The Dps Protein of Agrobacterium tumefaciens Does Not Bind to DNA but Protects It toward Oxidative Cleavage X-RAY CRYSTAL STRUCTURE, IRON BINDING, AND HYDROXYL-RADICAL SCAVENGING PROPERTIES*

Agrobacterium tumefaciens Dps (DNA-binding pro- teins from starved cells), encoded by the dps gene located on the circular chromosome of this plant patho-gen, was cloned, and its structural and functional properties were determined in vitro . In Escherichia coli Dps, the family prototype, the DNA binding properties are thought to be associated with the presence of the lysine-containing N-terminal tail that extends from the protein surface into the solvent. The x-ray crystal structure of A. tumefaciens Dps shows that the positively charged N-terminal tail, which is 11 amino acids shorter than in the E. coli protein, is blocked onto the protein surface. This feature accounts for the lack of interaction with DNA. The intersubunit ferroxidase center charac-teristic of Dps proteins is conserved and confers to the A. tumefaciens protein a ferritin-like activity that man-ifests itself in the capacity to oxidize and incorporate iron in the internal cavity and to release it after reduc-tion. In turn, sequestration of Fe(II) correlates with the capacity of A. tumefaciens Dps to reduce the production of hydroxyl radicals from H 2 O 2 through Fenton chemis- try. These data demonstrate conclusively that DNA protection from oxidative damage in vitro does not require formation

into plant cells, where the oncogene-containing T-DNA becomes integrated into the chromosomes. This transfer system has played important roles in plant biotechnology as well as in plant molecular biology (1,2). Recently, A. tumefaciens was shown to contain one circular and one linear chromosome in addition to the tumor-inducing plasmid (3), and the complete genome sequence was determined (4). Physical maps of the linear and circular chromosomal DNAs and a linkage map of the linear chromosomal DNA in A. tumefaciens MAFF301001 were constructed for further genome analysis (3). Chromosomally coded virulence genes were mapped exclusively on the circular chromosome.
Plant cells challenged with A. tumefaciens or other pathogenic microorganisms react by means of an oxidative burst that has H 2 O 2 as a major component and aims at restricting the bacterial growth conditions. In turn, A. tumefaciens uses H 2 O 2 as an intracellular and intercellular inducer for the catalase A (katA) gene expression. which serves to minimize the oxidative response of the infected plant cells by maintaining appropriate levels of catalase activity and H 2 O 2 (5,6).
The protection strategy against oxidative and nutritional stress has been studied extensively in Escherichia coli (7). A key component is a nonspecific DNA-binding protein named Dps (DNA-binding proteins from starved cells) that protects DNA from cleavage caused by reactive oxygen species such as the hydroxyl radicals produced during oxidation of Fe(II) by H 2 O 2 (8). When first described, the protective action of Dps on DNA was attributed to physical association between the two macromolecules since complex formation essentially abolishes oxidative DNA cleavage both in vitro and in vivo (8,9). Most recently however, E. coli Dps was shown to possess iron and H 2 O 2 detoxification capacity, and this novel property was proposed to act in concert with physical association to DNA to achieve its protection against oxidative assault (10).
The striking similarity of E. coli Dps with Listeria innocua ferritin, and in particular the presence of the unusual ferroxidase site first described in the latter protein, provide the structural basis for the ability to nullify the toxic combination of Fe(II) and H 2 O 2 (12,13). E. coli Dps and Listeria ferritin, like all members of the Dps family, are homo-polymers formed by 12 four-helix bundle subunits that assemble with 23 symmetry into a hollow shell whose cavity is about 90 Å in diameter (11,14,15). Relative to Listeria ferritin, many Dps sequences display an N-terminal extension of variable length that contains two or three positively charged lysine residues. In the E. coli protein, the family prototype, this N-terminal region extends from the four-helix bundle core into the solvent and is thought to play an important role in the stabilization of the complex with DNA (11). In keeping with this idea, Listeria ferritin (15), Dlp-1 Dlp-2 from Bacillus anthracis (16), HP-NAP from Heli-* This work was supported by Grant "Agenzia 2000" from the Consiglio Nazionale delle Ricerche (to A. I.), and Grants "Biologia Strutturale e Dinamica di Proteine Redox" and "Centro di Eccellenza Biologia e Medicina Molecolare" from the Ministero Istruzione Università e Ricerca, and a grant from the Agenzia Spaziale Italiana (to E. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The cobacter pylori (17), which lack the N-terminal extension, do not bind to DNA. In contrast to the variability of the N terminus, the ferroxidase site is highly conserved. In fact, the iron ligands are part of the so-called "DNA-binding signature." The Dps ferroxidase site is unusual in that it is not embedded in a four-helix bundle as is generally the case but is shared by 2-fold symmetry-related subunits providing the iron ligands. As mentioned above, this site was first described in Listeria ferritin, where it is involved in the iron oxidation/incorporation process (14), which in turn is related to the iron and H 2 O 2 detoxification properties of the protein (12). It is important to stress that the residues which coordinate the metal in the Listeria ferritin ferroxidase site have the same function in all the members of Dps family whose x-ray crystal structures have been solved so far, namely E. coli Dps (11), Dlp-1 Dlp-2 from B. anthracis (16), and HP-NAP from H. pylori (17). Accordingly, these proteins are endowed with the ability to oxidize and incorporate Fe(II).
In this report, we describe the cloning, x-ray crystal structure, and functional properties of A. tumefaciens Dps, encoded by the dps gene located on the circular chromosome of the bacterium. Two findings provide focus on the mechanism of DNA protection by Dps proteins. First, A. tumefaciens Dps does not bind DNA in vitro since the lysine-containing N-terminal extension, which is 11 amino acids shorter than in the E. coli protein, is not disordered but interacts with the four-helix bundle core. Second, A. tumefaciens Dps affords significant DNA protection against oxidative cleavage due to its capacity to bind iron and to inhibit hydroxyl radical formation. It follows that in vitro DNA protection against oxidative assault does not require formation of a Dps-DNA complex but rests on the iron binding/oxidation activity of the Dps proteins. This is the likely reason for conservation of the ferroxidase site.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Media-A. tumefaciens strain GV3101 was grown on LB liquid medium (10 g/liters tryptone, 5 g/liters yeast extract, 5 g/liters NaCl), and LB plates containing 25 g/ml gentamicin (Sigma), and 50 /ml rifampicin at 28°C. E. coli strain BL21DE3 was grown at 37°C on LB liquid medium and LB plates containing 50 g/ml ampicillin.
Cloning of the A. tumefaciens dps Gene into the Expression Vector pet-11a-The dps gene was amplified by PCR from the A. tumefaciens GV3101 genome using primers AGR1 (5Ј-CAGGAGACACATATGAAG-ACGCACAAG-3Ј) and AGR2 (5Ј-GCTGTGTGGATCCTGTTGGAAATT-C-3Ј). The restriction sites for NdeI and BamHI are underlined. Isolation of the A. tumefaciens genomic DNA was performed using the DNeasy tissue kit (Qiagen, Chatsworth, CA). The amplified fragment (505 bp) was digested with NdeI and BamHI, purified by the QIAquick PCR purification kit (Qiagen), and cloned subsequently into the expression vector pet-11a (Novagen) digested with NdeI and BamHI. This plasmid was introduced into E. coli BL21DE3 and sequenced by E. coli strain BL21DE3 dideoxy sequencing to confirm the presence of the correct gene.
Expression and Purification of Dps-E. coli BL21DE3 cells harboring the recombinant plasmid were grown in 1 liter of liquid LB containing ampicillin (50 g/ml) at 37°C to an optical density of 0.6 at 600 nm. After addition of 0.5 mM isopropyl-␤-D-thiogalactopyranoside to induce the transcription of the dps gene, the culture was incubated further for 3-4 h.
Cells were harvested (15,000 ϫ g for 20 min) and suspended in 10 ml of buffer A (50 mM Tris-HCl, pH 7.5, 0.5 mM dithiothreitol, 1 mM EDTA, 500 mM NaCl) and disrupted by sonication. The lysate was centrifuged at 15,000 ϫ g for 45 min. The supernatant was precipitated using two ammonium sulfate cuts at 30 and 60% saturation (w/v). At 60%, Dps remained in solution after centrifugation (15,000 ϫ g for 45 min); the supernatant was dialyzed overnight against 20 mM Tris-HCl, pH 7.5, 50 mM NaCl and then loaded onto a DEAE cellulose column (DE52) equilibrated with the same buffer. Dps was eluted with 250 mM NaCl, pooled, and stored at Ϫ75°C after controlling the purity of the preparation by Coomassie Blue staining of 15% SDS-PAGE gels.
Protein Crystallization-Crystallization was achieved at 293 K by the hanging drop vapor diffusion technique. A 2-l volume of the pro- Data Collection and Processing-Data were collected as 1.0 oscillation frames using the MAR CCD detector on the x-ray beamline at ELETTRA, Basovizza (Trieste, Italy) at a wavelength of 1.2 Å. Data were collected at 100 K using 25% glycerol as cryoprotectant. The data analysis, performed with DENZO (18), indicates that the crystals are orthorhombic (P2 1 2 1 2) with unit cell dimensions of a ϭ 106, b ϭ 90.349, c ϭ 105.6 Å. The data were scaled using SCALEPACK (18) and had an R sym ϭ 5.3% and an 2 ϭ 0.97. The crystal contains six monomers/ asymmetric unit, corresponding to half of the assembled molecule, with a V M ϭ 2.15 Å 3 /Da and a solvent content of about 45%.
Structure Solution and Refinement-The structure was solved by molecular replacement using as the search probe a polyalanine-truncated model of one-half of the E. coli Dps dodecamer (Protein Data Bank entry 1DPS). The rotational and translational searches, performed with the program AmoRe (19) in the resolution range of 10 -3.0 Å, produced a clear solution corresponding to a correlation coefficient between F calc and F obs of 62.0 and to an R fact of 46.2%. Refinement of the atomic coordinates and displacement parameters was carried out by the maximum likelihood method with the program REFMAC (20). The refinement statistics are presented in Table I. Model building was performed using the program package XTALVIEW (21). Water molecules were added to the model manually. The final model (a hexamer) includes 972 residues (162 residues/monomer), 740 water molecules, 6 Fe(III), 2 Tris (hydroxymethyl)-aminomethane molecules, and an ethylene glycol molecule. The final R crys at 1.45-Å resolution is 18%. The quality of the model was assessed by the program PROCHECK (22). The most favored regions of the Ramachandran plot contain 96.0% of non-glycine residues. The atomic coordinates and the structure factors have been deposited in the Protein Data Bank with the accession number 1O9R.
Iron Incorporation-Solutions of ferrous ammonium sulfate were prepared before each experiment by dissolving weighed amounts of the salt in AnalaR water deoxygenated by bubbling with nitrogen. The iron incorporation kinetics was monitored spectrophotometrically at 310 nm and 25°C on a Hewlett-Packard diode array spectrophotometer after the addition of five increments of 50 Fe(II)/dodecamer to Dps solutions (1 M) in 50 mM MOPS, 1 200 mM NaCl, pH 7.8, equilibrated in air. The solutions were kept under continuous stirring during the course of the reaction. As a control, the rate of Fe(II) autoxidation was measured in parallel. After the incorporation of iron, the samples were subjected to gel electrophoresis under non-denaturing conditions in 7.5% (w/v) polyacrylamide gels. The gels were stained for iron with potassium ferrocyanide and for protein with Coomassie Blue.
Kinetics of Iron Release-The kinetics of iron release was measured on Dps loaded with 273 Ϯ 40 iron atoms/dodecamer as assessed by means of the Fe(II) 2,2Ј-bipyridyl method (23). Sodium dithionite was used as reducing agent. Stock solutions of sodium dithionite (0.  A. tumefaciens Dps, Structure and Detoxification Properties reaction was monitored at 520 nm, the wavelength corresponding to the absorption maximum of the Fe(II)-bipyridyl complex using a Hewlett-Packard diode array spectrophotometer equipped with a thermostatted cell maintained at 25°C. The data were fitted using the program MATLAB (The Math Works Inc., Natick, MA). Measurement of Catalase Activity-Oxygen production was measured polarographically with a Clark type O 2 electrode (Ys Model 5300 YSI) at 25°C both in the absence and in the presence of A. tumefaciens or E. coli Dps (1 M) in the apo or in the 24 Fe(III)-containing form. The observed traces were analyzed with the software MATLAB.
Gel Retardation Assay-Supercoiled pet-11a DNA (20 nM) was incubated with 5 M A. tumefaciens or E. coli Dps in 20 mM MOPS (pH 7.5) for 30 min at room temperature in the presence or absence of 200 mM NaCl or 100 mM MgCl 2 . The complexes were resolved on 1% agarose gel run in TAE buffer (0.04 M Tris acetate and 0.002 M EDTA). The gels were stained with ethidium bromide and imaged by ImageMaster VDS (Amersham Biosciences).
DNA Protection Assay-DNA protection from oxidative damage was assessed in vitro using pet-11a plasmid DNA (5600 bp, 20 nM), purified by a Qiaprep spin plasmid miniprep kit (Qiagen). The total reaction volume was 15 l in 20 mM Tris-HCl, pH 7.3, 0.2 M NaCl. Plasmid DNA was allowed to interact with A. tumefaciens or E. coli Dps (5 M) for 10 min prior to the introduction of FeSO 4 (50 M) and H 2 O 2 (10 mM). The reaction mixtures were incubated for 5 min at room temperature. The reactions were stopped by incubation with 2% SDS at 85°C for 5 min. Plasmid DNA was resolved by electrophoresis on 1% agarose gel in TAE buffer. The gel was stained with ethidium bromide and imaged by ImageMaster VDS.
Protection against Hydroxyl Radical Formation-Iron-dependent hydroxyl radical formation was assayed as described by Halliwell and Gutteridge (24) and Yamamoto et al. (25). Fe(II) salts in solution generate hydroxyl radicals that degrade deoxyribose to a malondialdehydelike compound that forms a chromogen with thiobarbituric acid. Fluorescence of the chromogen was used as the assay for hydroxyl radical formation. The reaction mixture contained the following reagents in a total volume of 1.2 ml at the final concentration stated: potassium phosphate buffer (10 mM, pH 7.4), NaCl (63 mM), deoxyribose (1 mM). Variable amounts of E. coli or A. tumefaciens Dps were added to the basal reaction mixture. The reaction was initiated by adding ferrous ammonium sulfate (21 M), and the mixture was incubated at 37°C for 15 min. Thereafter, thiobarbituric acid (1 ml at 1% w/v) plus trichloroacetic acid (1 ml at 2.8% w/v) were added. The whole was heated at 100°C for 10 min, cooled, and the chromogen was measured by fluorescence emission at 553 nm ( ex 532 nm) using a Fluoromax fluorimeter (Spex Industry).

RESULTS
Expression and Purification of Dps-Using the amino acid sequence of E. coli Dps (P27430), a search of the available A. tumefaciens genome (www.expasy.org) revealed the presence of one Dps-like gene designated dps that encodes for a 162-aminoacid sequence (Fig. 1A). The pI and M r values, predicted using the ProtParam program, are 5.57 and 17,823 Da, respectively, as compared with 5.78 and 18,564 Da that pertain to E. coli Dps. Oligonucleotides were designed based on the available nucleotide sequence. The dps gene was amplified by PCR, cloned into the expression vector pet-11a, and transformed into E. coli BL21DE3. The protein was purified to homogeneity using the two-step purification protocol described under "Experimental Procedures." Monomer Fold and Dodecameric Assembly-The A. tumefaciens Dps monomer folds in the four-helix bundle typical of Dps proteins and ferritins. The four helices, A to D, are packed by hydrophobic interactions; helices B and C are connected by a long loop containing a short ␣-helix (BC). Twelve identical subunits assemble to form a hollow protein shell with 23 symmetry and an external diameter of 90 Å (Fig. 1B). Fig. 1, B and C, shows a comparison of the A. tumefaciens Dps monomer with those of E. coli Dps and L. innocua ferritin, respectively. A similar comparison with the monomers of Dlp-1, Dlp-2 from B. anthracis, and HP-NAP from H. pylori is given in Fig. 1D. The superposition of these structures yields root mean square deviations of the C␣ positions ranging from 1.2 to 1.5 Å for A. tumefaciens Dps relative to L. innocua ferritin, Dlp-1, Dlp-2, and HP-NAP, whereas the superposition between A. tumefaciens Dps and E. coli Dps results in a lower root mean square deviation, namely 0.822 Å (Fig. 1). Fig. 1 shows that the only significant differences are in the N-and C-terminal regions. Thus, the first 13-16 residues of A. tumefaciens Dps, which form an extension containing 3 Lys residues, are missing in the Dlp-1, Dlp-2, HP-NAP, and L. innocua monomers. In E. coli Dps, this extension is 11 residues longer than in the Agrobacterium protein; it is flexible (the first 9 amino acids are not visible in the x-ray crystal structure) and believed to protrude from the dodecamer and to be available for DNA binding. Surprisingly, in the A. tumefaciens Dps structure, the N-terminal extension is fully visible in the x-ray crystal structure as it interacts with the C terminus and with the C helix of the 3-fold symmetry-related subunit (Fig. 2). Hydrogen-bonding interactions are established by Gly 77 through the backbone carbonyl group to Thr 3 and by Lys 5 through the backbone carbonyl to NE2 of Gln 159 , whereas Lys 7 is hydrogen-bonded through the carbonyl oxygen to ND2 of Asn 119 and through the backbone nitrogen to OE1 of Glu 155 . In addition, OD1 of Asp 9 forms a salt bridge with NE of Arg 122 .
In the structural alignment presented in Fig. 1A, the residues forming the so-called DNA binding signature are evidenced (in red). These residues comprise, as mentioned above, the iron ligands in the intersubunit ferroxidase site, namely His 40 , His 52 , Asp 56 , Glu 71 , Asp 67 , and other residues that are likely to determine the correct fold of the iron binding site such as Trp 41 , Gly 45 , placed on the AB loop near His 40 , and Gly 77 on the first part of the CD loop near Glu 71 . It follows that DNA binding signature is a misnomer.
A last set of residues, such as Asp 133 and Asp 138 (Fig. 1A,  green), are at the outer rim of the hydrophilic channels along the 3-fold axes and provide the major contribution to the strongly negative electrostatic surface potential that surrounds the channels themselves to permit iron entry into the protein.
Although the external electrostatic surface potential is strongly negative, positive patches like that formed by Arg 112 , Lys 92 , and Lys 94 can be distinguished. They may have a role in the interaction with as yet unidentified molecular targets. As in all Dps proteins, the inner surface is strongly negative (data not shown) and is involved in the formation of the iron core.
Structural Features of the Ferroxidase Center-The F obs Ϫ F calc difference Fourier map, calculated before the introduction of the water molecules, shows a strong peak for each monomer with a value of 8 placed at the interface between the 2-fold symmetry-related subunits. The peaks have been assigned to an iron ion because the peak intensity is about three times stronger than that of water molecules and because x-ray fluorescence spectra measured at ELETTRA indicated that iron atoms are present in the crystals (data not shown). If one assumes an occupancy of 0.5, the iron ions refinement gives a reasonable thermal parameter of 27 Å 2 . The iron atom has a tetrahedral coordination sphere where two ligands are Asp 67 and Glu 71 of one subunit, which binds iron with side chain oxygen atoms. A third ligand is His 40 of the 2-fold symmetryrelated subunit, which binds iron with the side chain nitrogen atom (NE1), and the fourth ligand is a water molecule placed at about 3 Å from the iron ion and hydrogen-bonded to His 52 . This iron binding site therefore closely resembles the ferroxidase site identified in the L. innocua dodecameric ferritin (12) that is shared by all members of the Dps family.
Since the A. tumefaciens Dps iron binding site is not fully occupied by the metal, the x-ray structure furnishes the unique opportunity to visualize local conformational changes induced by iron binding. Thus, two distinct conformations for the Asp 67 side chain are apparent in the 2F obs Ϫ F calc electron density map (Fig. 3). When the site is empty, Asp 67 assumes a rotameric conformation that allows formation of a salt bridge with Lys 37 rather than coordination of the iron.
Iron Incorporation and Release-In accordance with the presence of the ferroxidase site, A. tumefaciens Dps is able to oxidize and incorporate iron. This is indicated by electrophoretic analysis of the products obtained upon incubation with ferrous iron in the presence of molecular oxygen in 50 mM MOPS, 200 mM NaCl, pH 7.8 (data not shown). A progress curve of iron uptake by A. tumefaciens Dps is shown in Fig. 4A; the half-time observed upon addition of 100 Fe/molecule corresponds to 27.7 s. Similar values were obtained with E. coli Dps under the same experimental conditions (data not shown).
As for Listeria ferritin (12,14) and for all Dps proteins tested (10, 15, 25, 26), up to 400 -500 Fe(II) can be loaded into the protein cavity. The iron-loaded protein was used to follow the kinetics of iron reduction, which results in the release of the metal into the solvent. For comparative purposes, dithionite was used as the reducing agent, and iron release was followed by measuring formation of the Fe(II) complex with 2-2Ј-bipyridyl at 520 nm. The time course of the reaction is biphasic with kinetic constants of 0.85 ϫ 10 Ϫ2 s Ϫ1 and 7.4 ϫ 10 Ϫ4 s Ϫ1 at a pH of 7.8 (Fig. 4B). The fast phase accounts for about 70% of the total absorbance change. This behavior resembles that of E. coli Dps, which releases iron with a biphasic process with rate constants 4.7 ϫ 10 Ϫ2 and 12 ϫ 10 Ϫ4 s Ϫ1 (13). A 2-fold variation of the bipyridyl concentration does not change the observed rates, indicating that formation of the Fe(bipyridyl) 2ϩ 3 complex is rapid compared with the iron reduction/release process.
Measurement of Catalase Activity-The catalase activity of A. tumefaciens Dps was tested in 50 mM MOPS, 200 mM NaCl, pH 7.5 and was compared with that of E. coli Dps. The O 2 evolution measured by electrode oximetry after addition of 80 DNA Binding Capacity Assayed in Gel Retardation Assays-The binding of A. tumefaciens Dps to DNA was analyzed in vitro by studying complex formation in agarose electrophoresis experiments (Fig. 5). Dps-DNA complexes were not detected when purified Dps (5 M) was added to 20 nM supercoiled pet-11a DNA in 20 mM MOPS (pH 7.5) in the presence or absence of 200 mM NaCl or 100 mM MgCl 2 . E. coli Dps analyzed in parallel gives rise to a complex with DNA that does not enter the gel in accordance with previous results (8). The same results were obtained in another set of experiments in which the Dps concentration was increased 10-fold while keeping DNA constant (data not shown).
DNA Protection against Hydroxyl Radical Formation-Dps proteins afford protection of DNA from cleavage due to radicals produced in Fe(II)-mediated Fenton reactions, a property attributed originally to physical association between the two macromolecules (8). To establish whether A. tumefaciens Dps is able to protect DNA although complex formation is not apparent in the gel retardation assays just presented, an in vitro DNA damage assay was set up. The combined effect of 50 M Fe(II) and 10 mM H 2 O 2 on the integrity of plasmid pet-11a (5600 bp) was assessed in the presence and absence of Dps. E. coli Dps was used as a control. Under the conditions employed, DNA is fully degraded, and either E. coli or A. tumefaciens Dps yields essentially complete protection (Fig. 6).
In a different type of experiment, iron-dependent hydroxyl radical formation in the presence or absence of A. tumefaciens or E. coli Dps was investigated using the fluorimetric assay (24,25) described under "Experimental Procedures." Addition of E. coli or A. tumefaciens Dps inhibits efficiently hydroxyl radical formation. The extent of inhibition varies from 55 to 70% depending on the Dps to Fe(II) molar ratio (Table II). DISCUSSION The present work on A. tumefaciens Dps brings into a new focus the structure-function relationships in the Dps family. It shows that the possible interaction between a given Dps protein with DNA may not be predictable based on simple sequence analysis of the N terminus. More importantly, it demonstrates that formation of Dps-DNA complex(es) is not a prerequisite for the protection of DNA against oxidative cleavage in vitro.
The current thinking on the interaction between Dps proteins and DNA rests heavily on the behavior of E. coli where it

FIG. 3. Electron density map of the ferroxidase center in
A. tumefaciens Dps. The ferroxidase center residues (Asp 67 , Glu 71 from subunit E, His 40 and His 52 from subunit F) and neighboring residues (Lys 37 from subunit F and Lys 149 from subunit A) are indicated. The iron atom is shown in yellow, and a water molecule is depicted in red. The 2F obs Ϫ F calc electron density map clearly shows two distinct conformations for the Asp 67 side chain. The picture was generated with BOBSCRIPT (35). results in the rapid formation of highly ordered and tightly packed Dps-DNA co-crystals (9). Since such structures have been detected in the cytoplasm of starved bacteria, they have been proposed to be a general mode of DNA protection based on its physical sequestration by Dps (8). The DNA binding ability of E. coli Dps has been associated to the presence of a flexible, disordered N-terminal tail containing three positively charged lysine residues and one arginine residue that can hook the negatively charged DNA backbone (11). Therefore, the positively charged N-terminal extension of A. tumefaciens Dps, albeit 11 residues shorter than in the E. coli protein, was expected to confer DNA binding capacity to this member of the family. However, there is no evidence of complex formation with DNA in band shifts experiments such as those presented in Fig. 5. This unforeseen behavior is explained by the x-ray crystal structure. In A. tumefaciens Dps, the N terminus is not disordered as in the E. coli protein but is blocked onto the four-helix bundle of the 3-fold symmetry-related subunit by several interactions that involve the backbone carbonyl oxygen and peptide nitrogen atoms of Thr 3 , Lys 5 , and Lys 7 and one salt bridge established by Asp 9 with Arg 122 (Fig. 2).
The behavior of A. tumefaciens Dps therefore indicates that an effective interaction of the Dps with DNA requires the positively charged N terminus to be disordered and endowed with flexibility. In line with this contention, the absence of the first N-terminal residues in Listeria ferritin, B. anthracis Dpl-1 and Dpl-2, and HP-NAP correlates beautifully with their inability to form a complex with DNA (15)(16)(17), whereas the short, two-lysine-containing N terminus of B. subtilis MrgA accounts for its binding to DNA (27). In Streptococcus mutans, the Dpr (Dps-like peroxide resistance) protein does not interact with DNA in accordance with the presence of a long N-terminal tail that does not contain positively charged residues apart from two lysines located near the predicted beginning of the A helix (25). Of interest is the formation of a Dps-DNA complex by Mycobacterium smegmatis Dps (28). This protein has a truncated, uncharged N terminus but contains an unusually long C Lane 2, plasmid DNA exposed to 50 M FeSO 4 and then exposed to 10 mM H 2 O 2 . Lane 3, plasmid DNA with 5 M A. tumefaciens Dps exposed to 50 M FeSO 4 and then exposed to 10 mM H 2 O 2 . Lane 4, plasmid DNA with 5 M E. coli Dps exposed to 50 M FeSO 4 and then exposed to 10 mM H 2 O 2 . terminus with 3 lysine and 2 arginine residues that thus is able to substitute the N terminus in the interaction with DNA (Fig.  1A). The behavior of Dps from Synechococcus sp. strain PCC 7942 remains unexpected. This heme-binding Dps-like protein is reported to bind DNA (29). However, the long N terminus does not contain lysine or arginine residues, and there is no C-terminal extension (Fig. 1A).
Despite the inability to bind DNA, A. tumefaciens Dps protects it from oxidative damage due to the ferritin-like activity that has its structural basis in the ferroxidase center. This contains the two canonical metal binding sites (A and B) required for the formation of an oxo-bridged intermediate during oxidation reaction but has an unusual intersubunit location as described first in L. innocua ferritin (15). In A. tumefaciens Dps, the A site is occupied partially by iron, which is coordinated by Asp 67 and Glu 71 from one subunit and His 40 from the 2-fold symmetry-related one, whereas the B site is occupied by a water molecule that is hydrogen-bonded to His 52 . In all known ferroxidase centers, the A and B sites have different affinities for the metal. In the x-ray structures of Dps proteins reported so far, site B contains a water molecule due to the lower affinity for iron, whereas site A is either totally or partially occupied by the metal or contains a water molecule. Thus, in Listeria ferritin (15), HP-NAP from H. pylori, and Dlp-1 and Dlp-2 from B. anthracis site A contains iron (16,17), whereas in E. coli Dps, a water molecule is in place of the metal even in crystals of the iron-loaded protein (13). According to Ilari et al. (15), the A site affinity is influenced by the nature and spatial arrangement of the residues surrounding the iron-coordinating ligands. Thus, the lower affinity of the A site in E. coli Dps relative to Listeria ferritin and the other Dps proteins just mentioned can be ascribed to the replacement of a histidine residue (His 28 in Listeria ferritin) with a lysine residue (Lys 48 ) that engages one of the iron-coordinating ligands (Glu 62 ) in a salt bridge. The differences in the residues surrounding the ferroxidase site may be reflected also in the different catalase activity observed in E. coli and A. tumefaciens Dps.
The partial occupancy of the A site by iron in A. tumefaciens Dps offers the unique opportunity to visualize the conformational changes induced by iron binding in the ferroxidase center itself. The electron density map shows that Asp 67 can adopt two configurations: in one, it coordinates iron, and the site architecture resembles that of L. innocua ferritin (15), and in the other, it forms a salt bridge with Lys 37 , and the site architecture resembles that of E. coli Dps (Fig. 3).
The similarity of the ferroxidase centers in all members of the Dps family has its functional manifestation in the similarity of the iron oxidation/incorporation and release processes. Thus, all Dps proteins tested are able to oxidize and incorporate up to 400 -500 iron ions in their internal cavity and to release them after reduction. Further, the kinetic features of both reactions are close to each other. The ferritin-like activity of Dps proteins has yet another facet of major biological relevance, namely the capacity to counteract the toxic action of H 2 O 2 , a common component of the defense systems put into action when organisms, including plants, are challenged by pathogenic bacteria. DNA damage in particular is due to the concerted action of H 2 O 2 and Fe(II), which generates hydroxyl radicals through Fenton chemistry. In turn, these radicals cause DNA strand breaks by oxidizing the sugar and base moieties. As in the case of E. coli Dps (10) and S. mutans Dpr (25), A. tumefaciens Dps reduces significantly the production of hydroxyl radicals generated by Fenton chemistry in vitro (Table II) and thus preserves the integrity of plasmid pet-11a (Fig.  6). Interestingly, the protective effect against oxidative damage does not appear to depend on the formation of a Dps-DNA complex as it is of very similar magnitude in A. tumefaciens and E. coli Dps, although the former protein does not interact with DNA. It follows that in vitro DNA protection by Dps proteins does not require the physical sequestration of DNA as hitherto believed but can be attributed solely to their Fe(II) scavenging capacity.
A number of experiments indicates that also in vivo, the Fe(II) capacity of Dps proteins is the major player in affording DNA protection against oxidative stress. Thus, in Campylobacter jejuni, where Dps is expressed constitutively, a Dpsdeficient mutant was found to be more vulnerable to H 2 O 2 than the parental strain, and the iron chelator Desferal was observed to restore H 2 O 2 resistance (30). It should be stressed that C. jejuni Dps is not likely to interact with DNA since it lacks the two known interacting modules, namely the N-and Cterminal extension. Another example is furnished by S. mutans where Dpr protects cells from peroxides and confers aerotolerance (31) in accordance with the inhibition of iron-dependent hydroxyl radical formation in vitro (25). As in the case of C. jejuni Dps, S. mutans Dpr does not bind DNA (25). On the basis of these data, A. tumefaciens Dps is expected to play a significant role in vivo when the bacterium has to face the production of H 2 O 2 by the infected plant. Dps can act in concert with catalase A in the elimination of the toxic effects of H 2 O 2 by removing Fe(II) from the intracellular medium and thereby inhibiting the generation of the DNA-damaging hydroxyl radicals. It remains to be established whether the DNA binding capacity of Dps proteins such as E. coli Dps (8,9) may serve an additional, specific regulatory role in the bacterium, as suggested by the highly pleiotropic phenotype of mutant cells lacking Dps with respect to the proteins synthesized during prolonged starvation (32).
In conclusion, the in vitro characterization of the structural and functional properties of A. tumefaciens Dps provides a deeper understanding of the DNA-protective action exerted by the members of the Dps family. It shows unequivocally that the Fe(II) binding ability of these proteins suffices to prevent the adverse effects of Fenton chemistry and thereby provides a molecular explanation for the conservation of their characteristic intersubunit ferroxidase site.