Iron and Hydrogen Peroxide Detoxification Properties of DNA-binding Protein from Starved Cells

The DNA-binding proteins from starved cells (Dps) are a family of proteins induced in microorganisms by oxidative or nutritional stress. Escherichia coli Dps, a structural analog of the 12-subunit Listeria innocua ferritin, binds and protects DNA against oxidative damage mediated by H2O2. Dps is shown to be a Fe-binding and storage protein where Fe(II) oxidation is most effectively accomplished by H2O2 rather than by O2 as in ferritins. Two Fe2+ ions bind at each of the 12 putative dinuclear ferroxidase sites (PZ) in the protein according to the equation, 2Fe2+ + PZ→ [(Fe(II)2-P] FS Z + 2 + 2H+. The ferroxidase site (FS) bound iron is then oxidized according to the equation, [(Fe(II)2-P] FS Z + 2 + H2O2 + H2O → [Fe(III)2O2(OH)-P] FS Z − 1 + 3H+, where two Fe(II) are oxidized per H2O2 reduced, thus avoiding hydroxyl radical production through Fenton chemistry. Dps acquires a ferric core of ∼500 Fe(III) according to the mineralization equation, 2Fe2+ + H2O2 + 2H2O → 2Fe(III)OOH(core) + 4H+, again with a 2 Fe(II)/H2O2 stoichiometry. The protein forms a similar ferric core with O2 as the oxidant, albeit at a slower rate. In the absence of H2O2 and O2, Dps forms a ferrous core of ∼400 Fe(II) by the reaction Fe2+ + H2O + Cl− → Fe(II)OHCl(core) + H+. The ferrous core also undergoes oxidation with a stoichiometry of 2 Fe(II)/H2O2. Spin trapping experiments demonstrate that Dps greatly attenuates hydroxyl radical production during Fe(II) oxidation by H2O2. These results and in vitro DNA damage assays indicate that the protective effect of Dps on DNA most likely is exerted through a dual action, the physical association with DNA and the ability to nullify the toxic combination of Fe(II) and H2O2. In the latter process a hydrous ferric oxide mineral core is produced within the protein, thus avoiding oxidative damage mediated by Fenton chemistry.

All aerobic organisms have evolved a variety of complex defense and repair mechanisms to protect their DNA from oxidative damage because of reactive oxygen species such as HO⅐, O 2 . , and H 2 O 2 . These include highly regulated enzymatic systems that recognize and repair damaged DNA or that prevent damage by detoxifying the reactants that produce radicals (1)(2)(3)(4). These inducible responses are associated with de novo protein synthesis, an energy requiring process. However, when bacteria are starved, their ability to cope with environmental assault by de novo protein synthesis becomes compromised by the lack of nutrients (1). A class of nonspecific DNA-binding Dps 1 (pexB) proteins is expressed in bacteria and accumulated to high levels prior to conditions of oxidative or nutritional stress, making the cell in stationary phase more resistant to H 2 O 2 than actively growing cells (1,(5)(6)(7)(8)(9). Studies have shown that Dps plays a central role in protecting DNA from oxidative damage both in vitro and in vivo by directly binding to DNA (1, 6 -9). It has been suggested that the protective activity of Dps may stem from its ability to sequester iron ions, but direct experimental evidence for this proposal has been lacking (1,7,10). Dps has a shell-like structure of 3:2 tetrahedral symmetry assembled from 12 identical subunits with a central cavity measuring ϳ45 Å in diameter (7). Each Dps subunit is folded into a 4-helix bundle through hydrophobic interactions. The 3:2 symmetry of the assembled protein leads to two inequivalent environments along the 3-fold axes. One corresponds to the 3-fold interactions found in the 24-mer ferritins having 4:3:2 octahedral symmetry and involves the N-terminal end of the subunit, whereas the other 3-fold interaction takes place at the C-terminal end of Dps but is not typical of the ferritins (7).
Escherichia coli Dps and the relatively recently discovered 12-mer Listeria innocua ferritin assemble in essentially the same way but share only 19% identity in primary sequence (10,11). Both proteins possess a negative electrostatic potential on the inner surfaces of their shells, providing an ideal microenvironment for iron mineralization (7,10). Unlike typical 24-mer ferritins, the putative dinuclear ferroxidase sites of L. innocua ferritin are not located in the four-helix bundle of individual monomers but rather have ligands provided by two symmetryrelated subunits (10). Most ligands comprising the putative ferroxidase center of L. innocua ferritin are conserved in E. coli Dps as well (7,10). Pb 2ϩ is bound at this location in the heavy atom derivative of the Dps crystal (7). The large "ferritin-like" 3-fold pores of 8 Å diameter are lined with Asp 121 , Asp 126 , and Asp 130 from the three symmetry-related subunits and are con-served between the two proteins (except that Asp 121 is replaced by Lys in Dps). The negatively charged hydrophilic pores are likely pathways for iron entry into the protein cavity (7,10). L. innocua ferritin has been shown to exhibit ferroxidase activity (12,13) and can rapidly accumulate up to ϳ500 iron atoms within its central cavity when O 2 is the oxidant (11).
The striking similarity in overall structure between E. coli Dps and L. innocua ferritin and the protective effect of E. coli Dps on DNA against oxidative damage mediated by Fenton chemistry (1) suggest that E. coli Dps might also function as an iron-binding and storage protein. However, to date the iron binding, oxidation, and hydrolysis/mineralization reactions of Dps have not yet been investigated. In the present study, we describe experiments that demonstrate a ferritin-like function for Dps. Unexpectedly, O 2 was found to be a relatively poor oxidant for Fe(II) in Dps, the rate of O 2 consumption being only marginally faster than Fe(II) autoxidation in the absence of protein. In contrast to O 2 , H 2 O 2 causes rapid and complete oxidation of Fe(II). The initial rate of iron oxidation reaches a maximum at 24 Fe(II)/Dps, a result implying that there are 12 dinuclear ferroxidase sites in Dps, one per each of the 12 subunits. As in E. coli bacterioferritin and L. innocua ferritin (12,14), the ferroxidase activity of Dps is not readily regenerated upon standing. A stoichiometry of 1 H 2 O 2 per 2 Fe(II) oxidized is obtained in both ferroxidase and mineralization reactions of Dps, thus avoiding the production of the HO ⅐ radical through the Fenton reaction as follows.
By using spectrophotometry, pH-stat, sedimentation velocity measurements, and iron analysis, the protein was shown to load up to ϳ500 Fe(III) within its cavity. The results are discussed in terms of possible mechanisms of core formation in Dps and of the relationship between its ferritin-like function and DNA protection from oxidative damage.

MATERIALS AND METHODS
Preparation and purification of Dps were performed as described by Almiron et al. (6) with one modification. The Sephadex G-100 gel filtration step was replaced by dialysis against 20 mM Tris-HCl, pH 8.0. At this low ionic strength Dps precipitates and can be collected easily after centrifugation at 15,000 rpm for 10 min. Dps resuspended in 50 mM Tris-HCl, 2 M NaCl, pH 8.0, was subjected to gel filtration on Sepharose 6B. ApoDps concentrations were determined spectrophotometrically using a molar absorptivity of 2.59 ϫ 10 5 M Ϫ1 cm Ϫ1 at 280 nm on a 12-mer protein basis (13). The concentration of freshly prepared solutions of hydrogen peroxide was determined from the amount of O 2 evolved upon addition of catalase (Roche Molecular Biochemicals) as measured by Clark electrode oximetry or from its absorbance at 240 nm (⑀ ϭ 43.6 M Ϫ1 cm Ϫ1 ) (15). EMPO was purchased from Oxis Research (Portland, OR). All chemicals were reagent grade or better, and were used without further purification.
The electrode oximetry/pH-stat apparatus and standardization reaction for its use have been described in detail elsewhere (14,16). In the present experiments, the consumption of O 2 during Fe(II) oxidation was followed at 25°C by a Clark-type oxygen microelectrode while maintaining the pH at 7.0 with a pH-stat. To accurately measure initial rates of proton production, a pH-stat proportional band setting of 0.1 was used; however, the rapid response of pH-stat at this setting resulted in an overshoot of the stoichiometric end point at the end of the reaction. Therefore, the stoichiometry was determined in a separate experiment using a proportional band setting of 0.2 or 0.5 where the response of pH-stat was slower and an accurate end point could be obtained. Typical conditions for the experiments were 0.1, 0.2, 1, 2, or 4 M Dps protein in 0.5 mM Mops and 150 or 200 mM NaCl, pH 7.0 (controlled by pH-stat), with increments of 6 -24 or 100 -1000 Fe(II) per protein added at 25°C as freshly prepared 10 -20 mM FeSO 4 or Fe(NH 4 ) 2 (SO 4 ) 2 (Baker Scientific) in pH 3.5 water. The use of 0.5 mM Mops buffer in the protein solution lends stability to the pH-stat control without significantly buffering the solution. Small corrections for free acid in the ferrous sulfate were made in all calculations.
The ultraviolet-visible difference spectrophotometric titration and kinetic experiments of Fe(II) oxidation by H 2 O 2 were performed on a Cary 500 spectrophotometer. The instrument was zeroed using 0.1 or 0.5 M Dps solution as the blank. Time-dependent absorbance kinetic traces at 25°C were collected using the Cary 500 kinetic Software. The kinetic data were further analyzed with Origin 6.0 software (Microcal Inc.). The initial rates of Fe(II) binding measured by the pH-stat and of Fe(II) oxidation measured by UV absorption spectroscopy were obtained from the linear A 1 term of third-order polynomial curve to the experimental data, namely Here Y is either the change in absorbance ⌬A at 305 nm or the amount of NaOH autotitrated into the sample solution at time t in seconds (16).
EPR spectra were recorded on a laboratory assembled EPR spectrometer based on a Bruker ER 041 XK-H X-band microwave bridge operating at 9.24 GHz with 100 kHz field modulation. Samples in 1-mm inner diameter quartz capillaries were placed in a Varian TE 102 cavity for measurement at room temperature. Typical spectrometer parameters were: microwave power, 5.0 mW; modulation amplitude, 0.8 G; time constant, 0.3; scan rate, 7.14 G s Ϫ1 . In the EMPO spin trapping experiments for hydroxyl radical (17), all spectra were recorded ϳ1 min after addition of the last reagent. The concentrations of reactants are indicated in the figure captions.
To prepare the ferric core of Dps using H 2 O 2 as the oxidant, 500 Fe(II) (as 12.5 mM FeSO 4 ) were added aerobically to 0.25 M apoDps in 50 mM Mops, 200 mM NaCl, pH 7, in 10 increments of 50 Fe(II)/Dps followed by 1 H 2 O 2 (as a 12.5 mM solution) per 2 Fe(II) 2 min later with a 10-min interval between iron additions. The protein solution was stirred only during the additions of Fe(II) and H 2 O 2 . The holoprotein prepared in this way was dialyzed three times against buffer and then analyzed for protein content by the Advanced Protein Assay (cytoskeleton.com) and the iron content by the ferrozine method (18). To prepare the ferrous core of Dps, Fe(II) was added in 8 increments of 50 Fe(II)/ Dps at 10-min intervals under an argon atmosphere and in the presence of 1 mM dithionite, Na 2 S 2 O 4 , in 50 mM Mops, 200 mM NaCl, pH 7.0, to give a total of 400 Fe(II)/protein. The iron(II) containing protein was ultrafiltered 3 times under argon with a 1000-fold total volume change and analyzed for protein and iron contents.
Formation of a ferric core with O 2 as oxidant was assessed by adding two increments of 200 Fe(II)/Dps to Dps solutions (1 M) equilibrated in air at 20°C. Ferrous ammonium sulfate solutions prepared in Thunberg tubes and kept under nitrogen were used. The iron incorporation reaction was followed at 310 nm; after reaching a constant absorbance (about 2 h) the samples were analyzed by sedimentation velocity.
Sedimentation velocity studies were carried out using the absorbance optics on a Beckman XLI analytical ultracentrifuge. Experiments were conducted at 30,000 rpm and 20°C at Dps concentrations between 0.25 and 1.0 M. Radial absorbance scans were obtained at both 280 and 310 nm using the continuous scan mode to provide an effective radial resolution of 30 m. The constant ratio of absorbance at 280 nm to that at 310 nm during the centrifugation run of the ferrous core sample confirmed that oxidation of the Fe(II) had not occurred. Sedimentation coefficient distributions, g(s*) were obtained using the linear least-squares algorithm (ls-g(s)) incorporated in Sedfit (version 8.3) (19). Sedimentation coefficients were interpreted using standard methods (20).
Fluorescence spectra were measured on 4 M Dps solutions in 50 mM Mops-NaOH, pH 7.0, containing 150 mM NaCl. The measurements were carried out at 25°C in gas-tight cells under a nitrogen atmosphere using a Fluoromax fluorimeter (Spex Industry). In the titration experiments, anaerobic 0.01 M FeSO 4 solutions were added in increments corresponding to 12, 24, 36, 48, and 100 Fe(II)/Dps. DNA protection from oxidative damage in vitro was assessed using pUHE21-2 plasmid DNA (3900 bp, 10 nM), purified by a Qiaprep spin plasmid miniprep kit (Qiagen, Chatsworth, CA). The total volume reaction was 10 l in 20 mM Tris-HCl pH 7.5. Dps (3 M) was allowed to interact with plasmid DNA for 10 min prior to introduction of FeSO 4 (50 M) and H 2 O 2 (10 mM). The reaction mixtures were incubated for 15 or 30 min at room temperature; the reaction was stopped by incubation with 2% SDS at 85°C for 5 min. Dps was extracted with phenol and plasmid DNA was resolved by electrophoresis on 1% agarose gel. The gel was stained with ethidium bromide and imaged by ImageMaster VDS (Amersham Biosciences).

Fe(II)
Binding to Dps-The anaerobic addition of Fe(II) to E. coli Dps produced no discernable change in the UV-visible spectrum, but resulted in significant quenching of the intrinsic protein fluorescence (Fig. 1A). It also resulted in a pH drop of the solution from the production of protons associated with Fe(II) binding to the protein. Therefore, a careful anaerobic titration of the protein with Fe(II) was carried out and proton production monitored by autotitration with the standard base, while maintaining the pH at 7.0 with the pH-stat apparatus. The initial rate of proton production as a function of the Fe(II)/ Dps ratio is plotted as curve a in Fig. 1B where each data point represents a separate Dps sample to which a single addition of Fe(II) was made. The initial rate of H ϩ production reaches a maximum at 24 Fe(II)/Dps, indicating that there are 24 Fe(II)binding sites on the protein, 2 per subunit (Fig. 1B, curve a). The same result was obtained from the fluorescence titration data (Fig. 1A).
A H ϩ /Fe(II) stoichiometry of 1.0 Ϯ 0.13 (n ϭ 3) was obtained (Fig. 2, first 24 M Fe(II) addition). A proton count of 1 H ϩ / Fe(II) has been reported for Fe(II) binding to L. innocua ferritin also (12), however, the production of protons from Fe(II) binding to L. innocua ferritin as well as to E. coli bacterioferritin occurs in less than 2 s (the 50% response time of the electrode) (12,14) compared with 40 Ϯ 3 s (t1 ⁄2 ϭ 14 s) for Dps at similar protein and iron concentrations (1 M Dps, 24 M Fe(II)) ( Fig.  2). We write the Fe 2ϩ binding (phase I) reaction in Dps as where [Fe(II) 2 -P Zϩ2 ] FS represents a diFe(II)-protein complex at each of the 12 putative ferroxidase sites. Fe(II) Oxidation-Initial experiments on Fe(II) oxidation in Dps were carried out using O 2 as the oxidant. Fig. 3 shows that the rate of O 2 consumption monitored by electrode oximetry was slow when Fe(II) was added aerobically to the apoprotein, being only marginally faster than Fe(II) autoxidation in buffer alone. Dps contrasts with all known ferritins investigated to date, including L. innocua ferritin, where O 2 is an efficient and rapid oxidant for Fe(II) (12,14,16,(21)(22)(23)(24). For example, when compared on a subunit basis (2 Fe(II)/subunit) under the conditions in Fig. 3, Fe(II) oxidation in Dps consumes 1.2 ϫ 10 Ϫ2 O 2 /subunit/min compared with 10 O 2 /subunit/min for human H-chain ferritin (16), 2 an 830-fold difference between proteins. Furthermore, addition of 24 Fe(II)/Dps aerobically to a 0.5 M protein sample caused a minimal absorbance change at 305 nm over a period of 7 min in contrast to ferritins where rapid Fe(II) oxidation by O 2 leads to formation of -oxo-bridged Fe(III) species that absorb at this wavelength (14,16). A gradual increase in absorbance with Dps occurs over an extended time period, however.
Experiments were subsequently undertaken to determine whether H 2 O 2 might be a more efficient oxidant of Fe(II) in Dps than O 2 . Fig. 4 shows the addition of 12 M Fe(II) to an aerobic solution of 0.50 M Dps. In the presence of air there was little change in absorbance; however, upon the addition of H 2 O 2 to the solution (2 H 2 O 2 /Fe(II)), the absorbance quickly increased to 0.35 units (Fig. 4). With H 2 O 2 as the oxidant under the conditions in Fig. 4, the specific rate of Fe(II) oxidation was 100-fold larger than for O 2 , 1.8 Fe/subunit/min versus 0.017 Fe/subunit/min, respectively. The molar absorptivity for the observed oxidation product was 29,200 M Ϫ1 cm Ϫ1 per iron at 305 nm, which is almost 10-fold higher than previously observed for the ferritins (ϳ3000 M Ϫ1 cm Ϫ1 per Fe) (14,16), 3 perhaps reflecting the presence of more than one -oxo bridge in the diFe(III) complex that was formed (see "Discussion").   (Fig. 6,  curve a) (Fig. 1B, curve b). The initial rate of Fe(II) oxidation (⌬A/min) reaches a maximum at 24 Fe(II)/Dps in accordance with the Fe(II) binding stoichiometry of 24 Fe(II)/ Dps ( Fig. 1A and 1B, curve a) Fig. 2 shows that 2.0 Ϯ 0.2 H ϩ were produced during the oxidation of the second 24 Fe(II) by H 2 O 2 , consistent with the value of 1.9 Ϯ 0.1 H ϩ per iron measured by pH-stat when 500 Fe(II) were oxidized (data not shown). Accordingly, we write the mineralization (phase III) reaction for Dps as follows.

FIG. 5. Dependence of the spectrum obtained from Fe(II) oxidation by H 2 O 2 on the order of addition of H 2 O 2 and Fe(II) to
To determine the maximum size of the ferric core in Dps, a spectrophotometric titration was carried out with Fe(II) using H 2 O 2 as an oxidant with a 1 H 2 O 2 /2 Fe(II) ratio (Fig. 7, curve a). A discontinuity in absorbance at ϳ500 Fe(III)/Dps was observed, suggesting that Dps can accumulate up to ϳ500 Fe(III) with H 2 O 2 as the oxidant 4 as was also found for L. innocua ferritin with O 2 as the oxidant (11). Iron analysis following dialysis of the protein ("Materials and Methods") gave 472 Ϯ 11 Fe(III)/Dps (n ϭ 3). The molar absorptivity at 305 nm for mineralized iron in Dps was 3120 M Ϫ1 cm Ϫ1 per iron compared with values in the range 2030 -3540 M Ϫ1 cm Ϫ1 per iron for horse spleen ferritin, human H-chain ferritin, and E. coli bacterioferritin (14,16). 3 The formation of an iron core was confirmed by analytical ultracentrifugation of apoDps and Dps loaded with iron (Fig.  8A). ApoDps sediments as a single species with s 20,w 0 ϭ 9.6 S similar to the value for apo-L. innocua ferritin (11). When 500 Fe(III) were added to apoDps in 10 increments of 50 Fe(II)/Dps at intervals of 10 min followed by 25 H 2 O 2 /Dps each time, the apoDps peak at 9.6 S was replaced by components sedimenting at s 20,w 0 ϭ 14.8, 21.9 and 28.5 S. Because s 20,w 0 is expected to increase by ϳ1 Svedberg per 100 Fe added to the core (25), we assign the 14.8 S species to Dps containing ϳ500 Fe(III). The minor component at 21.9 S is assigned to a cross-linked dimer of the 14.8 species, predicted to sediment at 20.9 S (ϭ1.414 ϫ 14.8 S). The lesser component at 28.5 S may be a trimeric species. Although elucidation of the specific nature of the crosslinks is beyond the scope of the present study, they probably arise from iron-induced radical chemistry (see below). Crosslinking is commonly observed for the ferritins (26 -28).
To determine whether Dps is capable of forming a ferric core with O 2 as oxidant, 400 Fe(II)/Dps were added to 1 M Dps solutions in two successive increments of 200 Fe(II) atoms. After about 2 h, a core was formed as indicated by the sedimentation velocity of the sample (Fig. 8B). The size of the core was similar to that obtained with H 2 O 2 as oxidant (s 20,w 0 ϭ 14.4 S), but the molecular mass distribution was somewhat wider; the latter finding indicates that the reaction leading to core formation was less cooperative in the case of oxygen and leads to increased heterogeneity in the sample.
Fe(II) Mineralization of Dps-To determine whether Dps is capable of forming a ferrous core, samples of Dps were placed in the pH-stat and various amounts of Fe(II) were added anaerobically. The production of H ϩ associated with Fe(II) incorporation into Dps was measured (Fig. 7, curve b). Fe(II) addition in the absence of Dps produces no H ϩ ions. A discontinuity in H ϩ production occurs at about 400 Fe(II)/Dps, suggesting that Dps can acquire up to 400 Fe(II). The H ϩ /Fe(II) ratios for points beyond 400 Fe(II)/Dps represent averages of values for Fe(II) bound (1 H ϩ /Fe(II)) and Fe(II) unbound (0 H ϩ /Fe(II)), hence the linear decrease in stoichiometric ratio beyond the end point. We attempted to measure ferrous core formation spectrophotometrically but the ferrous core was UV transparent from 250 to 400 nm. However, the analytical ultracentrifugation data of Fig. 8B demonstrates that Fe(II) was incorporated into the protein. The iron containing protein sediments as a single somewhat heterogeneous component at 13.7 S consistent with an average core size of 400 Fe(II). Iron analysis of the sample following ultrafiltration gave 381 Ϯ 13 Fe(II)/Dps (n ϭ 3). Because 1 H ϩ was produced per Fe(II) incorporated in the core (Fig. 7B) and the anion chloride is required for electroneutrality, we write the ferrous core reaction as follows.

Reduction of Hydroxyl Radical Production via Fenton
Chemistry-The ability of Dps to attenuate hydroxyl radical production from the Fenton reaction (Equation 1) was demonstrated through spin trapping experiments. Spectrum A of Fig. 9 is that of the EMPO-OH adduct in a control experiment using the chelator DTPA while adding reagents in the sequence DTPA ϩ EMPO ϩ H 2 O 2 ϩ Fe(II). The one-electron oxidation of Fe(II)-DTPA by H 2 O 2 is an efficient generator of hydroxyl radical (17) and an intense spectrum from trapped HO ⅐ was observed (Spectrum A). Spectrum B corresponds to another control experiment where the oxidation of Fe(II) was carried out in the absence of DTPA for the addition sequence EMPO ϩ H 2 O 2 ϩ Fe(II). Again a significant amount of EMPO-OH was produced, corresponding to 65% of that when DTPA was present. With Dps, a weaker EPR signal (25% of spectrum A) was produced for the addition sequence Dps ϩ EMPO ϩ H 2 O 2 ϩ Fe(II), Fe(II) being added immediately after H 2 O 2 . No EPR signal was observed (spectrum D) when Fe(II) was first allowed to bind to the protein before addition of H 2 O 2 for the sequence Dps ϩ EMPO ϩ Fe(II) ϩ H 2 O 2 . Relative to either control experiment, the presence of Dps substantially reduces the amount of hydroxyl radical trapped by EMPO (c.f. spectra A and B versus C and D).
To establish whether the ability of Dps to reduce hydroxyl radical production results in protection of DNA from cleavage because of Fe(II)-mediated Fenton reactions, an in vitro DNA damage assay was set up in which the effect of a combination of 50 M Fe(II) plus 10 mM H 2 O 2 on the integrity of plasmid pUHE21-2 (3900 bp) in the presence and absence of Dps was assessed. Prior to carrying out these assays, the amount of Dps necessary to saturate DNA by physical association was established in gel retardation experiments (data not shown). Fig. 10 shows that DNA was fully degraded in the absence of Dps (lane 2) relative to the control (lane 1). In contrast, the presence of Dps in sufficient amounts to saturate DNA yields essentially complete protection (lanes 3 and 4). The order of Fe(II) and H 2 O 2 addition has little influence on the protective effect exerted by Dps (lane 3 versus 4). The present in vitro assay differs from that reported in Ref. 1 where exposure of DNA to Fe(III)/ EDTA/H 2 O 2 resulted in DNA single strand nicks. DISCUSSION The present studies demonstrate that Dps possesses ferritinlike function and that iron incorporation is a multistep process involving Fe(II) binding, Fe(II) oxidation, nucleation, and growth of the mineral core as in classical ferritins. However, Dps is unique in that O 2 , which is an efficient oxidant for all known ferritins (12,14,16,(21)(22)(23)(24), does not rapidly oxidize Fe(II) in Dps ("Results"). In Dps, rapid and complete oxidation of Fe(II) in both ferroxidase and mineralization reactions occurs with H 2 O 2 at a stoichiometry of 2 Fe(II) per H 2 O 2 (Fig. 6,  Equations 3 and 4). Pairwise oxidation of Fe(II) avoids hydroxyl radical production, a finding confirmed by the spin trapping experiments (Fig. 9). The stoichiometry of Fe(II) oxidation by H 2 O 2 (Fig. 6), spin trapping measurements (Fig. 9), and the DNA protection experiments (Fig. 10)  and suggest a molecular basis for the protective effect of Dps on DNA when the bacterium is under conditions of oxidative stress (1,6).
The observed stoichiometries of 24 Fe(II)/Dps for both Fe(II) binding and oxidation (Fig. 1) are consistent with the binding of two Fe(II) at each of the 12 ferroxidase centers followed by Fe(II) oxidation. The significant quenching of the protein intrinsic fluorescence brought about by Fe(II) binding (Fig. 1A) supports this contention. In fact, the Trp 52 and Trp 160 residues in the 12-mer are all located within a radius of about 4 Å from the ferroxidase centers (7) and thus can function as reporter groups of the Fe(II) binding reaction. Because Fe(II) and its complexes do not readily hydrolyze at the pH 7.0 employed here (29), the production of ϳ1 H ϩ /Fe(II) upon Fe(II) binding is presumably derived from deprotonation of protein ligands, His 51 of the putative ferroxidase site being a primary candidate. The large difference between rates of Fe(II) binding to Dps versus L. innocua ferritin and the slow oxidation of Fe(II) by O 2 in Dps ("Results") but not in L. innocua ferritin (12) may reside in their somewhat different ferroxidase centers. Lys 48 is coordinated to the metal in the Pb 2ϩ derivative of Dps but this residue is absent in L. innocua ferritin (7,10). Despite these differences, the mineralization reaction is faster in both Dps and L. innocua ferritin than the ferroxidation reaction, a property uniquely different from 24-mer ferritins where mineralization is always slower and is preceded by a rapid ferroxidation reaction (14,16).
Whereas the order of addition of Fe(II) and H 2 O 2 has little influence on the protection of DNA (Fig. 10), it does affect Fe(II) oxidation to some extent. Fig. 5 shows that the addition of H 2 O 2 to Dps before Fe(II) results in incomplete formation of the specific Fe(III) complex at the ferroxidase center compared with the reverse order of addition. Because of the relatively slow rate of Fe(II) binding to the protein (Fig. 2), alternate reactions with H 2 O 2 can take place (e.g. Equation 1). The spin trapping experiments (Fig. 9) confirm that hydroxyl radical is produced when H 2 O 2 is added to the protein before Fe(II) (spectrum C) but less than in the controls (spectra A and B). No HO ⅐ was observed when Fe(II) was first bound to the protein prior to H 2 O 2 addition (spectrum D). Thus Fenton chemistry in Dps appears to be completely eliminated only when Fe(II) and H 2 O 2 are added in the proper order. In either case, hydroxyl radical production was significantly diminished by the presence of Dps.
The molar absorptivity corresponding to the ferroxidase center iron in Dps ( max ϭ 297 nm, 29,800 M Ϫ1 cm Ϫ1 per iron) is about 10-fold larger than those of human H-chain ferritin ( max ϭ 305 nm; 2,990 M Ϫ1 cm Ϫ1 ), horse spleen ferritin ( max ϭ 305 nm; 3,540 M Ϫ1 cm Ϫ1 ) (16), 3 and E. coli bacterioferritin ( max ϭ 300 nm; 3,380 M Ϫ1 cm Ϫ1 ) (14). Molar absorptivities of proteins and model complexes with -oxo or -hydroxyl bridges range from 2,000 to 12,000 M Ϫ1 cm Ϫ1 per iron dimer (30,31). The value for Dps is anomalously high and may reflect the presence of more than one oxo/hydroxyl bridging group, which would lend stability to the complex and account for the slow turnover of Fe(II) at the ferroxidase center as evidenced by the stability of its spectrum (Fig. 5). The release of 3 H ϩ upon oxidation of the di-iron(II) center is also suggestive of formation of multiple -oxo or hydroxyl bridges (Equation 3). The fact that the H ϩ / Fe(II) stoichiometry and the rate of iron oxidation are different for the first and second additions of 24 Fe(II)/Dps (Fig. 2) indicates a shift to a mineralization mechanism following saturation of the ferroxidase centers with 24 Fe(III).
The formation of a ferric core in Dps with H 2 O 2 as the Fe(II) oxidant was demonstrated by the spectrophotometric titration (Fig. 7, curve a), the measured oxidation/hydrolysis reaction (Equation 4), the ultracentrifugation data (Fig. 8A), and the similarity between the molar absorptivity of the Dps core ( max ϭ 300 nm, 3,200 M Ϫ1 cm Ϫ1 ) and those of known ferritins (14,16). Although a ferric core was also formed in the presence of dioxygen, a poor oxidant with respect to H 2 O 2 (Fig. 3), the O 2 -driven mineralization most likely does not take place in E. coli. The bacterium contains three kinds of ferritin, namely E. coli ferritin type A, E. coli ferritin type B, and E. coli bacterioferritin, that act as iron storage and detoxification proteins in aerobic environments (36). The role of Dps is distinct. Dps is synthesized under conditions of nutritional and oxidative stress to protect DNA from hydroxyl radical damage by preventing the coordination of Fe(II) to the phosphodiester backbone or to the bases of DNA (1,37). The unusual ferritinlike properties of Dps, which also permit H 2 O 2 consumption, therefore have to be coupled to the physical association with DNA.
Of particular interest is the observation of ferrous core formation (Figs. 7, curve b, and 8, B) as previously reported for horse spleen ferritin (32)(33)(34)(35). The hydrolysis chemistry suggests an FeOH ϩ species with associated chloride to maintain electroneutrality (Equation 5) to give a core composition FeO-HCl with the amount of associated water unspecified. This composition was the same as that determined analytically for horse spleen ferritin (32,33). Aqua Fe(II) itself normally does not undergo hydrolysis below pH 9 (29). However, the high negative electrostatic potential within the interior of the protein shell (7) perhaps facilitates Fe(II) hydrolysis and mineralization at pH 7. Because E. coli bacteria live in both anaerobic and aerobic environments, the ability of Dps to sequester Fe(II) may have survival value for the organism. When the ferrous core undergoes oxidation with H 2 O 2 , it does so with a stoichiometry of 2 Fe(II)/H 2 O 2 , again alleviating Fenton chemistry. Either the protein mediates the 2:1 oxidation stoichiometry of the ferrous core or the ferrous core itself undergoes pairwise Fe(II) oxidation through adsorption of H 2 O 2 on the mineral surface ("Results").
The  active metabolism; mode II seems to be first-order with respect to H 2 O 2 between 10 and 100 mM and does not require active metabolism. It is believed especially in mode I killing that DNA damage is mediated in part by Fenton chemistry (1,37) where intracellular generation of HO ⅐ causes oxidative DNA lesions (2). Indeed, upon the exposure of the Fe(II)-DNA complex to H 2 O 2 in vitro, DNA is completely degraded; in contrast, DNA saturated by Dps is fully protected (Fig. 10). From the present data, it is evident that there are at least two kinds of peroxideconsuming mechanism in Dps to cope with peroxide damage. As noted, both ferroxidase site and mineralized core play an important role in consuming H 2 O 2 (Figs. 2, 4, and 6). Furthermore, DpsA, a member of the Dps family, exhibits a weak heme-dependent catalase activity, which is believed to play a role in protecting DNA from peroxide damage (39,40). The E. coli Dps studied here does not contain heme although it does possess a low level catalase activity ("Results"), about a factor of 10 5 less than DpsA.
In closing, under conditions of oxidative assault, E. coli Dps may provide defense against Fe(II) and reactive oxygen species through a dual mechanism, its spatial association with DNA and its capacity to bind Fe(II) and consume H 2 O 2 without producing HO ⅐ radical, thereby ensuring survival of the genetic code of the organism.