The ferric uptake regulator (Fur) is a metal-dependent DNA-binding protein that acts as both a repressor and an activator of numerous genes involved in maintaining iron homeostasis in bacteria. It has also been demonstrated in Vibrio cholerae that Fur plays an additional role in pathogenesis, opening up the potential of Fur as a drug target for cholera. Here we present the crystal structure of V. cholerae Fur that reveals a very different orientation of the DNA-binding domains compared with that observed in Pseudomonas aeruginosa Fur. Each monomer of the dimeric Fur protein contains two metal binding sites occupied by zinc in the crystal structure. In the P. aeruginosa study these were designated as the regulatory site (Zn1) and structural site (Zn2). This V. cholerae Fur study, together with studies on Fur homologues and paralogues, suggests that in fact the Zn2 site is the regulatory iron binding site and the Zn1 site plays an auxiliary role. There is no evidence of metal binding to the cysteines that are conserved in many Fur homologues, including Escherichia coli Fur. An analysis of the metal binding properties shows that V. cholerae Fur can be activated by a range of divalent metals.
The ferric uptake regulator (Fur) is found in many pathogenic bacteria where it acts as a transcriptional regulator controlling the expression of multiple genes in response to iron levels. Regulation, transport and maintenance of iron are vital to the cell (Andrews et al., 2003; Wandersman and Delepelaire, 2004). Iron is poorly soluble at physiological pH and is not readily available. In mammalian hosts, low levels of free iron are maintained by the action of iron-binding proteins such as extracellular transferrin, lactoferrin and intracellular ferritin. Consequently, the free iron concentration in plasma is very low, at around 10−18 M (Sritharan, 2006). Although iron is important for microbes, a high concentration of iron in the cell is lethal to them as it leads to the production of reactive hydroxyl radicals via the Haber Weiss and Fenton reactions (Haber and Weiss, 1932). It is clear that the host contains very low levels of free iron and the pathogen requires iron in order to multiply in host environments. In this situation, bacteria activate virulence genes to gain nutrition from host cells. Large numbers of virulence genes are regulated by Fur; for example, in Pseudomonas aeruginosa, expression of an alternative sigma factor is regulated by Fur, which leads to the transcription of some virulence genes such as endoprotease PrpL and exotoxin A (Leoni et al., 1996; Ochsner et al., 1996; Wilderman et al., 2001).
Fur is a dimeric metalloprotein of monomeric size ∼17 kDa with an N-terminal domain involved in DNA recognition domain and a C-terminal domain involved in dimerization (Coy and Neilands, 1991; Stojiljkovic and Hantke, 1995). The N-terminal domain contains a winged-helix motif, and up to four of these motifs can bind a 19 bp inverted repeat sequence known as a Fur-box (Escolar et al., 1999). As a classical repressor, the main function of Fur is iron homeostasis. It also helps organisms to survive acidic environments or stress (Hall and Foster, 1996). When iron levels are high, Fur binds to the Fur-box and prevents access of RNA polymerase to the downstream iron acquisition genes. In the presence of low iron, derepression of the iron uptake genes occurs (Andrews et al., 2003). In vitro, a number of other divalent cations can also activate Fur to bind DNA (Bagg and Neilands, 1987; de Lorenzo et al., 1987), and their binding affinities for Fur have been studied in detail: in the case of Escherichia coli Fur (EcFur), Fe(II), Fe(III), Co(II), Mn(II) and Zn(II) all activate Fur binding to DNA (Mills and Marletta, 2005).
Vibrio cholerae, the causative agent of cholera, has an absolute requirement for iron and has multiple systems for iron acquisition from the human host and environmental niches (Wyckoff et al., 2007). Regulation of iron acquisition genes is mediated primarily through the V. cholerae Fur (VcFur) protein, encoded at gene locus vc2106 and composed of 150 amino acids. Iron and Fur regulation in V. cholerae has been extensively studied showing that VcFur has both negative and positive regulatory functions and is able to repress almost all of the iron acquisition genes during iron-replete conditions (Mey et al., 2005). In addition, genes encoding the toxin-co-regulated pilus (TCP) and the large integron are positively regulated by VcFur. A fur gene knockout strain of V. cholerae shows very weak autoagglutination that is possibly due to defective expression or assembly of the TCP, an important virulence factor. In infant mouse studies, it was found that the fur knockout strain colonized poorly compared with a wild-type strain of V. cholerae. In practice, VcFur can control the expression of more than 100 genes in V. cholerae in an iron-dependent or iron-independent manner (Mey et al., 2005).
Our understanding of the Fur protein was significantly advanced by the determination of the crystal structure of the P. aeruginosa Fur (PaFur) (Pohl et al., 2003). This confirmed a dimeric protein with an N-terminal DNA-binding domain and C-terminal dimerization domain, with the helices implicated in binding to the major grooves of DNA positioned one helix turn apart that allowed modelling of a Fur–DNA complex. The PaFur structure also revealed two zinc binding sites per monomer. Site 1, located in the dimerization domain, is co-ordinated in a distorted octahedral geometry by a water molecule and four residues that are conserved across the Fur family. Iron exchange experiments followed by X-ray absorption spectroscopy suggested that this site was readily exchanged by Fe(II) and was assigned as the regulatory metal binding site. Site 2, connecting the DNA-binding and dimerization domains, has a zinc co-ordinated by four conserved residues in a tetrahedral geometry and was assigned as the structural zinc binding site (Pohl et al., 2003). This interpretation of the roles of the two sites has been challenged recently, based on an analysis of reported mutagenesis experiments on several Fur proteins. The current proposal is that site 2 is the regulatory site, with site 1 playing a subsidiary role (Lee and Helmann, 2007). This proposal has been given support through a recent in silico study of EcFur involving free energy calculations and molecular dynamics (Ahmad et al., 2009). Structural studies on the related Mycobacterium tuberculosis zinc uptake regulator (Zur or FurB) (Lucarelli et al., 2007), the peroxide sensitive regulator (PerR) from Bacillus subtilis (Traore et al., 2006) and the nickel uptake regulator (Nur) from Streptomyces coelicolor (An et al., 2009) have shown similar DNA-binding and dimerization domains to PaFur, but variations in the metal binding sites.
Given the key role of VcFur in the survival and pathogenesis of V. cholerae, we embarked on a structural investigation of VcFur with the ultimate aim of exploring the potential of developing an allosteric inhibitor of VcFur as a possible treatment for cholera. In this study, we present the crystal structure of VcFur to a resolution of 2.6 Å. An analysis of the metal binding characteristics of VcFur shows that the protein can bind to promoter DNA in the presence of several divalent metals, and that the crystal structure contains two zincs per monomer. The residues co-ordinating the zincs are very similar to those in PaFur; however, the co-ordination details, particularly in site 2, are different and this is possibly linked to a dramatic difference in the orientation of the DNA-binding domains relative to the dimerization domain. The VcFur structure, together with other studies on Fur homologues and paralogues, suggests a geometry of metal binding consistent with the hypothesis that site 2 is the regulatory site at which Fe(II) can bind in vivo. In contrast to many reports on EcFur, the VcFur structure shows no evidence for metal binding to any of the four cysteines that are conserved across most of the Fur family as well as M. tuberculosis Zur and B. subtilis PerR.
Structure of VcFur
The asymmetric unit of the crystal contains a dimer of VcFur. In each chain of the dimer the first two amino acids and the last 17 amino acids (134–150) are not visible in the electron density map. Each chain has two domains: the DNA-binding domain (residues 3–82) and the dimerization domain (residues 83–133), the latter of which has a disulphide bond connecting Cys-93–Cys-133 (Fig. 1). The dimerization interface of VcFur has a buried surface area of ∼1350 Å2, similar to the 1300 Å2 reported for the same interface in PaFur (Pohl et al., 2003). Superposition of the two VcFur chains gives a root mean square deviation (rmsd) of 1.3 Å for 130 Cα atoms; however, separate superposition of the DNA-binding and dimerization domains gives rmsds based on Cα atoms of 0.5 and 1.0 Å, respectively, reflecting a slightly different orientation of the DNA-binding domains in each subunit relative to the dimerization domain. The crystallographic B-factor distribution throughout the dimer shows that the DNA-binding domain is relatively well ordered compared with the dimerization domain. The electrostatic charge distribution over the dimer is shown in Fig. 1, and reveals a predominantly negative charge apart from the helical regions involved in DNA recognition, particularly helix 4. Two metal ions, refined as zincs, are present in each monomer and have the same co-ordination in each monomer. The electron density around each zinc in chain B is shown in Fig. 2. The zinc in the dimerization domain, Zn1, is tetra-co-ordinated by the side-chains of His-87, Asp-89, Glu-108 and His-125 with interaction distances ranging from 1.9 to 2.3 Å (Fig. 2A); these are the equivalent ligands to those in PaFur. The zinc that connects the DNA-binding and dimerization domains, Zn2, is tetra-co-ordinated by His-33, Glu-81, His-88 and His-90 with interaction distances ranging from 2.0 to 2.4 Å (Fig. 2B). The significant difference in comparison with PaFur is that His-88 replaces Glu-100 (in PaFur) as a ligand to Zn2 which is discussed below.
Overall comparison with PaFur
VcFur and PaFur share 50% sequence identity. The crystal structure of PaFur has a monomer in the asymmetric unit, the physiological dimer being generated by a crystallographic twofold axis. Superposition of the PaFur monomer with the A and B chains of VcFur gives an rmsd of 2.8 and 2.3 Å, respectively, reflecting the slightly different relative orientation of the DNA-binding domains in the VcFur dimer. A dramatically different orientation of the DNA-binding domains in VcFur compared with PaFur is revealed by superimposing the dimerization domains of a PaFur dimer onto the dimerization domains of the VcFur dimer (Fig. 3A). The structures of the individual domains of VcFur and PaFur are quite similar. The DNA-binding domains of VcFur (residues 3–82) and PaFur (residues 1–82) share 59% sequence identity and superimpose with an rmsd of 0.9 Å for both chains of VcFur (Fig. 3B). The dimerization domains of VcFur (residues 83–133) and PaFur (residues 83–135), which share 51% sequence identity, superimpose with an rmsd of 1.1 Å for both chains A and B of VcFur.
The program lsqkab in the CCP4 suite (CCP4, 1994) was used to compare the relative positions of the DNA-binding domains in the VcFur and PaFur dimers. In chain A of VcFur, an almost pure 32° rotation (the angle between the centroid vector and the rotation axis being 87°) relates this DNA-binding domain to that in PaFur. In chain B of VcFur, a more complex relationship exists involving a 24° rotation around an axis that is 74° to the centroid axis. It may be significant that in the VcFur crystal chain B has many more crystal contacts than chain A. As the individual domains are structurally very similar, the program DynDom (Hayward and Berendsen, 1998) was used to analyse the domain movement between VcFur and PaFur. This suggested that the bending region encompasses residues 82–88 of VcFur, with the DNA-binding domain rotating of 30° relative to the dimerization domain, in agreement with the lsqkab analysis. The most significant consequence of the change in the orientation of the DNA-binding domains is the distance between helix 4 involved in DNA binding. The geometry between the two helices 4 of the dimer was analysed with the program interhelix (http://nmr.uhnres.utoronto.ca/ikura/resources/resources.html). In the PaFur dimer, the two helices are 33 Å apart and oriented 86° to one another. In the VcFur dimer the equivalent helices are 25 Å apart and oriented 103° to one another. If dimers are constructed from the VcFur A or B chains superimposed by their dimerization domains, then a dimer formed from two A chains has the helices 27 Å apart, oriented by 102°. A dimer formed from two B chains has the helices 23 Å apart, oriented by 108°. The change in orientation is illustrated in Fig. 3A by the change in distance between the Cα atoms of a conserved arginine on helix 4.
Comparison of metal binding sites of VcFur and PaFur
Figure 4 highlights the metal binding sites of VcFur (chain A) in comparison with PaFur following superposition of the two monomers via their dimerization domains. The co-ordination of Zn1 involves the same conserved residues (two histidines, a glutamic acid and an aspartic acid), but in VcFur the zinc is tetra-co-ordinated whereas in PaFur the zinc is reported to be hexa-co-ordinated through a bidentate interaction with Asp-88 and by a water molecule. Although there is a water molecule in a similar position in VcFur, it is 3.5 Å away from Zn1. The main difference in the co-ordinated side-chains of Zn1 is the position of His-87 (His-86 in PaFur). The second metal binding site, Zn2, shows significant differences. In PaFur, no electron density was observed beyond Cβ for the side-chain of His-87. In contrast, the equivalent residue in VcFur, His-88, is ordered in both monomers and co-ordinates to Zn2 in place of the glutamic acid Glu-100 in PaFur. The closest carboxylate oxygen of Glu-101, the equivalent residue in VcFur, is 3.9 Å from Zn2. Comparison of PaFur with chain B of VcFur reveals a very similar picture, with relatively minor changes around the Zn1 position and significant changes around Zn2.
Metal binding characteristics of VcFur
An analysis of EcFur showed that it contains 2.1 mol of Zn(II) per mol of EcFur monomer (designated as Zn2 EcFur) when zinc was used in purification steps, and that 1 mol of Zn per monomer can easily be removed upon dialysis against EDTA (Althaus et al., 1999). When zinc was used in the purification of VcFur, inductively coupled plasma-mass spectrometry (ICP-MS) analysis revealed 2.2 mol of Zn(II) per mol of Fur monomer (Zn2 VcFur) and also 0.1 mol of nickel per mol of Fur monomer (Table 1). When VcFur was purified without using zinc in the purification steps, there was 0.9 mol of Zn(II) per mol of Fur monomer (Zn1 VcFur) and also 0.04 mol of nickel per mol of Fur monomer (Table 1). The presence of nickel in VcFur indicates the purification of this protein through nickel column pull-down. In contrast to EcFur, when treated with EDTA, VcFur retains only 0.1 mol of Zn(II) per mol of VcFur monomer (Table 1).
Table 1. Metals analysis of native and EDTA-treated VcFur samples.
The EDTA-treated VcFur, containing almost no zinc, was dialysed with three different metals individually (Table 2). In reconstitution with Zn(II), 2.3 mol of zinc was incorporated per mol of VcFur, consistent with the previous amount of zinc found after purification of VcFur in the presence of zinc. In the case of Fe(III) reconstitution, 1.6 mol of iron was incorporated per mol of VcFur. In the case of Mn(II), 2.3 mol of manganese was incorporated per mol of VcFur, much higher than the 0.2 mol of manganese reported for EcFur (Mills and Marletta, 2005).
Table 2. Analysis of metal uptake into EDTA-treated VcFur samples.
A DNA mobility shift assay was used to examine the interaction of VcFur with duplex DNA corresponding to the fur promoter of the gene vc2694 (Mey et al., 2005). Zn1 VcFur, which is the form containing one zinc per monomer, binds to the promoter (Fig. 5A). When MnSO4, ZnCl2 or FeCl3 was added to Zn1 VcFur in the binding and running buffers, there was also a positive DNA mobility shift (Fig. 5B–D). When EDTA was used in the gel preparation and binding buffer with Zn1 VcFur, no shift was found (Fig. 5E), suggesting that apo-VcFur does not bind to DNA. Upon adding metal salts to the apo-VcFur (150 μM ZnCl2, or 200 μM MnSO4 or 150 μM FeCl3) in the binding and running buffer reaction, a mobility shift occurred (Fig. 5F–H). No shift was observed when non-promoter poly A was used as a control (Fig. 5I). Similar results showing DNA binding were obtained using the fur promoter of gene vc2209 (results not shown).
Recombinant VcFur, when purified without the addition of metal salts, contains essentially one zinc per monomer [0.9 mol of Zn(II) per mol of VcFur, Table 1], and this form, Zn1 VcFur, is able to bind to promoter DNA (Fig. 5A). The addition of Zn(II), Mn(II) or Fe(III) to the Zn1 VcFur form maintains the ability of VcFur to bind to promoter DNA (Fig. 5B–D). Zinc can be removed from VcFur by EDTA treatment, giving 0.1 mol of Zn(II) per mol of VcFur similar to the value of 0.19 mol reported for PaFur (Lewin et al., 2002), and this form does not bind to promoter DNA (Fig. 5E). Adding back metals to EDTA-treated VcFur produces 2.3 mol of Zn(II), 1.6 mol of Fe(III) or 2.1 mol of Mn(II) per mol of VcFur (Table 2), and these forms are able to bind to promoter DNA (Fig. 5F–H). As reported for other Fur proteins, VcFur shares the ability to bind DNA in the presence of different metals (Bagg and Neilands, 1987; de Lorenzo et al., 1987). Analysis of metal binding to EcFur showed that in the presence of Zn(II) it contains 2.1 (Althaus et al., 1999) or 2.3 (Mills and Marletta, 2005) mol of Zn(II) per mol of EcFur, and that EDTA could remove 1 mol of zinc per monomer. Removal of all zinc from EcFur required urea treatment in addition to EDTA (Althaus et al., 1999). VcFur therefore contains two metal sites per monomer, and the fact that Zn1 VcFur can bind to DNA suggests that the single zinc per monomer is binding to the regulatory metal binding site, and that this is the high-affinity site that requires EDTA to remove the metal. Other divalent metals can bind at both sites and promote binding to DNA. A study of EcFur concluded that in vivo zinc concentrations are too low for Zn(II) to be a regulatory ligand, and that Fe(II) is the physiologically relevant ligand that activates Fur (Mills and Marletta, 2005). This is also likely to be the case for VcFur, but Mn(II) may also be a ligand given its ability to occupy both metal binding sites and activate DNA binding; however, an analysis of metal concentrations in V. cholerae would be needed to confirm this.
The crystal structure of PaFur revealed the structural details of the metal binding sites for the first time for a member of the Fur family (Pohl et al., 2003). It was suggested in the PaFur study that of the two zinc binding sites per monomer, the Zn(II) at site 1 (Zn1 in Figs 1 and 3) is a low-affinity metal binding regulatory site that can readily exchange iron causing a conformational change in the DNA-binding domain of the protein, and that the Zn(II) at site 2 (Zn2) may be a structural metal binding site which helps to maintain the integrity of the protein (Pohl et al., 2003). This interpretation has been challenged recently suggesting that site 1 (Zn1) is the low-affinity metal binding site with site 2 (Zn2) being the high-affinity iron-sensing site (Lee and Helmann, 2007). This interpretation has been supported by a recent molecular dynamics and free energy calculation study of the binding of metals to PaFur that proposes site 2 as the regulatory iron-sensing site recognizing a hexa-co-ordinated Fe(II) in an octahedral environment (Ahmad et al., 2009). This in silico study also suggested that His-87 in PaFur might be an additional ligand of the metal at site 2, making five residues in total co-ordinating the metal. Further evidence that the Zn2 site is the regulatory site comes from studies on the Fur paralogue PerR from B. subtilis (BsPerR) that identified five ligands at the regulatory Fe(II) or Mn(II) binding site (Lee and Helmann, 2006), and these correspond exactly with the five ligands at the Zn2 site indicated in Fig. 6. The M. tuberculosis zinc uptake regulator (Zur or FurB) has what is believed to be a regulatory zinc binding site in a similar position to Zn2, involving four co-ordinating residues Asp-62, Cys-76, His-81 and His-83, the latter histidines being conserved in other Zn2 sites (Fig. 6) (Lucarelli et al., 2007).
The structure of VcFur reported herein also reveals two metal binding sites per monomer, occupied by Zn(II). Zn1 at site 1 in VcFur is tetra-co-ordinated by His-87, Asp-89, Glu-108 and His-125, residues that are conserved across all Fur proteins, including EcFur (Fig. 6). Although the corresponding residues in PaFur co-ordinate the Zn(II), the co-ordination is reported to involve six interactions achieved through a bidendate interaction with Asp-88 and by a co-ordinated water molecule. Zn2 at site 2 in VcFur is tetra-co-ordinated to His-33, Glu-81, His-88 and His-90 in contrast to PaFur where His-32, Glu-80, His-89 and Glu-100 tetra-co-ordinate the Zn(II). His-87 in PaFur (the equivalent of His-88 in VcFur) is disordered in the crystal structure of PaFur, but the fact that it is observed as a ligand to the Zn(II) in VcFur lends support to its involvement in co-ordinating Fe(II) at the regulatory site (Ahmad et al., 2009). Glu-101 in VcFur (the equivalent of Glu-100 in PaFur) is ordered, but too far away to co-ordinate to the Zn(II); however, all five potential ligands His-33, Glu-81, His-88, His-90 and Glu-101 are conserved across all Fur proteins (Fig. 6), and may therefore be involved in a hexa-co-ordinated octahedral interaction with Fe(II). A His-90–Ala mutant of VcFur showed that His-90 is critical for sensing iron (Lam et al., 1994), as have mutational studies of the same residue in Fur from other species (Hall and Foster, 1996; Bsat and Helmann, 1999; Lewin et al., 2002). Mutational studies on Vibrio alginolyticus Fur, which shares 94% sequence identity to VcFur, shows that mutation of His-33 or His-90, ligands of Zn2, completely inactivates Fur, whereas mutation of His-87 and His-125, ligands of Zn1, resulted in just a twofold reduction of Fur activity (Liu et al., 2007). It can be concluded that site 1 (Zn1) plays an auxiliary role and is not critical for activity, and site 2 (Zn2) is the regulatory site.
There have been many reports, based on a study of EcFur, that four cysteines (Cys-92, Cys-95, Cys-132, Cys-137 in EcFur numbering) conserved throughout most of the Fur family are involved in co-ordination of zinc (Jacquamet et al., 1998), specifically the Cys-92–XX–Cys-95 motif (Coy et al., 1994; Gonzalez de Peredo et al., 1999), and that the redox state of these cysteines and co-ordination of zinc are essential for maintaining EcFur in a dimeric state (D′Autreaux et al., 2007). PaFur appeared to be an exception as the second cysteine of this motif is a threonine (Fig. 6). Despite having the conserved cysteines, the structure of VcFur presented here does not reveal any metal binding site involving the cysteines, but shows that there is a disulphide linkage between Cys-93 and Cys-133 (equivalent to Cys-92 and Cys-132 in EcFur), in contrast to the disulphide analysis of the monomer to dimer transition of EcFur (D'Autreaux et al., 2007). It may be that this disulphide observed in VcFur would not form in the reducing environment of the cell, and we cannot rule out zinc binding in vivo. In the VcFur structure there is no electron density beyond Cys-133, although mass spectrometry analysis confirms that the C-terminal residues are present in the crystal. The similarity between EcFur and VcFur in the number of metals bound per monomer, and the observation that the two zinc binding sites are conserved in the crystal structures of PaFur and VcFur, involving residues conserved across all Fur proteins, suggest that EcFur may also share the same two metal binding sites. How this squares with the reports on zinc binding to the conserved cysteines in EcFur is difficult to explain, although there are reports that the cysteines in EcFur do not bind zinc (Saito et al., 1991).
The B. subtilis PerR and M. tuberculosis zinc uptake regulator (Zur or FurB) have a structural zinc co-ordinated by four cysteines that can only be removed by protein denaturation, and this has been confirmed in crystal structures of PerR (Traore et al., 2006) and Zur (Lucarelli et al., 2007). In both of these cases the four cysteines exist in two Cys–XX–Cys motifs shown in Fig. 6. One is equivalent to the Cys-93–XX–Cys-96 motif in VcFur (Cys-92–XX–Cys-95 in EcFur), the other motif occurring at the C-terminus immediately after the final β-strand of the dimerization domain: Cys-126–XX–Cys-129 in Zur and Cys-136–XX–Cys-139 in PerR. The presence of two Cys–XX–Cys motifs, however, does not always guarantee zinc binding, as illustrated in the recent structural analysis of the related Nur from S. coelicolor (An et al., 2009). The C-terminal cysteine motifs in EcFur and VcFur are quite different and occur in a more extensive structural element after the final β-strand of the dimerization domain that is predicted to be disordered in EcFur and VcFur when using the ronn disorder prediction program (Yang et al., 2005). In EcFur the C-terminal motif is Cys-132–XXXX–Cys-137, and in VcFur is Cys-133–XXXXX–Cys-139. Interestingly, a recent study of Vibrio harveyi Fur that shares 96% identity with VcFur, including conservation of the cysteines, shows that the Vibrio Cys-93 and Cys-96 are not essential for function, and that the final 12 residues at the C-terminus that includes Cys-139 are also not functionally important (Sun et al., 2008).
A crystal structure of Fur bound to Fur-box DNA remains elusive, but there is extensive evidence that more than one Fur dimer binds to the Fur-box. One study described the Fur-box as a head-to-head-to-tail repeat of the simple hexamer GATAAT (Escolar et al., 1998). An alternative view is that the Fur-box contains a conserved 7-1-7 inverted repeat presented twice within the 19 bp consensus sequence (Baichoo and Helmann, 2002), allowing each Fur-box to bind two Fur dimers on opposite sides of the DNA. Comparison of the crystal structures of VcFur and PaFur reveals a significant shift in the position of the DNA-binding domain relative to the dimerization domain, such that the mid-point distance between helix 4 in the dimer reduces from 33 Å in PaFur to between 23 and 27 Å in VcFur. For example, the distance between the two Cα atoms of the conserved tyrosine in helix 4 of the dimer (Tyr-55 in PaFur, Tyr-56 in VcFur), a residue implicated in DNA recognition of the Fur-box (Tiss et al., 2005), is 31 Å in PaFur and 21 Å in the VcFur. This closer distance in VcFur between the two recognition helices may be more consistent with the form that would bind a 7-1-7 inverted repeat sequence. It could be envisaged that one particular form may be better stabilized through the hexa-co-ordination of Fe(II) at site 2 (Zn2) involving the five conserved residues His-33, Glu-81, His-88, His-90 and Glu-101 (VcFur numbering), and this is supported by the recent molecular modelling studies (Ahmad et al., 2009). In both the PaFur and VcFur structures zinc is only seen to form tetra-co-ordination at site 2 with the consequence that in VcFur the DNA-binding domains may not be in an optimal position for DNA binding. Clearly the Zn(II) form can bind to promoter DNA, but the conformational changes required for association with DNA would provide an energetic penalty that would reduce the binding affinity in line with the reduced affinity seen for Zn(II) EcFur compared with Fe(II) EcFur (Mills and Marletta, 2005).
In conclusion, these studies on VcFur support the hypothesis that site 2 is the regulatory site where Fe(II), or possibly Mn(II), binds to control repression in vivo. The variation in the position of the DNA-binding domains compared with that observed in PaFur suggests that Zn(II) binding at site 2 (Zn2) may not lock the DNA-binding domains in the optimal position for DNA binding. The combined structural studies on Fur homologues suggest that Fe(II) binding to site 2 would employ five conserved residues to form a hexa-co-ordinated interaction rather than the tetra-co-ordinated interaction seen with Zn(II), and that these additional ligands would help generate the optimal orientation of the DNA-binding domains. A more complete understanding still awaits the elucidation of a Fur–DNA complex, ideally in the presence of iron, but these studies on VcFur provide further insights into this important family of gene activators and repressors.
Cloning of V. cholerae fur gene
The fur gene at locus vc2106 in the V. cholerae genome was successfully amplified using the polymerase chain reaction (PCR) from genomic DNA of V. cholerae O1 El Tor strain N16961 (kindly provided by Dr E. Fidelma Boyd) using standard protocols. Two primers, forward (5′-CAGGAAAGTCCATGGCAGACAATAAC-3′) and reverse (5′-CGTAAAGAATTCGGTTATTTCTTCGGC-3′), were designed corresponding to the 5′ and complementary 3′ ends of fur gene with specific recognition sites for the restriction enzymes NcoI and EcoRI (underlined). The primers produce one mutation at position 2 of the protein from a serine to an alanine. After amplification, the PCR product was digested with NcoI and EcoRI enzymes and ligated with the similarly digested pEHISTEV vector, an engineered variant of pET30 with an N-terminal six-histidine tag that is cleavable using tobacco etch virus (TEV) protease and leaves two additional residues (Gly–Ala) at the N-terminus (Liu and Naismith, 2009). Cloning was verified by colony PCR and DNA sequencing.
Protein expression, optimization of solubility and purification
The pEHISTEV vector containing the fur gene was transformed into E. coli strain C43 (DE3) using standard methods. The best soluble protein expression was achieved after 24 h growth at 25°C with the addition of 0.5 mM IPTG (isopropyl-beta-d-thiogalactopyranoside). The cell pellet was re-suspended in buffer A (50 mM HEPES, 500 mM NaCl, 10 mM imidazole and 10% glycerol) containing DNase and EDTA-free protease inhibitor cocktail (Roche Applied Science) and sonicated. After centrifugation, the soluble fragment was syringe filtered through a 0.22 μm filter unit (Millipore) and the target protein was purified using a HisTrap™ HP column (GE Healthcare), according to the manufacturers' instructions.
Eluted samples were dialysed against 4 l of buffer B [buffer A with addition of 1 mM DTT (dithiothreitol)] for 4 h at 4°C with two buffer changes. Next, TEV protease was added into the dialysis tube containing the His-tagged protein (1 μg of protease for 1 mg of protein at 25°C overnight). The digested protein sample was analysed by SDS-PAGE to check the efficiency of TEV protease digestion. At this stage protein samples were dialysed against buffer A to remove DTT. Next, protein samples were syringed filtered through a 0.22 μm filter unit (Millipore) and loaded into the nickel column as before and the flow-through was collected in fractions and analysed by SDS-PAGE. Corresponding fractions were pooled and dialysed against 4 l of buffer containing buffer B overnight at 4°C, and then concentrated to 4.5 ml by using a 10 000 Da MWCO centrifugal concentrator (Sigma-Aldrich). In the final stage of VcFur purification, 4.5 ml concentrated samples were loaded onto a Hiprep™ 16/60 Sephacryl™ S-200 gel filtration column (GE Healthcare). For crystallization trials, fractions containing VcFur were pooled and dialysed overnight at 4°C against 4 l of buffer A with the addition of 5 mM DTT and 1 mM ZnCl2. Next day, the sample was dialysed against buffer A for 4 h at 4°C with two buffer changes. Finally, the protein was dialysed in 10 mM HEPES, 50 mM NaCl, 5% glycerol. Mass spectrometry was carried out to confirm identity of VcFur.
Crystallization of VcFur
A pre-crystallization test (Hampton Research) indicated the appropriate protein concentration (7 mg ml−1) for crystallization screening that was carried out using the robotics of the Scottish Structural Proteomics Facility and a series of different screening conditions. After 4 weeks, initial hits with tiny rod shaped crystals were observed in Nextal PEGs screens (Qiagen). Crystal plates were set up at 10 mg ml−1 with a 192-condition optimization screen based on the initial hits. After 6 weeks, crystals suitable for X-ray analysis were produced in the condition 0.18 M magnesium nitrate hexahydrate and 16% PEG3350.
Structure solution and refinement
Crystals were cryoprotected in crystallization buffer with the addition of 20% (v/v) glycerol, and X-ray data to 2.6 Å were collected at 100 K on an in-house X-ray facility, a Rigaku/MSC MicroMax-007HF rotating anode equipped with focusing optics and a Saturn 944+ CCD detector. X-ray data were integrated and scaled using HKL2000 (Otwinowski and Minor, 1997) and the CCP4 (CCP4, 1994) suite. The crystals belonged to the space group P41/3212. The unit cell suggested two monomers in the asymmetric unit giving a Matthew's coefficient of 2.47 Å3 Da−1 with a 50% solvent content. The structure of VcFur was solved by molecular replacement using the program phaser (McCoy et al., 2007), by splitting the search model into two domains: the dimerization domain from PaFur (residues 84–134 from PDB code 1MZB), and the DNA-binding domain made up by superimposing the DNA-binding domains from EcFur (PDB code 2FU4) and PaFur (residues 1–83 from PDB code 1MZB). Zinc atoms were removed from the search models. A solution with Z-scores of 6.3 and 29.2 for the rotation and translation function steps, respectively, was obtained in the space group P43212. The final model was built manually using the program coot (Emsley and Cowtan, 2004) and the refinement was carried out using refmac (Murshudov et al., 1997) and phenix (Adams et al., 2002). The data collection and refinement statistics are summarized in Table 3. Atomic co-ordinates and structure factors have been deposited in the protein databank with code 2w57.
Table 3. Data collection and refinement statistics.
Rmerge = ΣhΣj|Ihj − 〈Ih〉|/ΣhΣj, where Ihj is the intensity of the jth observation of unique reflection h.
R factor = Σh||Foh| − |Fch||/Σh|Foh|, where Foh and Fch are the observed and calculated structure factor amplitudes for reflection h.
Free R factor is equivalent to R factor, but is calculated using 5% of reflections excluded from the maximum-likelihood refinement stages.
The fur promoter of gene locus vc2694 was chosen for the gel-shift assays, based on a study of fur promoters in V. cholerae (Mey et al., 2005), with forward and reverse oligonucleotides purchased from VHBIO, with one oligo labelled with FITC (fluorescein isothiocyanate). The oligonucleotide was 5′-TTGTTAATGATATTAATTATCATTAACAT, where the 19 bp Fur-box is underlined. The oligos were further purified using denaturing gel. Working solutions containing 2 μM of forward and reverse oligos were prepared in buffer (20 mM MES pH 6.5). Eppendorfs containing the oligo mix were placed into a water bath at 90°C then allowed to cool down overnight at room temperature. The resultant double-stranded DNA duplex was stored at −20°C.
The DNA mobility shift assay was performed using the published method with minor modifications (Ochsner et al., 1995). The FITC-labelled double-stranded DNA fragment (100 nM) was mixed into 20 μl of binding buffer which contains 10 mM bis-Tris borate (pH 7.5), 40 mM KCl, 0.1 mg ml−1 bovine serum albumin (BSA) (Sigma fraction V) and 10% glycerol. Different concentrations of native VcFur were added and the mixture was incubated at 37°C for 15 min using a static incubator. For the negative control, 100 nM FITC-labelled double-stranded DNA (poly A and poly T) was used. Two microlitres of 10% glycerol was added to the samples and 10 μl of samples in each sample were loaded into gel well under 60 V constant voltage. The gel was run at 10 mA constant current for 80 min in running buffer (20 mM bis-Tris borate). After the run, the gel was scanned at 473 nm using Fujifilm FLA-5000 series scanner.
Metal binding studies of VcFur
Two batches of VcFur were purified using standard methods described above into a final buffer containing 20 mM HEPES, 200 mM NaCl and 10% glycerol at pH 7.5. One batch, batch 1, was dialysed in 20 mM HEPES, 200 mM NaCl, 1 mM ZnCl2 and 2 mM DTT and 10% glycerol, followed by removal of extra metal ion by dialysis of protein samples in the same buffer without metal solution. Batch 2 was not dialysed against any zinc-containing buffer. Finally both samples were dialysed overnight at 4°C against 20 mM HEPES, 20 mM NaCl and 5% glycerol at pH 7.5. The concentration of protein was measured using NanoDrop (Thermo Scientific).
Removal of metals from VcFur
To remove metal ions (mainly Zn) from VcFur, the experiment was carried out with minor modification as previously described (Althaus et al., 1999). Both batches of VcFur (50 μM) were dialysed overnight against 200 mM EDTA, 20 mM HEPES, 20 mM NaCl and 5% glycerol at pH 8.0. To remove extra EDTA, the samples were dialysed again in same buffer without EDTA for 3 h followed by dialysis into 20 mM HEPES, 20 mM NaCl and 5% glycerol at pH 7.5 with three buffer changes. All the procedures were carried out at 4°C.
Addition of metals into EDTA-treated VcFur
Three different metals Zn(II) (as ZnCl2), Mn(II) (as MnSO4) and Fe(III) (as FeCl3) were added into EDTA-treated VcFur. The procedure for addition of metal into VcFur was followed with minor modifications (Mills and Marletta, 2005). To incorporate Zn(II), 50 μM VcFur was dialysed overnight with three equivalents of ZnCl2 (150 μM at pH 6.2) in 20 mM HEPES, 50 mM NaCl and 5% glycerol at pH 6.2 in the presence of a 20-fold excess of DTT. For Mn(II), 50 μM VcFur was dialysed overnight with four equivalents of MnSO4 (200 μM) in 20 mM HEPES, 50 mM NaCl and 5% glycerol at pH 7.5. For of the incorporation of the Fe(III), ferric chloride (FeCl3) was first dissolved with a 10-fold excess of citric acid which was adjusted to pH 7.0 and 50 μM VcFur was dialysed overnight with three equivalents of FeCl3 (150 μM) in 20 mM HEPES, 50 mM NaCl and 5% glycerol at pH 7.0. To remove the excess metal, all samples were dialysed against 20 mM HEPES, 20 mM NaCl and 5% glycerol at pH 7.5, without adding metal salts for 4 h followed by three buffer changes.
Analysis of metals
The metal contents of the proteins were analysed using ICP-MS at the University of Edinburgh. All the samples that were subject to metal analysis were finally dialysed against a buffer containing 20 mM HEPES, 20 mM NaCl and 5% glycerol at pH 7.5 and the machine was equilibrated with the same buffer to minimize any metal contamination from the buffer itself. Standard metal solutions were measured between every sample measurement to maintain higher sensitivity. Measurements were carried out in triplicate.
M.A.S. was supported by the University of St Andrews. Facilities of the Scottish Structural Proteomics Facility based in St Andrews, and funded by the Scottish Funding Council and the BBSRC, were used for the crystallization of VcFur. We thank Malcolm White, Melina Kerou and Christophe Rouillon for help with the EMSA experiments.