopies (8–14). The reaction in ferritins is shown in Scheme 1.

DFP decays to Fe(III) oxo or hydroxo dimers or multimers (15), representing reaction products from multiple sites, that move across the protein cage to the cavity where the Fe(III) oxide mineral forms. Parallel formation of diferric dimers and multimers (15), the effects on reaction rates of covalent crosslinks that decrease conformational flexibility (16), and the properties of Fe(II) binding measured directly by circular dichroism/magnetic circular dichroism (CD/MCD) (17) all suggest interactions among the polypeptide subunits during the reactions at the multiple catalytic centers in ferritin protein cages. The active site of a single ferritin polypeptide subunit, based on protein co-crystals with the Fe(II) analogue Mg(II) (18) and recent CD/MCD analysis (17) is shown in Fig. 1. Similar results for identification of the metal-binding amino acid residues, or ligands, were obtained with a Ca(II) co-crystal of ferritin (19) and with a solution study of an L ferritin [no Fe(II)/O₂ oxidoreductase catalytic center] where the putative catalytic center ligands replaced the wild-type (WT) residues (20). A distinguishing feature of the ferritin diiron substrate catalytic centers is the weaker Fe(II) ligands in the Fe2 site that contrast with diiron cofactor catalytic centers, in proteins such as methane monooxygenase (MMO), ribonucleotide reductase (RNR) and Δ⁹-stearoyl-acyl carrier protein desaturase (Δ⁹ desaturase) (21–23) and the heme-containing Bfr (2). Ferritin diiron substrate centers also lack one of the two carboxylate ligands of the diiron cofactor sites, which, like iron in the ferritin substrate sites, react with O₂ to form a DFP intermediate (8–10).

Fig. 1.

Comparisons of diiron sites for Fe substrate (ferritin) and cofactors reveal significant differences in iron ligands and second-shell residues. The ferritin Fe2 site ligands are Glu¹⁰³, Gln¹³⁷, and Asp¹⁴⁰, with a bridging carboxylate, Glu⁵⁸. The distinctive ferritin Gln¹³⁷ is compared with Glu²⁴³ in methane monooxygenase (MMO) and Glu²³⁸ in ribonucleotide reductase (RNR). Multiple histidine bonding interactions in diiron cofactor centers are absent from ferritin diiron substrate centers. The figure uses the frog M ferritin model [Protein Data Bank (PDB) file 1MFR, a crystal structure with Mg(II) as an Fe(II) homologue (18)] and CD/MCD analysis of diferrous binding (17). Diiron cofactor site structures include diferrous, as in MMO (1FYZ) (21) and RNR (1PFR) (22).

Ferritins are isoforms that vary not only in overall polypeptide sequences but also in the sequence at the catalytic centers. The selective expression of the different ferritin isoforms in specific tissues or organelles indicates that the variations are physiologically relevant (24, 25). Four of the 6 iron ligands at the F_ox catalytic centers of ferritins are present in diiron enzyme cofactor centers (Fig. 1), except in Bfr with all six diiron cofactor site ligands. The two divergent residues at the ferritin catalytic centers are both in the Fe2 site: a conserved glutamine at position 137* and a variable residue at position 140; in Fig. 1, residue 140 is aspartate from the extensively characterized frog M ferritin model.

We now report the results of investigating the function of the ferritin-specific residues in the Fe2 site of ferritin catalytic centers by comparing the catalytic properties of natural variants in residue 140, aspartate, serine, or alanine, in a single polypeptide sequence (frog M) to that in the natural polypeptide (frog M, Asp¹⁴⁰; frog H, Ser¹⁴⁰; and human H, Ala¹⁴⁰). We also examined the diiron cofactor residue, histidine, in the ferritin context. We observed that k_cat, and K_app for DFP formation, and DFP decay (t_1/2), determined from the progress curves of A₆₅₀, were sensitive both to residue 140 and to the polypeptide backbone, suggesting coevolution of second-shell or distant residues, while histidine inhibited Fe(II) oxidation in the ferritin polypeptide background. Changing ferritin-specific invariant glutamine to glutamate inhibited DFP formation (A₆₅₀) at either 20 °C or 4 °C but not Fe(II) oxidation, measured by A₃₅₀, although the Fe(III)O product(s) had altered spectral properties. The absorbance at 350 nm is not a specific transition but is in the middle of the broad absorbance at 300–400 nm, due to Fe(III)O species that include DFP, diferric Fe(III) oxo/hydroxo mineral precursors, and Fe(III) oxide mineral with large absorbance of protein aromatic amino acids between 280 and 300 nm. The results illustrate the importance of ferritin Fe2 site-specific residue DFP formation and in directing DFP decay away from higher-valence iron species, such as Fe(IV), as in diiron cofactor proteins, and toward the diferric mineral precursors.

Results

Natural Asp¹⁴⁰ Variants in Ferritin Fe2 Site Alter Catalytic Rates.

The variations at residue 140 in the Fe2 site of the catalytic centers (1, 20, 25–27) are inconsistent with the high conservation of the other residues as well as the conservation of histidine at the analogous position in the catalytic centers of related diiron oxygenases. The variations in residue 140 at the ferritin Fe2 sites appear to be physiologically relevant based on patterns of expression (24, 25); the physiological parameter driving the selection is unknown, but it could be temperature, oxygen concentration, or supramolecular interactions with other cellular components that are cell or organelle specific. To examine the effects of residue 140 variants, we compared the activity of proteins with ferritin Fe2 site Asp¹⁴⁰Ala (Fe site2 with Glu¹⁰³, Gln¹³⁷, and Ala¹⁴⁰; EQA) and Asp¹⁴⁰Ser (EQS) in the same ferritin backbone, the frog M-ferritin subunit, to the WT, Asp¹⁴⁰ (EQD).

Rapid Fe(II) oxidation with DFP formation was observed with all natural residue 140 variants, measured either as the increase in A₆₅₀, at the absorption maximum of ferritin DFP (8, 9, 15, 28) (Fig. 2A), or as the increase in A₃₅₀, the less specific Fe(III)O species described above (Table 1). Fitting the data for initial rates at different concentrations of iron showed positive cooperativity (Fig. 2B), as previously observed with WT protein (20). Note that the Hill coefficient for DFP formation is smaller than for the binding of Fe(II) in the absence of O₂ (n ∼ 3) (17), indicating that the basis for cooperativity is different for binding of the first substrate, Fe(II), to the protein in the absence of the second substrate, O₂, compared with the reaction between Fe(II) and the second substrate, O₂.

Fig. 2.

Effects of different Fe2 site residues with natural ferritin variants or with diiron oxygenase ligands, on rates in ferritin. (A) Progress curves of A₆₅₀. Apoferritin (0.1 mM subunit) (4.2 μM nanocages) in 200 mM Mops, 200 mM NaCl at pH 7.0 was mixed with an equal volume of 0.2 mM ferrous sulfate (48 Fe per nanocage) in 1 mM HCl, at room temperature (20 °C) in a stopped-flow spectrophotometer (mixing time <5 ms). Traces are from the following: 1, WT frog M (EQD); 2, EQA; 3, EQS; 4, EQH; 5, EEH; and 6, WT human H (EQA). (Inset) Time course of A₃₅₀ for the following: 7, WT frog M (EQD); 8, EQH; and 9, EEH. (B) The initial rate (V_i) for DFP formation as a function of iron concentration. The conditions were the same as for A. The curves represent the results of the Hill equation fitting. ○, WT frog M (EQD); □, EQA; ◇, EQS; and ●, WT human. (C and D) Effect of Gln¹³⁷ → Glu substitution. The mixing conditions were the same as for A and B. (C) Progress curve for WT frog M (EQD). (D) Progress curve for EED. Traces shown here are 1, A₆₅₀; and 2, A₃₅₀. (Inset) Kinetic traces obtained at 4 °C for WT frog M (EQD) (black) and EED (gray). The trace for EED was multiplied by 10 for comparison.

Table 1.

Effects of altering Fe2 sites in ferritin Fe(II)/O₂ oxidoreductase catalysis

Protein	Single turnover	Multiple turnover (48–480 Fe per protein cage)
	DFP (A650)	DFP (A650)			Fe3+O formation (A350)
Fe2 site and backbone*	t1/2 (ln2/kdecay),†s	kcat,§ s−1	Kapp, mM	n	Max Vi,‖ Δ A350·s−1	Kapp, mM	n
EQD-Fr M	0.37 ± 0.02	89 ± 2	0.20 ± 0.02	1.4 ± 0.1	7.4 ± 0.2	0.13 ± 0.01	1.6 ± 0.1
EQS-Fr M	0.15 ± 0.01	30 ± 2	0.32 ± 0.12	1.4 ± 0.2	5.8 ± 0.2	0.55 ± 0.08	1.4 ± 0.1
EQA-Fr M	0.21 ± 0.02	77 ± 7	0.64 ± 0.17	1.3 ± 0.2	7.5 ± 0.2	0.62 ± 0.05	1.3 ± 0.1
EQS-Fr H¶	—	25 ± 6	0.20 ± 0.03	1.6 ± 0.1	—	—	—
EQA-Hu H	1.5 ± 0.2	49 ± 7	0.20 ± 0.02	1.3 ± 0.1	4.1 ± 0.1	0.18 ± 0.03	1.5 ± 0.1
EQH-Fr M	—	—	—	—	0.33 ± 0.10	1.4 ± 1.3	0.72 ± 0.29
EEH-Fr M	—	—	—	—	0.20 ± 0.04	0.81 ± 0.56	0.71 ± 0.19
EED-Fr M	—	—	—	—	8.6 ± 0.2	0.37 ± 0.06	1.2 ± 0.1

*Single-letter amino acid code for Fe2 sites in ferritins with polypeptide backbone as: human (Hu) or frog (Fr). Boldface letters indicate amino acids present in the Fe2 site of ferritins and also in diiron cofactor active sites.

^†t_1/2 values were obtained by single-exponential fitting to DFP decay process of A₆₅₀; 48 Fe/protein cage (2Fe/active site). The results compare, within experimental error, to double exponential fitting (data not shown) for all the proteins except EQS-Fr M and EQA-FrM where t_1/2, using double exponential fitting, is 0.069 ± 0.04 and 0.14 ± 0.04, respectively but the fitting quality, judged by nonrandom residuals, is poor. Regardless of the model used to fit the data, t_1/2 was significantly different (P<0.01) in the mutants and wild type Hu H compared to Fr M.

^§The value for k_cat is reported/site (24 sites/protein nanocage) for comparison to single site diiron oxygenases, but contrasting with values per ferritin nanocage reported previously (20), which are 24-fold higher. The values presented were computed assuming a constant 1,000 M⁻¹·cm⁻¹ (9) as before (20). Using double exponential fitting as in (9) to compute ε, the values were within experimental of the reported value, ε = 1040 ± 80, 960 ± 140 and 1060 ± 80 M⁻¹·cm⁻¹ for frog M WT, D140A and human H WT, respectively, except for EQS where ε = 720 ± 120; non-random residuals indicated that the goodness of fits were poorer for the double exponential fitting model. Nevertheless, k_cat values results with each fitting model were within 10% of each other except for EQS where the vaues differed by 30%. All the computed values of k_cat were significantly different (P<0.01) in the mutants, EQS FrM, EQA, Fr M, and in EQS-Fr H and EQA-Hu H compared to EQD-Fr M.

^‖Maximum V_i (ΔA_{350 nm}·s⁻¹) is used instead of V_max because the spectra of the Fe(III)O species (DFP, diferric product, mineral) have not, to date, been spectroscopically resolved.

^¶Data from ref. 20.

The initial rates and the kinetic parameters such as k_cat, K_app, and decay of DFP (t_1/2) were dependent on the Fe2 residue 140 (Fig. 2B and Table 1). Asp¹⁴⁰ was the most efficient among natural variants in residue 140 with frog M backbone, which is the WT combination. DFP kinetic stability, t_1/2 values, varied only slightly among the proteins with variable residues at position 140 (Fig. 2A and Table 1), Asp¹⁴⁰, 0.34 ± 0.02 s; Ala¹⁴⁰, 0.21 ± 0.02 s; and Ser¹⁴⁰, 0.15 ± 0.01 s. The k_cat value in the Ala¹⁴⁰ ferritin was comparable to WT (Asp¹⁴⁰), but k_cat for Ser¹⁴⁰ was decreased to less than 50% of WT. Among the three types of catalytic centers, the order of K_app, Asp¹⁴⁰ < Ser¹⁴⁰ < Ala¹⁴⁰, suggests that Fe(II)/O₂ binding was stabilized by a ligand and/or longer side chain. Slower DFP formation, higher K_app [lower affinity for Fe(II)], smaller k_cat, and faster DFP decay result in lower accumulations of DFP and can explain the lower peak absorbance in Ala¹⁴⁰ or Ser¹⁴⁰ than Asp¹⁴⁰ (Fig. 2A).

When the same measurements were made in proteins where residue 140 was matched to the homologous polypeptide backbone, the results contrasted with those in the frog M polypeptide, illustrating the contribution of the polypeptide backbone. For example, K_app was constant among WT ferritins, independent of residue 140 for frog M (Asp¹⁴⁰), frog H (Ser¹⁴⁰), and human H (Ala¹⁴⁰), indicating that binding stability with weak ligands at the active centers could be compensated for by changes elsewhere in the protein structure (Table 1). In the Ala¹⁴⁰ ferritins, in the human H and frog M polypeptide backbones, the kinetic differences were particularly large (Fig. 2A and Table 1). For example, Ala¹⁴⁰ in the frog M polypeptide context had a 7-fold lower DFP kinetic stability, and a 3-fold larger K_app than in Ala¹⁴⁰ in the homologous human H polypeptide. In addition, Ser¹⁴⁰ in the frog M ferritin polypeptide, which predominates in liver, had a significantly higher K_app than in the homologous frog H ferritin polypeptide, which predominates in red blood cells (Table 1). These observations, which may reflect ferritin physiology at the different temperatures of warm-blooded or cold-blooded animals, or subcellular supramolecular interactions, emphasize the role of residues beyond the active site in ferritin catalysis.

His¹⁴⁰, Required for Diiron Cofactor Oxidation, Eliminates Fe(II) Oxidation in Ferritin.

To test the range of variability in residue 140 compatible with ferritin catalysis, given that all of the combinations of natural ferritin residues with various polypeptide backbones were kinetically competent (Table 1), we studied the effect of histidine at position 140 in the frog M polypeptide backbone. [Histidine is the WT residue at the equivalent position of the diiron oxygenase cofactor sites that form the same DFP intermediate from Fe(II) and O₂ as ferritins.] All of the polypeptides with histidine substitutions at the catalytic centers assembled normally into the protein nanocages, based on gel filtration and iron mineralization properties. In the ferritin protein environment, histidine at position 140 prevented DFP formation (2 Fe per active site or 48 Fe per nanocage under standard, aerobic conditions) (Fig. 2A trace 4]. No DFP was detected, even with high concentrations of Fe(II) (up to 1,200 Fe per nanocage) (data not shown). Thus, to determine if Fe(II) was oxidized by the His¹⁴⁰ protein, we measured the ΔA₃₅₀. All Fe(III)O species, including DFP, Fe(III) oxo or hydroxo dimers, and Fe(III) oxide mineral have spectral changes in the region 300–400 nm. Fe(II) was oxidized at a very slow rate in the His¹⁴⁰ protein, retarded by 40-fold compared with the frog M protein (Fig. 2A Inset). Moreover, the kinetic parameters such as maximum V_i and K_app for Fe(II) oxidation, determined from the initial rate of ΔA₃₅₀ at various concentrations of Fe(II), were greatly affected by His¹⁴⁰ substitution. For example, the maximum V_i was decreased by 95%, and K_app increased 10-fold (Table 1). A double mutant that replaced both ferritin-specific residues in the Fe2 substrate site with amino acids that completed the iron binding set of the cofactor sites in diiron oxygenases was as inactive as a single histidine substitution (Fig. 2A and Table 1) in the ferritin polypeptide context. Such data, although unexpected, show that even though histidine functions in very similar reactions between Fe(II) and O₂ in the environment of diiron oxygenase centers, the variations in residue 140 compatible with reactivity in ferritins are restricted to aspartate, serine, alanine, or possibly another amino acid, but not histidine.

Gln¹³⁷, Conserved in the Ferritin Fe2 Site, Controls the Diferric Peroxo Pathway.

Gln¹³⁷ in the Fe2 site of all ferritins (Fig. 1) is a conservative substitution of an amide side chain for the absolutely conserved carboxylate side chain in glutamate at the equivalent position in diiron cofactor sites. A glutamine/glutamate substitution might produce an active protein with an activity distinct from either ferritin or a diiron oxygenase. Therefore, we created the Gln¹³⁷ → Glu mutant of frog M that generates a ferritin Fe2 site with EED.

There were two effects of the Gln¹³⁷ → Glu substitution: (i) No DFP detected; (ii) Fe(III)O product(s) with different spectral properties. We know of no other observation of a ferritin protein with such properties. The absence of DFP (Fig. 2D trace 1] could be caused by decreased kinetic stability of the intermediate or a change in the pathway of the reaction of Fe(II) with O₂ (Fig. 2C trace 2 and Fig. 2D trace 2). For Fe(III)O product, the maximum initial rate, V_i, K_app, and the Hill coefficient (n), determined from ΔA₃₅₀, were comparable to the WT Fe2, EQD site (Table 1). When the reaction was measured at 4 °C, instead of the standard 20 °C, in an attempt to trap any reaction intermediates, we observed the appearance of a species (Fig. 2D Inset) with increasing absorbance from 350–700 nm, without the 650-nm absorbance maximum characteristic of the DFP complex of ferritin (data not shown). Even at 4 °C, the EED-specific, Fe(II) + O₂ intermediate was only 1/15 as kinetically stable (t_1/2: 0.15 ± 0.02 s) as DFP in the natural EQD site (t_1/2: 2.0 ± 0.1 s). In addition, the maximum ΔA₃₅₀ of the Fe(III)O products observed in Fe2 EED were only ∼50% that of WT Fe2 EQD (Fig. 2 C and D). Identification of the product of the Fe2 EED site in ferritin catalytic centers is outside the scope of this report, but existence of the product illustrates both a possible evolutionary intermediate between ferritin Fe(II)/O₂ oxidoreductase sites and diiron cofactor oxygenase sites and the importance of Gln¹³⁷ in control of the DFP reaction pathway for iron mineralization.

Discussion

Selectivity of Fe(II)/O₂ chemistry at the diiron substrate sites in ferritin [Fe(II)/O₂ oxidoreductase] catalytic centers, compared to the diiron cofactor centers in oxygenases that produce the same initial reaction intermediate Fe(III)–O–O–Fe(III), appears to depend on differences in only 2 of the 6 conserved residues at the diiron cofactor sites: (i) a conserved, glutamine in ferritins vs. glutamate in the cofactor sites, and (ii) a set of variable residues, aspartate/serine/alanine in ferritins vs. a conserved histidine in the cofactor sites. The two types of differences in the ferritin catalytic centers residues are associated with a weaker Fe(II) binding environment and a more open structure with a single carboxylate bridge between the two Fe atoms in structures of ferritin diiron substrate sites compared with the stronger binding environment and more closed structures in the oxygenase diiron cofactor sites (Fig. 1 and refs. 18 and 21–23). The kinetic studies presented here show that each natural variation among ferritin active sites alters kinetics of the Fe(II) + O₂ reaction, but not the mechanism, and that each diiron catalytic center appears to have coevolved with the sequence of the polypeptide backbone, based on comparisons of the catalytic properties of natural variants in residue 140, aspartate, serine, or alanine in a single polypeptide sequence (frog M) with that in the natural polypeptide (frog M Asp¹⁴⁰, frog H Ser¹⁴⁰, and human H Ala¹⁴⁰). However, substitution of the two Fe ligands from the diiron cofactor sites for the ferritin-specific residues effectively inactivated the ferritin oxidoreductase activity, which is facilitated by the same ligand set in the oxygenases, MMO, RNR, and Δ⁹ desaturase (11–14, 21–23).

Positive cooperativity, observed during DFP formation (Table 1), and in earlier studies, e.g., ref. 20, likely reflects interactions among the multiple catalytic sites of ferritin as recently observed during diferrous binding to the active sites in the absence of dioxygen (17). Local protein conformational changes, indicated by the differences among the Fe–Fe distances (EXAFS) of diferrous (3.4 Å), DFP (2.5 Å) and diferric mineral precursor (3.9 Å) (10), may also contribute to cooperativity.

One of the residues at the ferritin-specific Fe2 site, residue 140, varies coincidentally with tissue-specific or organelle-specific ferritin expression (24, 25), without altering the reaction pathway (Scheme 1) (9, 10, 29). However, each ferritin active site in the homologous background had variable k_cat and DFP kinetic stability, but constant K_app (Table 1). In contrast, when the polypeptide sequence was the same, and just residue 140 was changed at the catalytic center by using the natural active center residues Ser and Ala as substitutions for Asp, the WT match with the polypeptide backbone, a different set of kinetic properties was observed (Table 1), indicating contributions of the global protein sequence to catalysis. Sequence and structure comparisons indicated that proximal residue 143, covarying with ferritin Fe2 site residue 140 (Fig. 3 A and B), might contribute to catalysis. However, changing residue 143 had no significant effects on DFP formation (T.T. and E.C.T., unpublished observations). Key residues matched to variations in residue 140 that modulate catalysis in ferritins are, thus, more likely to be some distance from the catalytic centers as discussed for other protein catalysts (reviewed in ref. 30).

Fig. 3.

Models of the environments around ferritin Fe2 site related to Fe(II)/O2 oxidoreductase activity. (A) The Fe2 environment with Asp at variable position 140. Arg¹⁴³ hydrogen-bonded to Asp¹⁴⁰ and His⁵⁴, near Asp¹⁴⁰, covary among ferritins. (B) The environment of variable residue with Ala^140(144); same view as C. The hydrogen-bond network Arg¹⁴³···His⁵⁴ would be absent with Glu¹⁴³ and Gln⁵⁴. [The figure was created from PDB 1MFR, frog M-ferritin, for Asp¹⁴⁰, His⁵⁴, Arg ¹⁴³ Mg(II) cocrystals (18) and PDB 2CEI, Ala^140(144), Glu^143(147), and Gln⁵⁴⁽⁵⁸⁾, human H-ferritin Zn(II) cocrystals (35).] Dotted lines indicate presumed hydrogen bonds. (C) The inner cavity surface around the catalytic center. Fe2 site Asp¹⁴⁰ is located at the inner cavity surface; Mg(II) (green) is exposed to cavity solvent. (D) A plausible hydrogen bonding network: the invariant ferritin Fe2 amide side chain of Gln¹³⁷ hydrogen-bonds to the carboxylate group of conserved Glu¹⁰³, which is hydrogen-bonded with the conserved Tyr³⁰ phenoxyl.

Histidine at variable position 140 in the Fe2 site of the ferritin Fe(II)/O₂ oxidoreductase centers marked a limit in the range of functional residues possible at Fe2 in ferritin catalytic centers, because His¹⁴⁰ ferritins were ineffective not only in the coupling reaction that forms the DFP catalytic intermediate but also in Fe(II) oxidation itself (Fig. 2A and Table 1); histidine is conserved at the analogous site in diiron cofactors. Preliminary X-ray structural analysis (protein crystallography) on the Gln¹³⁷ → Glu/Asp¹⁴⁰ → His mutant showed that the active site was intact and the His¹⁴⁰ configuration was similar to Asp¹⁴⁰ in WT (T.T, H.-L. Ng, E.C.T., and T. Alber, unpublished observation). In addition, CD/MCD data indicate binding of 2Fe(II) in the catalytic center of Gln¹³⁷ → Glu/Asp¹⁴⁰ → His mutant (J. K. Schwartz, T.T, E.C.T., and E. I. Solomon, unpublished observation). Therefore, the activity loss of His¹⁴⁰ ferritin is independent of Fe(II) binding or collapse of the active site structure. Several ferritin properties could explain the inactivation caused by inserting histidine at position 140. First, because residue 140 is exposed to nanocage cavity (Fig. 3C) it is in contact with the air-saturated, saline buffer in the cavity (31, 32), suggesting a role for residue 140 in active site interactions with small molecules such as O₂ and H₂O and possibly with the diferric oxo/hydroxo catalytic product; His¹⁴⁰ might physically block transfer of product to the cavity. Second, the basicity of histidine is modulated by multiple protein side chain interactions in diiron cofactor sites (21–23) that are absent from WT ferritin around His⁶¹ in the Fe1 site (18) (Fig. 1) and may also be absent from the engineered His¹⁴⁰ ferritin. A strong Fe–histidine interaction in the ferritin Fe2 site may impede O₂ binding or electron transfer between Fe(II) and O₂ or product release. Such effects illustrate the structural differences between the diiron cofactor centers in MMO, RNR, and Δ⁹ desaturase and the Fe(II)/O₂ oxidoreductase centers in ferritin that coincide with different pathways from the common DFP catalytic intermediate.

Gln¹³⁷, the conserved, ferritin-specific, active site residue with an amide side chain, contrasts with the glutamate carboxylate side chain at the equivalent site in the diiron cofactor oxygenases and controls DFP formation in ferritins, based on effects of glutamate substitution (Fig. 2D and Table 1), even though glutamate in the context of the diiron cofactor enzymes also forms DFP (11–14). Because the DNA code for glutamine and glutamate can differ by only one nucleotide (glutamine/glutamate codons can be CAG/GAG) (10, 18), ferritin with Glu¹³⁷ could model an evolutionary intermediate on the path between diiron substrate and diiron cofactor catalytic centers. The important role of Gln¹³⁷ in DFP-mediated iron mineralization by ferritin reflects the multiple interactions of Gln¹³⁷ within the ferritin protein. First, crystal structures suggest a conserved hydrogen-bond network, Gln¹³⁷···Glu¹⁰³···Tyr³⁰ (Fig. 3D). Gln¹³⁷ → Glu catalysis is unusual in producing Fe(III)O products with a lower average absorbance than normal but with a normal rate of Fe(II) oxidation itself (Table 1), whereas substitutions for Glu¹⁰³ or Tyr³⁰ decrease Fe(II) oxidation (20, 33–35). The second Fe1–Fe2 bridge, formed by glutamate in the diiron oxygenases at the position analogous to Gln¹³⁷ in ferritins (Fig. 1), not only is prevented by the amide but the side chain orientation itself differs because of constraints imposed by the hydrogen bond network. Second, in a diferrous model for the catalytic center ferritin developed with variable-temperature, variable-field magnetic circular dichroism (VTVH MCD) (17), the carbonyl O in Gln¹³⁷ is coordinated to Fe(II) in the Fe2 site. Third, the VTVH MCD model (17) indicates bonding between the amide of Gln¹³⁷ and H₂O coordinated to Fe1 as shown in Fig. 1. Protonation at the DFP active site, required for decay of the intermediate and product release (15, 29), would be facilitated by a hydrogen bond between Gln¹³⁷ and H₂O coordinated to Fe1 (17). In addition, the oxo or hydroxo bridge suggested by the short Fe1–Fe2 distance (2.5 Å) in DFP (10) might also be facilitated by a hydrogen bond between Gln¹³⁷ to H₂O at Fe1. Thus, the particular orientation of Gln¹³⁷ imposed by the hydrogen bond network could facilitate the proton transfer associated with DFP formation from Fe(II) + O₂, or DFP decay to form Fe(III)–O(H)–Fe(III), or possibly the release of the catalytic product, Fe(III)–O(H)–Fe(III) during catalysis.

Catalysis at the ferritin Fe(II)/O₂ oxidoreductase centers depends on combining ferritin-specific residues in the Fe2 site with residues common to diiron cofactor oxygenases. The reaction of Fe(II) with dioxygen that produces the DFP intermediate occurs for both diiron substrate and diiron cofactor protein catalysts. The ferritin-specific Fe2 site uses Glu¹⁰³, also present in diiron cofactor sites, ferritin-specific conserved residue (Gln¹³⁷), and a ferritin-specific set of ferritin-variable residue Asp¹⁴⁰, Ser¹⁴⁰, or Ala¹⁴⁰, but cannot function with the cofactor residue, His¹⁴⁰. Gln¹³⁷ is required for DFP formation/decay and for the more open catalytic site structure (Fig. 1). The short Fe1–Fe2 distance (2.5 Å) and small Fe–O–O angle (8, 10) in the ferritin DFP compared to DFP in diiron cofactor enzymes, appears to facilitate peroxide release rather than the formation of high-valence reaction intermediate observed in diiron cofactor enzymes. The results reported here show that weaker iron binding at the ferritin-specific Fe2 site is required for the formation of DFP in ferritin and may contribute to a particular reaction pathway distinct from the heme-containing Bfr with an active site containing the diiron cofactor ligands (2) and a geometry of the ferritin DFP that is incompatible with the diiron cofactor ligand set. The two ferritin-specific ligands in the Fe2 site of the catalytic center, Gln¹³⁷/Asp¹⁴⁰, are related to Glu and His, the two diiron cofactor site ligands at equivalent positions, by 2 of 3 codon nucleotides in DNA. Thus, the results also illustrate the potential simplicity of evolving active sites for diiron cofactors or diiron substrates.

Materials and Methods

Cloning and Mutagenesis.

Site-directed mutagenesis used QuikChange Mutagenesis (Stratagene) on frog M-ferritin in pET-3a as previously described (20). Primers were Q¹³⁷E (EED): forward, 5′-GAA TAT CTG GAG GAA GAG GTG AAG GAT ATT AAA CGC-3′; reverse, 5′-GCG TTT AAT ATC CTT CAC CTC TTC CTC CAG ATA TTC-3′; D¹⁴⁰H (EQH): forward, 5′-G GAA CAG GTG AAG CAT ATT AAA CGC-3′; reverse, 5′-GCG TTT AAT ATG CTT CAC CTG TTC C-3′; Q¹³⁷E/D¹⁴⁰H (EEH): forward, 5′-GAA TAT CTG GAG GAA GAG GTG AAG CAT ATT AAA CGC-3′; reverse, 5′-GCG TTT AAT ATG CTT CAC CTC TTC CTC CAG ATA TTC-3′, D¹⁴⁰A (EQA): forward, 5′-GAA TAT CTG GAG GAA CAG GTG AAG GCT ATT AAA CGC-3′; reverse, 5′-GCG TTT AAT AGC CTT CAC CTG TTC CTC CAG ATA TTC-3′, D¹⁴⁰S (EQS): forward, 5′-GAA TAT CTG GAG GAA CAG GTG AAG TCT ATT AAA CGC-3′; reverse, 5′-GCG TTT AAT AGA CTT CAC CTG TTC CTC CAG ATA TTC-3′ (underlines indicate mutation sites). The human H-ferritin coding region, obtained by PCR amplification of Marathon-Ready HeLa cell cDNA (Clontech), was cloned in pUC19, followed by PCR to insert NdeI and BamHI sites, digestion with NdeI and BamHI, and cloning into pET-3a DNA, which was analyzed to confirm the sequence of human H-ferritin insert.

Protein Expression.

Recombinant ferritin samples, isolated as previously described (20, 36–38) from Escherichia coli BL21(DE3) incubated in LB medium, used 20 mM Bis-Tris-propane, pH 7.5, during sonication, coagulation of soluble nonferritin proteins (15 min, 65 °C), concentration by precipitation with 45% saturated (NH₄)₂SO₄, ion-exchange fractionation (mono-Q column, Amersham Pharmacia) with a linear NaCl gradient of 0–1.0 M, and detection with SDS/PAGE. Pooled, homogenous, ferritin fractions were dialyzed against 100 mM Mops (pH 7.0), 100 mM NaCl, stored at 4 °C, and analyzed within 2 weeks.

Kinetic Analysis of Ferrous Oxidation and DFP Formation.

Fresh solutions of FeSO₄ (0.1–5 mM in 1.0 mM HCl) were mixed with an equal volume of ferritin solutions [0.1 mM subunits in 200 mM Mops (pH 7.0), 200 mM NaCl] in a stopped-flow UV/vis spectrophotometer (Applied Photophysics) at 20 °C and 4 °C. Data were collected for A₆₅₀, the absorbance maximum of DFP in ferritin (9), and A₃₅₀, the midpoint of the Fe(III) oxy absorption, an unresolvable mixture of DFP, Fe(III) oxo/hydroxo dimer, and Fe(III) biomineral (20); the absorbance range 300–420 nm is often used in ferritin F_ox analyses (20, 29, 33, 39–42). Initial rates (V_i) were calculated by linear fitting to the early time domain (∼50 ms) of A₃₅₀ and A₆₅₀ data. V_i values obtained at varied Fe concentrations were fitted to the Hill equation (20), V_i = V_max[Fe(II)]ⁿ/(K_app + [Fe(II)]ⁿ), where n is the Hill coefficient. The k_cat values (= V_max/[ferritin]) for DFP formation were estimated by two fitting methods. First, by computing the k_cat value (= V_max/[ferritin]) for DFP formation with V_max and the DFP molar extinction coefficient (ε) of 1,000 M¹·cm¹ at 650 nm (9). Second, k_cat was determined by using the extinction coefficient of each protein, under the conditions used here, computed from fitting the kinetic trace of A₆₅₀ under single-turnover conditions (48 Fe per ferritin) with a double-exponential equation assuming two irreversible consecutive reactions; ΔA₆₅₀ = A₀k_fwd × [exp(−k_fwdt) − exp(−k_decayt)]/(k_decay − k_fwd) (9), where A₀ = ε × [ferritin]. The computed values for ε were within experimental error of 1,000 M¹·cm¹, ε = 1,040 ± 80, 960 ± 140, and 1,060 ± 80 M¹·cm¹ for frog M WT (EQD-Fr M), D140A, and human H WT, respectively, except for EQS-Fr M, where ε = 720 ± 120 M¹·cm¹; nonrandom residuals indicated that the goodness of fits were poorer for the double-exponential fitting model. The values of k_cat were significantly (P < 0.01) different from the control WT frog M by either method. DFP decay, fit well by a single-exponential (20) and less well with the double-exponential equation (see above), gave similar results, using the equation t_1/2 = ln2/k_decay. Fitting used Igor software (WaveMetrics). Except for EQA-Hu H, where only one protein preparation was studied, the data were from 3–5 protein preparations; each protein preparation was analyzed 2–7 times