Structural Changes of Escherichia coli Ferric Uptake Regulator during Metal-dependent Dimerization and Activation Explored by NMR and X-ray Crystallography*

Ferric uptake regulator (Fur) is a global bacterial regulator that uses iron as a cofactor to bind to specific DNA sequences. Escherichia coli Fur is usually isolated as a homodimer with two metal sites per subunit. Metal binding to the iron site induces protein activation; however the exact role of the structural zinc site is still unknown. Structural studies of three different forms of the Escherichia coli Fur protein (nonactivated dimer, monomer, and truncated Fur-(1-82)) were performed. Dimerization of the oxidized monomer was followed by NMR in the presence of a reductant (dithiothreitol) and Zn(II). Reduction of the disulfide bridges causes only local structure variations, whereas zinc addition to reduced Fur induces protein dimerization. This demonstrates for the first time the essential role of zinc in the stabilization of the quaternary structure. The secondary structures of the mono- and dimeric forms are almost conserved in the N-terminal DNA-binding domain, except for the first helix, which is not present in the nonactivated dimer. In contrast, the C-terminal dimerization domain is well structured in the dimer but appears flexible in the monomer. This is also confirmed by heteronuclear Overhauser effect data. The crystal structure at 1.8Å resolution of a truncated protein (Fur-(1-82)) is described and found to be identical to the N-terminal domain in the monomeric and in the metal-activated state. Altogether, these data allow us to propose an activation mechanism for E. coli Fur involving the folding/unfolding of the N-terminal helix.

Iron is an essential element for almost all living organisms. However, the Fe(III) form is highly insoluble, and an excess of Fe(II) is toxic for the cell because of the formation of highly reactive radicals via the Fenton reaction. This is why the intracellular iron concentration has to be tightly regulated. In Gramnegative bacteria, the ferric uptake regulator (Fur) 3 protein plays the key role in regulating iron homeostasis. Escherichia coli Fur is a global transcriptional regulator that controls the expression of more than 90 genes mainly implicated in iron uptake but also in other fundamental processes, such as the regulation of oxidative stress, acid tolerance, and bacterial virulence determinants (1).
E. coli Fur is mainly described as a dimeric protein containing two metal-binding sites per subunit. Binding of a divalent cation (ferrous iron in vivo) to the regulatory site enables the protein to bind to a specific DNA sequence in the promoter region of iron-regulated genes, the Fur box (2), thus repressing gene transcription. In the regulatory site, the ferrous iron can be replaced by other divalent metal ions, such as Mn(II), Co(II), Zn(II), and Ni(II) (2)(3)(4). Previous spectroscopic data (EPR, Mössbauer, and x-ray absorption) indicated that this site forms an axially distorted octahedral environment (5,6). A second metal site was identified in the 1990s in the E. coli Fur protein, which binds Zn(II) with a very high affinity (7,8). Extended x-ray absorption fine structure studies showed that this zinc ion is coordinated by two cysteines together with one carboxylate and one histidine or two carboxylates (7). The two cysteines, Cys 92 and Cys 95 , have been identified as ligands of the zinc atom using chemical modification and mass spectrometry (9). This confirmed the importance of these two cysteines that have been shown to be essential to the E. coli Fur activity by site-directed mutagenesis (10). These cysteines are conserved in a large number of Fur proteins, but only one is conserved in Pseudomonas aeruginosa Fur. Although the zinc site probably plays a struc-* 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 atomic coordinates and structure factors ( tural role, to date no clear function of this zinc site was elucidated. The only three-dimensional structure of a Fur protein described to date is the metal-activated form of the P. aeruginosa Fur (11). The protein was crystallized in the presence of Zn(II) ions. Four metal ions were observed per monomer, two of them were bound to conserved ligands and were therefore proposed to occupy the two characterized relevant sites. From the metal-site geometry in the presence of zinc and after its substitution by iron in the regulatory site, the regulatory and the structural site could be identified (11). In the regulatory site, the zinc ion is hexacoordinated in an octahedral geometry, with a water molecule occupying one of these positions. The structural zinc-binding site contains a zinc ion in a tetrahedral geometry bound to two glutamic acids (Glu 80 and Glu 100 ) and two histidines (His 32 and His 89 ) and is therefore different from the one in E. coli Fur. It is noteworthy that the single cysteine residue of P. aeruginosa Fur was shown to be dispensable for the in vivo activity of P. aeruginosa Fur by site-directed mutagenesis (12).
Because most Fur proteins contain the conserved cysteines found to be bound to the structural zinc in the E. coli protein, it seems essential to study the structural properties of the biochemically well characterized E. coli Fur. In addition, no structural information is available on the nonactivated form of Fur.
Here we describe for the first time the structural properties of the nonactivated dimeric E. coli Fur (named Zn S FurD for dimeric Fur protein containing zinc in the structural metal site) by using NMR spectroscopy. In addition, we also studied the so far uncharacterized monomeric form (13) and its conversion to dimer. Finally, a truncated protein comprising the first 82 amino acids, corresponding to the DNA binding domain, has been characterized by NMR and x-ray crystallography. Comparison of the different Fur forms yields essential information on the structural reorganization of this important E. coli global regulator during dimerization and activation processes.

MATERIALS AND METHODS
Protein Sample Preparation-Protein purification was performed at 4°C and centrifugation at 6°C. Native E. coli Fur was purified to homogeneity according to the published procedure (14). A higher yield in monomeric Fur can be obtained by increasing the EDTA concentration in the extraction buffer to 100 mM. The different oligomeric states of Fur were separated by gel filtration (Superdex 75 16/60 HR; Amersham Biosciences) in 0.1 M Tris/HCl, 0.1 M KCl, pH 8 (4°C). 10% v/v glycerol was added before protein concentration and storage. Dimer/monomer fractions were pooled separately and concentrated to a final volume of 1 ml by using an ultrafiltration support (Ultrafree 5K, Millipore). Protein concentrations, consistently expressed in subunit concentration unless indicated otherwise, were determined spectrophotometrically by using an absorption coefficient of 0.4 mg Ϫ1 ml Ϫ1 cm Ϫ1 at 275 nm for one monomer of native Fur (15). The Fur-(1-82) mutant was generated by using the QuikChange site-directed mutagenesis kit from Stratagene with the two synthetic oligonucleotide primers 5Ј-CCGTATTTGAACTGACATAGCAA-CATCACCACGATC-3Ј and 5Ј-GATCGTGGTGATGTT-GCTATGTCAGTTCAAATACGG-3Ј and the methylated plasmid pET30c-FUR containing the wild-type protein (14). The constructs were confirmed by DNA sequencing (Genome Express, France). Freshly transformed BL21(DE3) competent cells were plated on LB (0.1% agar) containing 50 g/ml kanamycin and were grown at 37°C. Cells were grown at 37°C from an overnight culture in 1 liter of LB containing 50 g/ml kanamycin. Fur-(1-82) expression was induced at A 600 nm ϭ 0.7-0.8 by isopropyl ␤-D-thiogalactoside at a final concentration of 0.5 mM for 3 h. Cells were harvested at 5000 ϫ g for 5 min, and the pellet was suspended in 30 ml of 0.1 M MOPS, pH 8, 0.1 M EDTA, 10% w/v sucrose, 10% glycerol (v/v) containing 10 mg/ml trypsin-chymotrypsin inhibitor, 10 g/ml phenylmethylsulfonyl fluoride, and 4 g/ml pepstatin A (all from Sigma). Cells were disrupted by sonication, and the supernatant obtained from centrifugation at 20,000 ϫ g for 20 min was treated with 1% w/v streptomycin sulfate (Sigma). The suspension was centrifuged for 30 min at 35,000 ϫ g. The first step of protein purification consisted of an ammonium sulfate precipitation to 50% of saturation for 3 h followed by centrifugation for 30 min at 25,000 ϫ g, the supernatant was brought to 77% ammonium sulfate saturation, and proteins were precipitated under stirring overnight before centrifugation. The precipitate was redissolved in 5 ml of 0.1 M Tris/HCl, pH 8, containing 10% glycerol (v/v). Homogeneous Fur-(1-82) was obtained after a gel filtration on a Superdex 75 HR 16/60 (Amersham Biosciences) equilibrated with 0.1 M Tris/HCl, pH 8.0, 1 M KCl. Pure protein fractions, as revealed by SDS-PAGE and mass spectrometry, were pooled and concentrated with Ultrafree (5K Millipore). Protein concentration was determined with the BCA assay (Bio-Rad) using a FUR and a bovine serum albumin calibration curve. The yield was 30 mg of pure protein per liter of bacterial culture.
Crystallization Conditions and Structure Determination of the Truncated Fur-(1-82)-Screening of the crystallization conditions was initially performed with a TECAN robot at the Institut of Structural Biology using various commercial screens from Hampton Research and Qiagen. Crystals were obtained at 20°C in 100 mM sodium acetate, pH 4.6, 30% (v/v) PEG 400 containing 100 mM CdCl 2 (classic suite from Qiagen). Crystallization conditions were then optimized, and finally, crystals were grown at 20°C from a 1:1 l mixture of 10 mg/ml protein solution with 100 mM sodium acetate, pH 4.6, 30% (v/v) PEG 200 containing 100 mM CdCl 2 . For data collection, the crystals were transferred into a solution composed of the mother liquor containing 25% glycerol as cryoprotectant. X-ray crystallography was then used to obtain the three-dimensional structure of Fur-(1-82) at high resolution.
The final data set was collected at 1.8 Å resolution on beamline FIP-BM30A at the European Synchrotron Radiation Facility, Grenoble, France (16). Data reduction was performed using the XDS program (17). The space group was P2 1 2 1 2 (a ϭ 38.50 Å, b ϭ 159.27 Å, c ϭ 28.65 Å) with potentially two molecules per asymmetric unit (solvent content of 48%). The structure was solved by molecular replacement performed with the Phaser program (18) using the model of the first 84 amino acids of the P. aeruginosa protein (Protein Data Bank code 1MZB). The first model obtained with Phaser was then introduced in ARP-WARP (19) with the E. coli Fur sequence. ARP_WARP built 139 amino acids over the 164 present in the asymmetric unit, and the remaining amino acids have been incorporated with the coot program (20). Refinements of the different models were carried out with the Refmac5 program of the CCP4 Suite (21) in the 30 -1.8 Å resolution range without any NCS restraints. The atomic coordinates and structure factors for N-terminal domain of Fur have been deposited with the Brookhaven Protein Data Bank under the accession number 2FU4.
For deuterated Fur, 20 ml of LB medium prepared with 50% D 2 O were inoculated with a colony of freshly transformed bacteria, and cells were grown at 37°C to saturation overnight. Bacteria were harvested at 1500 ϫ g, washed twice with 20 ml of cold M9 medium, and resuspended in a final volume of 100 ml of 13 C/ 15 N-labeled M9 medium prepared with Ϸ95% D 2 O. Growth occurred overnight at 37°C. Cells were harvested at 1500 ϫ g and resuspended in 1 liter of 13 C/ 15 N-labeled M9 medium in Ϸ95% D 2 O. The culture was incubated at 37°C. Expression of the Fur protein was induced with 0.5 mM isopropyl ␤-D-thiogalactoside when the A 600 nm reached a value of 0.5. Cells were kept at 37°C for 5 h and were finally harvested and stored frozen at Ϫ80°C until protein purification.
An additional NMR sample of highly deuterated Fur was also prepared using a commercial triple-labeled medium (OD2 medium from Silantes, Germany). After two 10-ml precultures in LB medium prepared with 100% D 2 O, 40 ml of triple-labeled OD2 medium were inoculated with 400 l of cell suspension and incubated overnight at 37°C. Cells were harvested at 1500 ϫ g and resuspended in 920 ml of Silantes OD2 medium. Protein expression was induced by isopropyl ␤-D-thiogalactoside (final concentration of 0.5 mM) at 37°C at an A 600 nm of 0.8 for 5 h before harvesting the cells.
For the NMR experiments, the buffer was exchanged to 0.1 M MOPS/NaOH, 0.5 M KCl, pH 7.5, using NAP10 columns (Amersham Biosciences). 7% D 2 O and either DTT (10 mM, Fur dimer) or EDTA (20 mM, Fur monomer) were added to the sample after the buffer exchange step.
The high KCl concentration employed was necessary to prevent protein oligomerization at high concentrations. 4 NMR sample concentrations varied between 1.5 and 2.5 mM. All NMR samples were prepared in a glove box under argon and sealed in the NMR tubes.
NMR Spectroscopy-All NMR experiments were performed at 298 K on VARIAN INOVA 600 and 800 spectrometers equipped with triple resonance ( 13 C, 1 H, 15 N) probes including shielded z-gradients. The data were processed using Felix 2000 (Accelrys), and the resonance assignment was achieved using the ALPS program (24). HNCO, HN(CA)CO, HN(COCA)CB, and HN(CA)CB pulse sequences were provided by the Varian protein pack. The three-dimensional MQ-HNCOCA (25) and the three-dimensional MQ-COHNCA (26) were performed as described in the literature. The secondary structure elements were identified from secondary 13 CЈ and 13 C-␣ chemical shifts (27). In the case of the dimeric protein, two NOESY experiments were recorded at 600 MHz to deduce the topology of the ␤-strands as follows: a 15  { 1 H}-15 N NOE were measured at 600 MHz using a standard two-dimensional pulse sequence (28). Two spectra with on and off resonance 1 H saturation were recorded in an interleaved manner. The saturation time and the recycle delay were set to 4 and 7 s, respectively. Only well resolved peaks were taken into account for the intensity measurements. Resonance assignments for Zn S FurD and monomeric Fur have been deposited with the BMRB databank under the accession numbers 6947 and 6948, respectively.
Zinc Titration of Monomeric Fur-20 M eq of DTT were added to a solution of 0.8 mM monomeric 15 N-labeled Fur in 0.1 M MOPS/NaOH, 0.5 M KCl, pH 7.5. Up to 1.5 M eq/monomer of ZnSO 4 dissolved in water was then added drop by drop by steps of 0.5 M eq to the sample in an NMR tube before recording 1 H, 15 N-HSQC spectra.

RESULTS
E. coli Fur exists in solution under several oligomeric states (29). 4 Monomeric and dimeric species were purified separately because they are not in equilibrium (13,29). 4 The E. coli Fur monomer contains the four cysteines oxidized in two disulfide bridges Cys 92 -Cys 95 and Cys 132 -Cys 137 (29). 4 Here we describe the structural properties of the dimeric and monomeric forms using NMR and x-ray crystallography, and we investigate the structural changes related to the monomer-dimer conversion.
The 1 H, 15 N-HSQC spectra of the two forms are shown in Fig.  1. The spectrum of the dimeric protein shows ϳ150 peaks, from which it can be concluded that Zn S FurD forms a symmetric dimer, in agreement with the P. aeruginosa crystal structure (11). Even if we consider a variation of the chemical environment for residues at the dimer interface, the spectra obtained for the monomer and the dimer are too different for the two species to expect the same three-dimensional structure. In particular, it can be seen that the spectrum of the monomeric protein is less extended both in the proton and in the nitrogen dimension, indicating a lower degree of structuralization of the monomer compared with the dimer.
However, under our NMR experimental conditions, Fur dimer, prepared in the absence of DTT and without strictly respecting anaerobic conditions, slowly and irreversibly formed some monomeric species. To avoid formation of the monomer, DTT was added to the Fur dimer samples to maintain the cysteines in their reduced form. Furthermore, to avoid formation of oligomeric Fur in the NMR dimer samples, they were prepared in 0.5 M KCl, pH 7.5. The monomeric sample appears stable with time and only shows some proteolytic degradation after several days at room temperature.
NMR Characterization of Nonactivated Fur Dimer (Zn S FurD)-Using standard three-dimensional NMR experiments, 91% of the backbone resonances could rapidly be assigned. No resonance assignments could be obtained for the five N-terminal residues as well as Val 67 , Gly 74 , His 86 , His 87 , Asp 104 , Ser 105 , and Lys 116 residues. The secondary chemical shifts derived from the difference between the experimental chemical shift and the one corresponding to a random coil conformation for a specified amino acid indicate the presence of four ␣-helices (residues 18 -24, 35-44, 51-62, and 106 -114) and five ␤-strands (residues 66 -71, 77-82, 88 -91, 97-102, and 119 -131), which are detailed in Fig. 2A. The first three helices and the subsequent two ␤-strands belong to the winged helixturn-helix motif required for DNA interaction (30). However, the crystal structure of P. aeruginosa Fur contains an additional N-terminal helix (11) that we do not observe for the Zn S FurD protein in solution.
271 and 196 NOEs between amide groups were identified from the 15 N-edited NOESY-HSQC and the 15 N-HSQC-NOE-15 N-HSQC, respectively. This allowed us to deduce the topology of the ␤-sheets; strands ␤1 and ␤2 as well as ␤3 and ␤4 form two short antiparallel sheets, and the NOEs observed in the last strand can only be explained by the presence of an antiparallel inter-subunit ␤-sheet at the dimer interface. These observations are in good accordance with the crystal structure of the Zn(II)-activated P. aeruginosa Fur protein (11), with the exception of the N-terminal helix preceding the winged helix-turnhelix motif.
Unfortunately, further structural investigations of the protein by NMR were made impossible by the very unfavorable relaxation properties of this protein. The difficulty of the NMR study of Fur was already described in the literature (31)(32)(33). Only about 40% of the side chain resonances could confidently be assigned. Attempts to obtain additional structural information from 13 C-edited NOESY spectra failed. 15 N relaxation measurements revealed that the ratio of transversal and longitudinal relaxation rates was much higher than expected from the molecular mass of 34 kDa (results not shown), and we suppose that under our experimental conditions there is still some amount of oligomeric protein formed in the sample during data acquisition.
NMR Characterization of Fur Monomer-In case of the monomeric Fur, the backbone resonance assignment could be achieved easily for the N-terminal domain. Four ␣-helices (residues 5-9, 17-24, 35-44, and 51-63) and two ␤-strands (residues 67-71 and 76 -80) were deduced from the analysis of the secondary chemical shift values (Fig. 2B). The C-terminal domain (residues 81-147) was difficult to assign, and 26 residues of 67 could not be identified. Moreover, 16 peaks were missing in the 1 H, 15 N-HSQC, indicating that the monomer probably undergoes conformational exchange. The secondary chemical shifts do not indicate any well defined secondary structure elements in the C-terminal domain, which is known to be involved in the dimer interface.
Compared with Zn S FurD, the N-terminal domain has globally the same secondary structure except for the first residues, for which the secondary chemical shifts reveal the presence of an additional ␣-helix (residues 5-9), situated N-terminal to the helix-turn-helix motif, analogous to the one seen in the P. aeruginosa protein.
Comparison of Monomeric and Dimeric Species-The comparison of the backbone chemical shifts between monomer and dimer is shown in Fig. 2E. It can be seen that there are no deviations or only very weak deviations in the region corresponding to the helix-turn-helix motif (residues 14 -65), suggesting that the local structure of this part of the protein is not influenced by the dimerization. On the other hand, there are significant differences for the C-terminal half and for the 18 N-terminal residues, regions for which the analysis of the secondary chemical shift has already indicated very different structural behavior (see Fig. 2, A and B). To confirm the structural flexibility of the C-terminal domain in the monomeric species and to obtain further information on the intriguing changes in secondary structure in the N-terminal part between mono-and dimer, { 1 H}-15 N heteronuclear NOE (het-NOE) spectra were recorded for both forms. The het-NOE allows us to obtain information on the reorientation of the amide bond on the pico-to nanosecond time scale. Generally, { 1 H}-15 N het-NOE values above 0.7 indicate that the N-H vector is rigid with respect to the rest of the protein, whereas smaller values are due to increased backbone mobility.
In both the dimer and the monomer, residues previously determined to be situated in secondary structure elements are characterized by { 1 H}-15 N het-NOE values greater than 0.7 (Fig.  2, C and D). However, even in the more structured dimer, loops connecting the secondary structure elements are found to be quite flexible with { 1 H}-15 N het-NOE values below 0.5. In addition, both the N-terminal and the C-terminal residues appear mobile in the dimeric Fur protein. The low { 1 H}-15 N het-NOE values measured for the 15 N-terminal residues confirm the absence of the additional N-terminal helix identified in the monomeric protein. So, overall the dimeric nonactivated Fur does not appear as a structurally well ordered protein, which may explain the difficulties in getting crystals for x-ray analyses.
In the monomeric protein, the N-terminal DNA-binding domain appears to be rigid, including residues 6 -10, thus con-firming the presence of an additional ␣-helix. On the contrary, the C-terminal domain is characterized by { 1 H}-15 N het-NOE values below 0.5, demonstrating its poorly structured nature.
Monomer to Dimer Conversion-In the dimer, the four cysteines are reduced in contrast to the monomer, in which they were found oxidized 4 (data not shown) and involved in disulfide bridges between cysteines Cys 92 and Cys 95 on one hand and Cys 132 and Cys 137 on the other hand. We found that formation of dimer required reduction of the disulfide bonds and incorporation of Zn(II), but reduction of the four cysteines by DTT alone did not promote dimerization. 4 We further investigated the structural changes following reduction of cysteines and addition of Zn(II) ions using NMR. 1 H, 15 N-HSQC spectra of the monomer recorded in the absence or presence of DTT are shown in Fig. 3A. These spectra indicate that no major structural changes occur upon reduction. Only a few 1 H, 15 N-correlation peaks shift, which are mostly located around the Cys 132 and Cys 137 . The addition of Zn(II) to the reduced monomer leads to the appearance of 1 H, 15 N-correlation peaks that belong to the dimer spectrum. Under nonsaturating conditions with zinc, the 1 H, 15 N-HSQC spectra of both forms coexist, indicating that both forms are in slow exchange with respect to the NMR time scale (compare Fig. 3B). The spectrum obtained in the presence of 0.95 M eq of Zn(II) still shows the presence of monomeric Fur, probably because of the excess of DTT also chelating competitively the Zn(II) in solution and thus decreasing the amount of metal available for the protein. It was not possible to completely saturate the zinc site of the Fur protein because at higher Zn(II) concentrations precipitation occurred.
We therefore confirm that the incorporation of Zn(II) is essential for a large structural reorganization related to Fur dimerization and that the dimer formed from the reduced monomer is identical to the native dimeric protein.
Structural Characterization of the Truncated Fur, Fur-(1-82)-As more detailed structural studies of the dimeric E. coli Fur protein were extremely difficult, we decided to further characterize the monomeric form. A truncated protein formed only by the N-terminal DNA-binding domain was produced as the C-terminal domain has been shown to be disordered in monomeric Fur. 15 N-Labeled truncated Fur-(1-82) was obtained and an 1 H, 15 N-HSQC spectrum acquired. Fig. 4 shows the superposition of the 1 H, 15 N-HSQC spectra of monomeric and truncated E. coli Fur. These spectra are almost identical, and we can therefore conclude that the structure of the isolated N-terminal domain is only slightly changed with respect to the corresponding fragment of monomeric Fur.
X-ray crystallography was then used to obtain the three-dimensional structure of Fur-(1-82) at 1.8 Å resolution, which is shown in Fig. 5. The final model contains residues 1-82 from each monomer present in the asymmetric unit as well as 180 water molecules. The asymmetric unit includes two monomers with the interaction surface formed by helix 4 and strand 1 and two Cd(II) cations. This interface, involving two cadmium ions, is certainly not biologically relevant as it contains helix 4, the DNA recognition helix. In addition, this dimeric form is incompatible with the native dimeric protein in which the C-terminal domain provides the inter-subunit interface (11). The structure of one subunit is composed of four ␣-helices (residues 3-10, 16 -25, 35-45, and 51-64) and two ␤-strands forming an antiparallel sheet (66 -72 and 76 -82), a secondary structure composition identical to the one found in the dimeric and zincactivated P. aeruginosa Fur (11). All residues fall within the allowed regions of the Ramachandran plot as defined by the PROCHECK program (34). Statistics for all the data collections and refinement of the different structures are summarized in Table 1. The structure of truncated E. coli Fur-(1-82) can be remarkably well superimposed with the N-terminal domain of the P. aeruginosa Fur protein, with a root mean square deviation of 0.8 Å (calculated for 312 backbone atoms). Local variations of the backbone conformation are observed for two loops, the one connecting helix 2 and helix 3 and the turn between the two ␤-strands (Fig. 5). It should be mentioned that the N-terminal helix of Fur-(1-82) is perfectly superimposed to the one in the P. aeruginosa Fur. In the truncated Fur-(1-82), residues from this helix are involved in hydrophobic contacts with Leu 19 , Leu 22 on helix 2, Phe 61 , Ala 64 on helix 4, as well as Ile 66 and Leu 12 , both situated in loop regions. As also observed for the P. aeruginosa Fur, positively charged residues situated at the end of the N-terminal helix  (Lys 8 , Lys 9 , and Lys 13 in E. coli Fur) contribute to a basic surface patch that can be assumed to be essential for protein-DNA interaction. Comparison of the electrostatic surfaces of E. coli Fur-(1-82) and P. aeruginosa Fur is shown in Fig. S1. Again, these surfaces are very similar, indicating a conserved DNA-interaction mechanism.

DISCUSSION
During the last 2 decades, numerous research groups have contributed interesting and complementary results on the structural and functional aspects of Fur proteins, global transcriptional regulators ubiquitous in Gram-negative bacteria. In 2003, the crystal structure of the Fur protein from P. aeruginosa was published (11), providing the first detailed structural information of a Fur protein on a molecular level. This structure corresponds to the zinc-activated protein, containing zinc in the two metal-binding sites (structural zinc and regulatory iron sites). However, spectroscopic and biochemical data have shown that at least for the structural zinc site, P. aeruginosa and E. coli Fur proteins show differences, which might be related to their specific function in the different organisms. The importance of the cysteines is clearly different in E. coli and in P. aeruginosa Fur because the two of them bound to the zinc in E. coli Fur have been shown to be essential to E. coli Fur activity by site-directed mutagenesis (10). On the contrary to the single cysteine residue of P. aeruginosa Fur that was shown to be dispensable for the in vivo activity of P. aeruginosa Fur also by site-directed mutagenesis (12).
Therefore, structural information on additional Fur proteins is necessary for a better understanding of the structure-function relationship within the Fur family. In addition, no structural data were available for the nonactivated protein. Biochemical data obtained on the E. coli protein have clearly shown that metal binding induces conformational changes (30) 4 related to the activation mechanism. For these reasons, we started to investigate the structural properties of the E. coli Fur protein in its nonactivated dimeric form. Fig. 6 shows the sequence alignment of P. aeruginosa and the E. coli Fur together with the experimentally determined secondary structure elements. By comparing the structural information available from the solution study of dimeric, nonactivated E. coli Fur with the crystal structure of P. aeruginosa Fur, we have shown that the overall architecture is conserved; the size, the localization, and the topology of the secondary structure elements seem to be identical, as far as we can conclude from our NMR data. There is only one remarkable exception: the additional N-terminal helix, which has also been predicted from the E. coli Fur protein sequence. This helix is present in the activated P. aeruginosa protein, but our results show that it is absent in the nonactivated form of Fur from E. coli. We first thought that the presence of this helix in the P. aeruginosa crystal structure might be related to an altered N terminus (addition of two residues resulting from the cleavage of the glutathione S-transferase fusion protein) or to the crystal environment. However, our observation of an analogous helix in the monomeric form of the E. coli Fur protein in solution as well as in the truncated Fur N-terminal domain demonstrates that this region is able to fold as an ␣-helix also in the E. coli protein. We thus became interested in this N-terminal helix and its possible role within the DNA-binding domain.
It has already been suggested by Pohl et al. (11) that the N-terminal helix is required for efficient DNA binding. This is further supported by mutagenesis studies; for example, P. aeruginosa Fur, having Ala 10 mutated to glycine, a much poorer helix former, was found to be unable to bind to the pvdS promoter (35), and in E. coli it has been shown that proteolytic cleavage of the 8 or 9 N-terminal residues resulted in a protein with reduced DNA-binding affinity and specificity (36). By comparing the sequences of different Fur proteins it can be seen that in addition to a highly conserved basic residue in the loop connecting the first two helices (Lys 13 in E. coli), one or two lysine or arginine residues are often found at the end of the N-terminal helix. These lysine side chains are surface-exposed and could thus be involved in DNA recognition and interaction. One can imagine that folding of the N-terminal residues in an  ␣-helical conformation stabilizes these side chains in a position suitable for DNA recognition and interaction.
The absence of this N-terminal helix in the nonactivated protein in solution together with the structural considerations mentioned above suggest that the presence of the N-terminal helix is related to the activation state of the dimeric Fur protein.
Metal binding to the functional iron site would induce conformational changes resulting in the stabilization of the N-terminal residues in an ␣-helical conformation, thus optimizing DNA recognition and interaction. However, looking at the crystal structure of dimeric and activated P. aeruginosa Fur protein, it becomes difficult to imagine how the C-terminal regulation domain could communicate with the distant N-terminal helix in the DNA-binding domain. The results we obtained in this study on the three different forms of Fur may provide a clue for a possible activation mechanism of the Fur protein, involving the N-terminal helix. Comparison of the 1 H, 15 N-HSQC spectra of the three forms indicates that the structure of the winged helix-turn-helix motif (helices 2-4, strands 1 and 2) is independent of the state of the C-terminal domain: folded (Zn S -FurD), unfolded (Fur monomer), or absent (truncated Fur-(1-82)). The only difference between these forms is the absence or presence of the additional N-terminal helix, which we have identified in the two monomeric forms (Fur monomer and truncated Fur-(1-82)). From this observation we can conclude that it is the sole presence of the structured C-terminal domain in the dimeric nonactivated protein, which triggers the unfolding of the N-terminal helix. The unfolded state of the corresponding residues has experimentally been demonstrated by the NMR relaxation measurements (compare Fig. 2D). Furthermore, EDC cross-linking experiments have shown the spatial proximity of the N-terminal amino group of one subunit and a carboxyl group within the C-terminal last 9 residues of the second (37). 4 In addition, experiments using a dimethyl adipimidate cross-linker demonstrated intersubunit contacts between the N-terminal extremity of one subunit and Lys 76 and Lys 97 (37). 4 These contacts are not compatible with the crystal structure of activated P. aeruginosa Fur with an ␣-helical N terminus (see Fig. 7) but could be explained by a more elongated and flexible conformation of the N-terminal residues. Metal binding to the C-terminal activation site abolishes effective intersubunit cross-links involving EDC as well as dimethyl adipimi-date (37), 4 providing additional experimental evidence for the reorganization of the N-terminal residues. We therefore suggest that the activation of E. coli Fur proceeds by metal binding to the C-terminal domain that triggers local conformational changes. This then unlocks the experimentally observed interactions with residues in the unstructured N-terminal part, leading to formation of helix 1 and thus of the DNA recognition site involving the basic residues directly adjacent to this helix. This mechanism may involve a relative reorientation of the two subunits, which could contribute to the decrease of the distance between the N terminus and parts of the other subunit.
To clarify this issue, additional structural studies of the nonactivated form of the Fur protein are necessary. A structural characterization at high resolution of this form would help to gain insight into the global structural reorganization leading to such a drastic change in secondary structure because of the binding of the activating metal ion.
It should be mentioned that the presence of this additional N-terminal helix is very untypical for transcriptional regulators with a winged helix-turn-helix motif. Proteins of the MarR family like MarR (38) or MexR (39) have been shown to possess an extra N-terminal helix, but in these two proteins this helix is involved in dimerization by directly interacting with the C-terminal dimerization domain. To our knowledge, the implication of such an N-terminal helix in DNA recognition and the regulation of DNA binding activity have never been reported in the literature and might be a unique feature of the Fur proteins.
Besides the switch from the nonactivated to the metal-activated form, E. coli Fur also undergoes a metal-induced, redoxcontrolled monomer-dimer transition that depends on Zn(II) binding to the structural metal site. By using NMR spectroscopy, we show that the monomer is characterized by an unstructured C-terminal dimerization domain and that addition of Zn(II) to the reduced monomer leads to the regeneration of Zn S FurD. Therefore, zinc binding induces structural changes that enable the structuralization of the C-terminal domain and dimerization.
In E. coli Fur, the structural zinc is bound by two cysteine residues, as formerly shown by extended x-ray absorption fine structure spectroscopy (7). They were identified and correspond to Cys 92 and Cys 95 (9). One of these two cysteines is missing from the P. aeruginosa Fur protein, and the FIGURE 6. Sequence alignment of Fur from P. aeruginosa and E. coli and secondary structure comparison. Metal-binding ligands are in boldface in the corresponding sequence as follows: underlined boldface for the regulatory iron site and boldface for the structural zinc site. Experimentally determined secondary structure elements are indicated with E for ␤-strands, H for ␣-helices, and -for unstructured. Blanks correspond to unassigned residues in E. coli Fur.
crystal structure reveals that zinc is tetra-coordinated by four conserved amino acids. two histidines and two glutamic acids. In P. aeruginosa Fur, the zinc ligands are located both in the N-and in the C-terminal domain, holding together the two antiparallel ␤-sheets. In a homology-based structural model of E. coli Fur, Cys 92 and Cys 95 are situated in the loop connecting the two strands of the second antiparallel ␤-sheet and are far from the zinc-binding site in P. aeruginosa Fur (Fig. 7). Thus, the zinc-binding site is clearly different from the one in E. coli Fur and may not be strictly conserved among the Fur proteins.
This observation is also confirmed by recent results obtained on the Fur protein from Bradorhizobium japonicum. By using atomic absorption spectroscopy, the authors showed that Br. japonicum Fur does not contain a structural zinc ion, although they were able to correctly demonstrate the presence of zinc in E. coli Fur (40). In addition, they performed mutagenesis studies in which the residues corresponding to either of the two metal-binding sites identified in the crystal structure have been mutated. These mutants still bind DNA with high affinity and repress transcription in an iron-dependent manner (39), which led the authors to suggest that the Fur protein might be a structurally and functionally more diverse protein than previously supposed.
Looking at the sequence alignment of the Fur family (41), it can be realized that most of the Fur proteins in which the two cysteine residues corresponding to Cys 92 and Cys 95 in E. coli are conserved also contain a second pair of cysteines in a C-terminal extension. The first of these two additional cysteines (Cys 132 ) could be mod-eled in the three-dimensional structure for E. coli Fur and is separated by less than 8 Å from Cys 92 and Cys 95 (Fig. 7). From our results in E. coli Fur, we know that this cysteine 132 is not implicated in zinc binding (9) and is not essential for the in vivo activity (10). However, in Fur-like proteins functioning as oxidative stress sensors, like FurS or PerR, the cysteine homologous to Cys 132 in E. coli is found in a CX 2 C motif in contrast to the CX 4 C motif encountered in E. coli Fur. A very recent structural model based on mutagenesis studies was proposed for PerR of Bacillus subtilis (42), in which the zinc was bound by these four cysteines.
All these data provide evidence that the zinc site is not strictly conserved among the Fur and Fur-like proteins and may have evolved in function of the specific requirements in the whole regulation network of a given bacterial strain. For instance, zinc bound to sulfur ligands can be involved in redox regulation of important protein functions (43). The present NMR studies of the monomeric protein confirm the importance of the zinc site for the structural integrity of the dimeric protein. We can hypothesize that this observed redox and metal dependence of the dimerization may have a physiological relevance. By taking together the structural results obtained for the nonactivated dimeric and the monomeric forms of E. coli Fur, new insights are gained into the molecular basis of Fur activation because of metal binding to the regulatory iron site. We propose that Fur activation implicates folding and unfolding of the untypical N-terminal helix observed in the metal-activated and in the monomeric forms of the Fur protein.
Further experimental data are required to verify these hypotheses and to obtain a more profound comprehension of the function of this still very challenging protein.