FUR (Ferric uptake regulator) proteins are among the most important families of transcriptional regulators in prokaryotes, often behaving as global regulators. In the cyanobacterium Anabaena PCC 7120, FurB (Zur, Zinc uptake regulator) controls zinc and redox homeostasis through the repression of target genes in a zinc-dependent manner. In vitro, non-specific binding of FurB to DNA elicits protection against oxidative damage and avoids cleavage by deoxyribonuclease I. The present study provides, for the first time, evidence of the influence of redox environment in the interaction of FurB with regulatory zinc and its consequences in FurB–DNA-binding affinity. Calorimetry studies showed that, in addition to one structural Zn(II), FurB is able to bind two additional Zn(II) per monomer and demonstrated the implication of cysteine C93 in regulatory Zn(II) coordination. The interaction of FurB with the second regulatory zinc occurred only under reducing conditions. While non-specific FurB–DNA interaction is Zn(II)-independent, the optimal binding of FurB to target promoters required loading of two regulatory zinc ions. Those results combined with site-directed mutagenesis and gel-shift assays evidenced that the redox state of cysteine C93 conditions the binding of the second regulatory Zn(II) and, in turn, modulates the affinity for a specific DNA target. Furthermore, differential spectroscopy studies showed that cysteine C93 could also be involved in heme coordination by FurB, either as a direct ligand or being located near the binding site. The results indicate that besides controlling zinc homeostasis, FurB could work as a redox-sensing protein probably modifying its zinc and DNA-binding abilities depending upon environmental conditions.
The FUR (Ferric uptake regulator) family is a widespread group of prokaryotic transcription factors mainly controlling the homeostasis of metals such as iron (Fur subfamily), zinc (Zur, Zinc uptake regulator), manganese (Mur, Manganese uptake regulator) and nickel (Nur, Nickel uptake regulator). FUR proteins also intervene in the regulation of peroxide-stress response (PerR) and heme availability (Irr) . Anabaena sp. PCC 7120 (hereinafter Anabaena sp.) contains three paralogues of the FUR family, namely FurA, FurB and FurC . FurA is a global regulator in cyanobacteria and is mainly responsible for controlling iron homeostasis  and FurC has been proposed to act as a PerR regulator . FurB, the protein under study in this work, acts as a Zur regulator in Anabaena sp., controlling zinc homeostasis in a zinc-dependent manner . Also, FurB participates in the control of the oxidative stress response regulating the expression of key antioxidant enzymes  and protects DNA against DNase I (deoxyribonuclease I) cleavage and ROS damage in vitro .
The most extended model for FUR–DNA interaction assumes that FUR proteins generally act as classical repressors, binding in the dimeric form to palindromic A/T-rich sequences at the promoter regions of target genes. Thus, under metal repletion conditions, FUR proteins co-ordinate the regulatory metal acting as co-repressor and undergo conformational changes that trigger FUR binding to the A/T-rich sequences (FUR-boxes), blocking transcription of target genes by RNA polymerase . Resolution of the X-ray crystallographic structures from several FUR proteins supports the hypothesis that the binding of the co-repressor metal provokes a local conformational change leading to a pivotal motion of the DNA-binding domain with respect to the dimerization domain [9–16]. In some cases, this motion would allow the regulator to acquire a ‘closed conformation' in which DNA recognition by the two N-terminal domains in the dimer is possible. X-ray structures of FUR proteins, such as Pseudomonas aeruginosa Fur , Streptomyces coelicolor Zur and Nur [10,11], Vibrio cholerae Fur  and Escherichia coli Fur , present a closed conformation and have been suggested to present optimal DNA binding. Conversely, structures of Streptococcus pyogenes PerR , Helicobacter pylori Fur  and Mycobacterium tuberculosis FurB/Zur  containing structural and regulatory zinc ions exhibit an open conformation. In these cases, the N-terminal domains are too far from each other to permit optimal interaction with DNA, which probably results in a weaker DNA–protein interaction .
As derived from crystallographic studies, metal coordination by FUR proteins affects their DNA-binding activity and is therefore an essential aspect to study when characterizing a FUR protein. Typically, FUR proteins contain one structural zinc ion per monomer and a second regulatory metal-binding site. However, some of the crystal structures available in the PDB (Protein Data Bank) exhibit a third binding site, whose functionality is still under debate. Thus, the location and ligands involved in metal coordination usually help to figure out their function . Most of the FUR proteins known to date contain a structural binding site for zinc (site 1), which normally involves four cysteine residues arranged in CXXC motifs and that is located in the C-terminal domain [10,11,14–17,19,20]. In all structures, the site present in the region between N- and C-terminal domains is responsible for high-affinity coordination of the regulatory metal and is usually called site 2. Finally, some proteins, such as Fur from H. pylori , FurB/Zur from M. tuberculosis  and Zur from S. coelicolor , present a third binding site in the core of the dimerization domain. The function of this site is still controversial; it has been proposed to have a role in dimer formation, but it could also work as a second regulatory site, allowing the protein respond to a broader range of metal concentrations.
Apart from the coordination of regulatory metals, the ability of FUR proteins to bind DNA can also be modulated by the interaction with other ligands, such as heme. Thus, the interaction of FurA and FurB from Anabaena sp. with heme has been shown to inhibit protein activity in vitro [7,21]. Binding of FurA to heme involves cysteine C141 . This residue is part of a CP motif, classically described as a ‘heme regulatory motif'  that is also present in the FurB paralogue . The interaction of FurA and FurB with heme would have interesting implications in vivo because, under oxidative stress conditions when heme levels increase, they would be released from DNA and would permit transcription of genes involved in the oxidative stress response, such as dpsA  and prxA .
Although FurB has been described as a Zur regulator , little is known about its zinc-binding ability and its influence in the interaction with DNA. With the aim to gain new insights into these processes, we first developed a purification procedure that allowed us obtaining enough pure and active protein for biochemical characterization. Our analyses revealed that recombinant FurB contains one structural zinc ion per monomer and is able to bind up to two additional regulatory zinc ions. Evidence for the consequences of the redox environment of FurB on zinc loading and its affinity for DNA, as well as for the role of C93 in FurB functionality is provided.
Bacterial strains and culture conditions
Bacterial strains used in the present study are described in Supplementary Table S1. E. coli strains were grown in LB medium at 37°C with constant shaking of 180 r.p.m. When necessary, the medium was supplemented with kanamycin 50 µg ml−1. Anabaena sp. was grown at 28°C in BG-11 medium , under illumination of 50 µE m−2 s−1 and constant shaking of 120 r.p.m.
Protein expression and purification
Anabaena sp. genomic DNA was extracted as previously described  and used as a template to amplify the furB gene with oligonucleotides all2473N-2 and all2473C (Supplementary Table S2). The furB gene was cloned into NdeI–HindIII restriction sites of a pET-28a(+) plasmid (Novagen), so that FurB protein was expressed with a N-terminal His-tag. A FurB–C93A mutant was obtained with the Quik-Change site-directed mutagenesis kit (Stratagene) and oligonucleotides C93A_up and C93A_dw (Supplementary Table S2). Plasmids were verified by DNA sequencing and transformed into E. coli BL21-competent cells for protein overexpression. Cells were grown until the exponential phase and treated with 1 mM IPTG (isopropyl β-d-1-thiogalactopyranoside) during 3 h for optimal protein expression.
For recombinant protein purification, 10 g of FurB/C93A-overexpressing cells were resuspended in 50 ml of buffer A [0.1 M NaH2PO4, 0.01 M Tris, 2 M GdnHCl (guanidine hydrochloride) pH 8] containing 1 mM PMSF (phenylmethylsulfonyl fluoride). Cells were lysed by 10 cycles of sonication at 190 W and the suspension was centrifuged to remove cell debris. The supernatant was loaded into a Chelating Sepharose Fast Flow column (Amersham) that was previously charged with zinc according to the manufacturer's conditions. The column was washed with five volumes of a solution containing 0.5 M (NH4)2SO4 in buffer A. Then, the column was subjected to an additional wash with 35 mM glycine in buffer A until OD280 became lower than 0.1. Recombinant proteins were eluted from the column with a linear gradient of 0–1 M imidazole in buffer A. Fractions containing the higher amount of target protein were pooled, dialysed against 10 mM acetic acid/acetate buffer (pH 5.5) and stored at −20°C until use. When necessary, proteins were concentrated with an Amicon Ultra 3 K device (Millipore). Proteins were quantified by using the bicinchoninic acid protein assay (Pierce).
4-(2-Pyridylazo)-resorcinol staining and Inductively Coupled Plasma Mass Spectrometry assay
The presence of Zn(II) in the protein was verified by PAR [4-(2-Pyridylazo)-resorcinol] staining. PAR is a reagent that forms an orange-coloured complex with zinc . A denaturing gel containing 5 μg of the protein was incubated for 5 min with a solution [20 mM HCl–Tris (pH 8), 100 mM NaCl, 5% glycerol] containing 500 µM PAR. To visualize the complex, 50 mM H2O2 was then added. Zn(II) was quantified by inductively coupled plasma mass spectrometry (ICP-MS) at the Servicio de Ionómica from CEBAS-CSIC (Murcia, Spain). For the determination of Zn(II) content, 22.16 nmol of precipitate-free FurB were scaled up to 10 ml with ICP-grade HNO3.
Electrophoretic mobility shift assay
Gene promoters were obtained by PCR amplification, using Anabaena sp. genomic DNA as a template and with oligonucleotides in Supplementary Table S2. EMSA (electrophoretic mobility shift assay) was performed as previously described , using the all4725 gene promoter as target DNA . To evaluate the influence of heme in the DNA-binding ability of the proteins, a 2 µM heme solution in 50 mM Tris–HCl (pH 8) was prepared and added to the proteins at 1:5 or 1:10 protein:heme molar ratio.
DNA protection assays
The non-specific binding of FurB to DNA was analysed by in vitro DNA protection assays, with a similar procedure to that described in . FurB (80 pmol) was incubated with 100 ng pUC18 plasmid for 10 min at room temperature. Reaction volume was scaled up to 60 µl with a solution of 10 mM Bis–Tris/HCl (pH 7.5), 40 mM KCl, 0.1 mM MgCl2 and 5% (v/v) glycerol. Then, 100 µU of DNase I was added to the sample and incubated for 20 additional minutes. The reaction mixture was divided into two 30 µl samples and proteins were removed from one sample by using the GFX PCR DNA and Gel Band Purification kit (GE Healthcare). As a control, the same experiment was performed with BSA instead of FurB and negative controls (without FurB/BSA) were also included. All samples were analysed by electrophoresis in a 0.8% agarose gel and ethidium bromide staining.
Difference spectroscopy was used to study the interaction of FurB/C93A with heme. To prepare heme solutions, 3 μg of heme (protoporphyrin IX) were dissolved in 1 ml of 0.1 N NaOH/ethanol (1:9 v/v). The suspension was filtered through a 0.22 µm filter and heme concentration was determined by spectrometry, using the extinction coefficient for heme at 385 nm in 0.1 N NaOH (58.44 mM−1 cm−1) . From that stock solution, a 2 µM heme solution in 50 mM Tris–HCl (pH 8) was prepared. Owing to the tendency of heme to dimerize, fresh solutions were prepared for each experiment. Heme solutions were also routinely protected from light. To obtain protein–heme complexes, successive aliquots of a 70 µM protein solution were added into a quartz cuvette containing the 2 µM heme solution and spectra were recorded. To correct volumes, equivalent aliquots of protein buffer (10 mM acetic acid/acetate, pH 5.5) were added to the reference cuvette.
Atomic force microscopy
Adequate DNA for AFM (atomic force microscopy) experiments was constructed as follows: the all4725 gene promoter was amplified from Anabaena sp. genomic DNA using oligonucleotides Pall4725_F, Pall4725_R in Supplementary Table S2. The PCR product was cloned into a pGEM®-T Easy Vector System (Promega) and the construct was verified by DNA sequencing. To prepare binding reactions, FurB was incubated with the 5 ng/µl construct in binding buffer, 10 mM Bis–Tris, 40 mM KCl, 5 mM MgCl2, 0.1 mM MnCl2 with or without 5 µM ZnSO4, pH 7.5 and freshly prepared 2 mM DTT (dithiothreitol), for 20 min at room temperature. Biomolecules were then immobilized on freshly cleaved muscovite mica sheets (Electron Microscopy Sciences). Mica pieces were pretreated with 200 mM MgCl2 for 5 min and rinsed with binding buffer prior to biomolecule binding as optimized previously to slightly immobilize DNA strands on such negative surfaces . The concentration of the DNA incubated on the substrates was suitable to get isolated complexes that could be analysed individually.
Images were taken in binding buffer with a Cervantes Fullmode Scanning Probe Microscope (Nanotec Electrónica S.L.). V-shaped silicon nitride cantilevers with integrated pyramidal 2 nm ultrasharp tips exhibiting a spring constant of 0.06 N/m (SNL, Bruker Probes) were used. Jumping mode applying a very low force to work in a repulsive regime was used . Several controls were performed to define the adsorption properties of FurB on naked and pretreated substrate; the effect of FurB concentration on DNA binding; the effect of absence and higher zinc concentrations on binding; and the effect of the protein net charge on binding through incubation of DNA with avidin (Sigma–Aldrich) that has a similar isoelectric point as that of FurB. Image processing was carried out with the WSxM software . Furthermore, each feature or associate was analysed in detail with the zoom function of the program, performed without losing image information and discarding possible artefacts.
Isothermal titration calorimetry
FurB/C93A interaction with zinc was analysed with a MicroCal Auto-iTC200 calorimeter (Malvern) at a constant temperature of 20°C. Protein solutions (FurB and C93A) were used at a final concentration of 20 µM in freshly prepared 10 mM acetic acid/acetate buffer (pH 5.5) to avoid protein precipitation. Ligand solution contained 400 µM ZnSO4 in 10 mM acetic acid/acetate buffer (pH 5.5). Although this is not a physiological pH, it was selected to avoid the formation of insoluble zinc hydroxides. Extrapolating at neutral pH, a considerable increase in zinc-binding affinity is expected. Despite the effect of the pH on the binding affinity, the stoichiometry of binding will not be affected. Protein and ligand solutions were degassed for 2 min at 10°C in a ThermoVac (MicroCal) prior to loading into the system. A sequence of 2-µl injections (0.5 µl/s injection flow) spaced 150 s and a stirring speed of 750 r.p.m. were programmed. The dissociation constant was obtained through nonlinear regression of experimental data to a model for either a single ligand-binding site or two independent and identical sites, as explained elsewhere . No evidence for more complex models (e.g. non-identical or cooperative binding sites) was observed. When necessary, a reducing agent (DTT) was added to the solutions at a final concentration of 2 mM. Considerable interaction between DTT and zinc has been reported before, but in our case, DTT has not competitively affected the interaction of zinc with FurB; in fact, the binding affinity for FurB was higher in the presence of DTT. Appropriate controls were performed: calcium-EDTA titrations to calibrate/test the calorimeter, and zinc dilution into buffer in order to evaluate the heat effect associated with that process (heat effects were small and uniform along this dilution experiments, and there was no need to subtract them in the protein-zinc titrations).
Homology modelling was driven with Modeller  using the structures of H. pylori Fur (PDB ID: 2XIG) , M. tuberculosis FurB/Zur (PDB ID: 2O03)  and S. coelicolor Zur (PDB ID: 3MWM)  as templates. Molecular structures and structural models were inspected, analysed and plotted with PyMol 1.4.1 (Schrodinger LLC). The C93 residue in Anabaena sp. structural model and the conservative positions in H. pylori (P114), M. tuberculosis (H98) and S. coelicolor (P102) are shown in sticks. Secondary structure prediction was made with PSIPRED . Surface electrostatic potentials were calculated and represented using the PyMol 1.4.1 Program.
FurB purification and assessment of specific and non-specific DNA binding
In order to study FurB interaction with different ligands, a sufficient amount of pure soluble protein was required. Previous attempts to obtain recombinant FurB with an ionic exchange column followed by metal affinity chromatography  or with a heparin column followed by ionic exchange  yielded very low-protein concentration. Since FUR proteins present a remarkable stability upon chemical denaturation with guanidine hydrochloride [GdnHCl] , purification of fully active FUR proteins in the presence of a subdenaturing concentration of this chaotropic agent has been successfully used . In this work, purification of FurB by means of a one-step metal affinity chromatography with Zn(II) in the presence of GdnHCl yielded FurB with purity above 90% (Figure 1a). Most of the protein was obtained in the monomeric form although, to a much lesser extent, dimers (∼34 kDa) could also be observed. After chromatography, FurB was dialyzed in the presence of 10 mM acetic acid/acetate (pH 5.5) since it conferred higher protein stability and resistance to freeze–thawing cycles than other buffers tested. Further attempts to dialyze the protein against higher pH buffers led to protein precipitation.
Purification of FurB and analysis of dual DNA binding.
To confirm proper folding of FurB after dialysis, different experiments were performed to determine the activity of the protein. Those experiments were based on the ability of FurB to bind DNA in vitro, both in a specific and in a non-specific manner, regardless of the purification procedure . The tests included EMSA, DNA protection assay and AFM imaging. As shown in Figure 1b, EMSA assays demonstrated that recombinant FurB was able to recognize and specifically bind the all4725 gene promoter, a previously described FurB target encoding porphobilinogen synthase , leading to the formation of well-defined protein–DNA complexes in the presence of a reducing agent (DTT) and the Zn(II) co-repressor.
Non-specific binding of FurB to DNA was verified by protection assays, incubating FurB with a pUC18 plasmid and subjecting the mixture to DNase I treatment. The results established that FurB protected DNA because the incubation of pUC18 with protein decreased the amount of nicked plasmid upon addition of DNase I (Figure 1c). In fact, protein bound the DNA so effectively that results could only be easily interpreted when FurB was removed from the sample and plasmid could freely run into the agarose gel.
Apart from EMSA and protection assays, AFM experiments were also performed to obtain a deeper insight into both specific and non-specific binding of FurB to DNA. Figure 2a shows a FurB monomer bound to a single DNA molecule, presumably to the Zur-box of the all4725 promoter cloned in the pUC18 plasmid (Supplementary Figure S1). The size observed for FurB coating DNA, ∼6 nm, agrees with previous measurements obtained for FurB (4.0 ± 1 nm) and free DNA (1.5 ± 0.5 nm) on naked mica (Supplementary Figure S2). No protein was found unbound to DNA on Mg2+ pretreated mica, as can be deduced from the height profiles drawn in the images with DNA. These results were corroborated with those obtained in control measurements scanning FurB on positively charged pretreated mica, in the absence of DNA, where no protein molecules remain bound (not shown).
Analyses of FurB–DNA interaction by AFM.
Probably due to the high affinity of FurB for DNA, non-specific binding of FurB along the plasmid was already observed at protein concentrations ∼1–2 nM (Figure 2b) even in the absence of zinc (Figure 2c). Hence, it was not possible to observe binding of FurB in the dimeric form that, according to the classical model, is the oligomeric state of FUR proteins acting as transcriptional regulators. When a higher concentration of regulator was assayed, from 10 to 500 nM, the DNA strand was covered by FurB oligomers of higher degree. This phenomenon was similar in the absence of zinc (Supplementary Figure S3) or when zinc concentration was increased (Figure 2d). Since avidin did not bind to the DNA strands at the working conditions, it can be concluded that the binding interaction is not based just on an electrostatic adsorption (Supplementary Figure S2). Furthermore, it should be noted that a previous analysis of FurA–DNA interaction showed still specific binding of FurA to the PfurA (Promoter region of the furA gene) target at protein concentration as high as 5 µM .
Analyses of zinc binding by FurB and its C93A derivative
To get a better understanding of the metalloregulatory properties of FurB, we analysed its zinc content with a PAR staining assay and ICP-MS. Recombinant protein was subjected to SDS–PAGE electrophoresis and the resulting gel was stained with a PAR solution (Supplementary Figure S4). The formation of an orange-coloured band evidenced the presence of zinc in FurB and, consequently, an ICP-MS analysis was carried out in order to establish the stoichiometry of zinc ions. Calculations derived from mass spectrometry results determined a content of ∼1.2 moles of zinc per mole of protein, that is, 1 atom of Zn(II) per monomer. As will be discussed later, the zinc ion detected is likely to be the structural metal that, according to the alignment of FurB with the sequences of FUR proteins from different bacteria, it is probably co-ordinated by the highly conserved cysteines C81, C84, C121 and C124 (Figure 3).
Alignment of a subset of FUR proteins from different bacteria.
Once established that purified FurB contained one zinc ion, the presence of additional binding sites was examined by ITC (isothermal titration calorimetry) under reducing and non-reducing conditions (Figure 4a,b). The analysis of the experimental data through nonlinear regression  led to the conclusion that in the absence of reducing agent FurB exhibited one binding site with affinity in the micromolar range. However, in the presence of DTT, FurB displayed two binding sites with similar affinities. DTT also increased the binding affinity and made the enthalpy of the interaction less unfavourable (Table 1). The fact that zinc binding was affected by the redox state of FurB led us think about the possible implication of the remaining cysteine residue, C93, in the coordination of one of the regulatory zinc ions. From now on, we will consider regulatory binding sites/ions to those detected by ITC, while the Zn(II) detected by PAR staining and ICP-MS will remain considered structural.
ITC experiments for Zn(II) interacting with FurB/C93A.
|Sample .||K (M−1) .||Kd (µM) .||ΔH (kcal/mol) .||n .|
|FurB||2.9 × 104||35.0||6.3||0.99|
|FurB + DTT||1.5 × 105||6.5||0.7||2.20|
|C93A||4.7 × 104||21.0||5.0||0.93|
|C93A + DTT||1.3 × 105||7.8||0.8||0.90|
|Sample .||K (M−1) .||Kd (µM) .||ΔH (kcal/mol) .||n .|
|FurB||2.9 × 104||35.0||6.3||0.99|
|FurB + DTT||1.5 × 105||6.5||0.7||2.20|
|C93A||4.7 × 104||21.0||5.0||0.93|
|C93A + DTT||1.3 × 105||7.8||0.8||0.90|
With the aim of studying the potential entanglement of C93 in the coordination of regulatory zinc and therefore in FurB metalloregulation, a C93A mutant was obtained and purified using the same procedure already described for FurB (Figure 5a). As verified by EMSA assays, the C93A mutant was also active (Figure 5b). Although C93A activity is lower than that of the WT FurB, Figure 5b shows that the C93A mutant is also able to bind the all4725 promoter, while it does not interact with the nifJ promoter. In the line 1 (control), there is no shift of the all4725 promoter in the absence of protein (C93A), while in lanes 2 to 6, containing DNA samples previously incubated with increasing concentrations of C93A, an additional band consisting in a C93A-all4725 DNA complex can be observed. In lanes 7 and 8 higher molecular weight complexes due to the higher C93A/DNA ratio are formed and the disappearance of the free all4725 DNA is even more evident. Conversely, the status of the nonspecific nifJ promoter does not change. Analogous ITC experiments were carried out with C93A (Figure 4c,d) and data derived from the results are shown in Table 1. As it was observed with FurB, the C93A mutant exhibited only one binding site in the absence of DTT but, conversely, addition of DTT did not lead to the appearance of a second binding site in the mutant. This result supports the hypothesis that C93 could be involved in zinc coordination at one of the regulatory sites. Consequently, substitution of this residue by alanine leads to a decrease in zinc affinity and the detection of only one zinc binding event in the calorimetric experiments under reducing and non-reducing conditions.
Purification and specific DNA binding of C93A.
In order to confirm whether the location of C93 would be compatible with zinc coordination by this residue, we performed a tridimensional modelling of FurB based on solved X-ray structures of FUR proteins crystallized with three zinc ions: H. pylori Fur (Figure 6a), M. tuberculosis FurB/Zur (Figure 6b) and S. coelicolor Zur (Figure 6c). Interestingly, C93 was located in all FurB models at the beginning of an α-helix close to a flexible loop. This position could be highly flexible as predicted in particular models from H. pylori Fur and S. coelicolor Zur, allowing C93 to rotate and bind to zinc at site 3 if necessary. The location of C93 at the beginning of an α-helix in Anabaena sp. FurB was also corroborated with the PSIPRED tool for secondary structure prediction (Supplementary Figure S5).
FurB modelling based on the structure of different FUR proteins.
Influence of redox status in the specific FurB–DNA interaction
Our previous studies consistently showed that DTT was required for the optimal interaction of FUR proteins with DNA [2,6]. In order to study the influence of FurB redox state on its interaction with DNA, EMSA assays under different redox conditions were conducted. Figure 7 shows that the presence of DTT substantially favoured and stabilized complex formation (lanes 2 and 3). DTT was also able to partially revert the effect of H2O2 on FurB (lanes 4 and 5), as well as in the preformed FurB–DNA complex under oxidizing conditions, clearly indicating that the specific interaction of FurB with DNA is modulated by its redox status.
Effect of redox conditions on FurB–DNA interaction.
As previously shown in the EMSA assays with FurB, upon titration of Pall4725 with the regulator, two clearly defined FurB–DNA complexes could be observed (Figure 1b). This was in contrast with the smearing obtained when the C93A derivative was incubated with the promoter (Figure 5b), indicative of complex dissociation along the experiment. Furthermore, the mutant presented a weaker affinity for DNA when compared with FurB (Figure 5c). Taken altogether, these results indicate that FurB interaction with DNA is redox regulated through C93, since the presence of this residue in the reduced form appears to be necessary for the formation of defined protein–DNA complexes.
Study of FurB/C93A interaction with heme
As previously described, addition of heme to FurB impaired its DNA-binding ability, pointing to an interplay between FurB activity and the pool of unbound cellular Fe(II/III) heme . Since FurB contains a putative heme regulatory motif involving C93, EMSAs in the presence of heme were performed with the C93A mutant as well as with the wild-type protein (Figure 8). Addition of the ligand affected C93A binding to DNA, indicating that both FurB and C93A are able to interact with heme. Owing to the high aggregation of FurB in neutral and basic pH buffers, additional controls (lanes C) were performed so as to corroborate that DNA binding was not affected by the buffer in which heme was dissolved (50 mM Tris–HCl, pH 8). Hence, the decrease in FurB/C93A DNA-binding ability after heme addition was due to the existence of a protein–heme interaction.
EMSA assays in the presence of heme.
Differential spectroscopy experiments were then carried out with the aim of determining the affinity of the protein–heme interaction. Consecutive aliquots of FurB were added to a heme solution and differential spectra were recorded (Figure 9a). Analogous experiments were performed with C93A to establish whether C93 substitution affected affinity and to infer its implication in heme binding (Figure 9b). As derived from spectra, both FurB and C93A interacted with heme, exhibiting a shift of the 385 nm band (corresponding to the maximum absorbance of free heme) to ∼420 nm (maximum absorbance of a protein–heme complex). The ΔAbs420-385 values obtained in both experiments were represented against protein concentration and adjustment to the typical saturation curve of a 1:1 stoichiometry complex was attempted. Unfortunately, data did not lead to a proper adjustment and affinity constant for the FurB/C93A–heme interaction could not be calculated by this method (not shown).
Interaction of FurB and C93A with heme.
Determination of the affinity for heme with the ITC technique was also attempted but, again, data obtained did not allow calculation. In this case, ITC results could not be used because either the enthalpy of reaction was too small or there was no interaction under the conditions of the assay. Nevertheless, comparison of both sets of differential spectra shows that the maximum ΔAbs420-385 achieved with C93A was smaller than in the case of FurB, indicating that even though C93 is not essential for heme–FurB interaction, its substitution by an alanine results in a decreased affinity of FurB for heme. The occurrence of C93 either as a direct heme-coordinating amino acid or in its proximal environment was also supported by the presence of two cavities nearby C93 in a protein model based on the average of the structures from H. pylori Fur and M. tuberculosis Zur (Figure 9c). In the absence of the zinc ion in site 3, those two potential gateways for heme conformed by C93, H77, H91, H96 and H113, as well as by a hydrophobic pocket surrounded by Q103, F108, F111 and Y113 could certainly contribute to the formation and stabilization of the heme-protein.
The proclivity of FUR proteins to form inclusion bodies when recombinantly overexpressed makes its purification challenging. Previous purification procedures yielded low amounts of FurB not suitable for biochemical or structural studies [2,7]. In this work, we have taken advantage of the high stability of FUR proteins against chemical denaturation using mild concentrations of GdnHCl in the purification procedure to obtain a high yield of active FurB. The presence of GdnHCl has been proposed to decrease interactions favouring protein aggregation and inclusion body formation , increasing the yield of the purification process when compared with the previous attempts without denaturing reagent. GdnHCl could also work reducing the non-specific binding of proteins to the column, allowing for a higher purity of the target protein.
Purified FurB retained its ability to specifically bind DNA in vitro, corroborating proper folding of the protein. Recombinant FurB also exhibited the ability to non-specifically bind and protect DNA against DNase I cleavage. Hence, according to our previous reports [2,6,7], the results here evidence that the non-specific binding to DNA in vitro is an intrinsic ability of FurB, not depending on the purification procedure employed.
Although AFM experiments confirmed the duality in FurB binding to DNA, only protein monomers could be observed specifically bound to DNA. It is notable that the presence of protein monomers bound to DNA was also observed in FurA from Anabaena sp., which binds the PfurA target in a sequential fashion . Unlike FurA, non-specific binding of FurB to DNA starts by binding of FurB monomers along the DNA chain (Figure 2b) and only when DNA has been coated, the recruitment of additional FurB molecules to form higher aggregates is observed (Figure 2d). Notably, non-specific FurB binding does not require zinc as cofactor (Figure 2c and Supplementary Figure S3). The possible in vivo implications of the non-specific DNA-binding ability of FurB have already been discussed . However, further work should be done to identify the environmental factors that trigger massive binding to DNA, as well as the cellular pathways potentially involved in this process.
PAR-staining and ICP-MS analyses indicated that recombinant FurB contained one zinc ion per monomer of protein. PAR staining showed the presence of zinc even under SDS-denaturing conditions, pointing to the structural role of the zinc detected. As previously proposed for other FUR proteins , under denaturing conditions, residues involved in structural zinc coordination could form a hairpin-like structure that would allow metal binding (Supplementary Figure S4). Thus, we propose that this zinc ion detected in FurB would correspond to the structural one, which, in other members of this family, is tightly bound to the protein even under denaturing conditions [28,38].
As already mentioned, structural zinc coordination in FUR proteins usually takes place through the four highly conserved cysteine residues belonging to CXXC motifs in the C-terminal domain . Regarding FurB, there are four residues fulfilling those conditions: C81, C84, C121 and C124. The equivalent cysteines in B. subtilis Zur (C95, C98, C132, C135), S. coelicolor Zur (C90, C93, C130, C133), M. tuberculosis FurB/Zur (C86, C89, C126, C129) and H. pylori Fur (C102, C105, C142, C145) are involved in structural zinc coordination [11,15,16,39]. This, together with residue disposition in the different FurB models points to C81, C84, C121 and C124 as ligands to the structural zinc at site 1.
ITC experiments show that, apart from the structural site, reduced FurB presents two additional binding sites with an affinity for zinc in the micromolar range. Intriguingly, in the absence of DTT FurB exhibited a single regulatory binding site. This indicates that at least one cysteine residue should participate in the coordination of the second regulatory zinc ion, which is only bound to FurB under reducing conditions. Since C81, C84, C121 and C124 are probably involved in structural zinc coordination, C93 would be the only cysteine available. Calorimetry experiments with the C93A mutant supported the hypothesis of C93 being involved in regulatory zinc coordination, because the mutant presented only one regulatory site even under reducing conditions. FurB modelling based on different FUR X-ray crystallographic structures revealed that C93 position would be compatible with regulatory zinc coordination, specifically, at site 3. In all the models used in this work, C93 is located in the extreme of an α-helix and close to flexible regions that would allow residue rotation to co-ordinate zinc if necessary. A shortening of α-helix length and a rotation of cysteine-binding metal have been observed previously for the apoprotein to metal-binding protein transition . Interestingly, a previous bioinformatic analysis based on Zur from Prochlorococcus marinus  also locates the equivalent to C93 at the sensory site of Zur. Thus, taking all those results into account together with the comparison of EMSA profiles from FurB and the C93 variant, we suggest that FurB activity is influenced by the redox state of C93 that, in turn, determines the loading of the second regulatory zinc ion.
Regarding the other residues involved in zinc coordination at site 3, D75, H77, H96 and H113 occupy equivalent positions to the coordinating residues of site 3 in both M. tuberculosis FurB/Zur and S. coelicolor Zur , so that some of these residues could be implicated in zinc coordination at site 3. Concerning site 2, only H78 is highly conserved among FUR proteins. However, the high proximity between sites 2 and 3, only separated by a flexible loop, would make it possible for residues in the loop (D75 and H77) to intervene in zinc coordination at site 2 instead of site 3. Other residues such as Q21, Q58 and Q73 could also be involved in zinc coordination at site 2. Although coordination through glutamine residues is less frequent , metals can interact with the oxygen in the carbonyl group. Another possibility is that the coordination sphere is completed with a water molecule bound to a glutamine residue. This kind of interaction has been found in Fur from Magnetospirillum gryphiswaldense, where Q111 is involved in manganese coordination through a water molecule . Only additional mutagenesis studies together with the resolution of the X-ray crystallographic structure of zinc-replete FurB would allow determining residues involved in zinc coordination.
The effect of oxidative stress in zinc uptake, as well as the role of Zur in protecting bacteria from oxidative damage, has been established in organisms with very different physiology [6,44–46]. More recently, an interconnection between FUR proteins and thiol-mediated redox homeostasis has been reported in B. subtilis, where Zur and Spx regulons partially overlap . Furthermore, metal and DNA binding of cyanobacterial FurA is controlled by an intramolecular cysteine-based redox switch that integrates iron availability and redox environmental signals . Thus, the integration of metal availability and redox status through redox modulation of cysteine residues could be a feature of cyanobacterial FUR proteins to manage a more precise control of cell homeostasis.
The ability of FurB to bind heme has previously been described . Heme is an essential cofactor and a signal molecule in a source of biological processes. FUR–heme interaction modulates the activity of several members of this family through different mechanisms. While heme binding to Irr triggers protein degradation in rhizobia , the interaction of heme with cyanobacterial FurA and FurB results in dampening of protein activity [7,21]. In Anabaena sp., FurA displays the features of a heme-sensor, binding the cofactor through the regulatory motif CP with an affinity in the micromolar range [21,22]. In a similar way, it was proposed that the CPV (cysteine–proline–valine) motif present in FurB, which contains C93, was involved in FurB–heme interaction. However, further studies were required to verify this hypothesis. Results from the present work show that the presence of heme in a 10-fold molar excess severely affected the DNA-binding ability of FurB in vitro. A similar excess of heme is necessary for the inhibition of the activity in other FUR proteins such as FurA from Anabaena sp.  and Fur from H. pylori (Ángela Fernández-Otal, personal communication).
Attempts to calculate the affinity of the FurB–heme interaction by means of differential spectroscopy and ITC were unsuccessful, perhaps due to the liability of the interaction. In fact, a common feature of heme-containing proteins with a CPV motif is the transient character of this interaction, essential for an efficient effector-mediated regulation of protein function. Results from differential spectroscopy experiments with C93A, the FurB-derivative with a mutation in the CPV motif, indicated that C93 is probably involved in heme coordination. Whereas differential spectra from the wild-type protein presented a clear increase in the absorbance at 420 nm when protein was added to heme, the maximum absorbance at this wavelength was much lower in C93A. Therefore, as previously suggested in FurA from Anabaena sp., where C141 belonging to the CP motif was proposed as an axial ligand for the Fe(III) high-spin heme , C93 could be directly involved in heme coordination by FurB. If not directly implicated, C93 could be located in the proximities of the truly binding site, favouring heme recruitment. Histidine residues are usually involved in heme coordination  and, hence, it cannot be ruled out that heme binding in FurB could be taking place through some of the histidine residues in close proximity with C93, such as H77, H96 and H113.
Free-heme levels inside the cell increase along with oxidative stress , so that the FurB–heme interaction would make sense from a biological point of view. Heme interaction with FurB under oxidative stress would diminish protein affinity for DNA, increasing the transcription of genes such as sodA and alr0998, which are FurB-regulated and directly involved in the oxidative stress response (Figure 10) . It is also notable that all4725, whose promoter is used as a positive control in the present study because of the high affinity of FurB by its sequence, codes for the porphobilinogen synthase hemE, a key enzyme in the heme-synthesis pathway. Interestingly, all4725 is strongly de-repressed in a Δzur mutant , supporting the occurrence of a FurB–heme regulatory interplay.
Role of C93 in the relationship between oxidative stress, ligand binding and the regulation of target genes by FurB.
In summary, our data indicate that Anabaena sp. FurB, in addition to controlling zinc homeostasis and the oxidative stress response through direct gene repression, is a redox-responsive protein based on the status of C93 (Figure 10). This residue seems to play a critical role in FurB activity, contributing to heme recruitment and/or binding and tuning FurB interaction with zinc through its redox state which, in turn, determines the affinity of FurB for DNA. The results reported here establish, for the first time, a relationship between the redox status of a FUR protein, the functionality of a metal-binding site and their contribution to optimal DNA binding. These properties, together with the ability of some FUR paralogues to bind heme, conform an additional layer of regulation of these multifunctional proteins.
atomic force microscopy
- DNase I
electrophoretic mobility shift assay
ferric uptake regulator
inductively coupled plasma mass spectrometry
isothermal titration calorimetry
manganese uptake regulator
nickel uptake regulator
promoter region of the all4725 gene
Protein Data Bank
promoter region of the furA gene
promoter region of the nifJ gene
zinc uptake regulator
V.C.S.-E. planned, performed experiments and wrote the manuscript; M.C.P. and A.L. performed AFM experiments; I.Y. performed protein modelling; A.V.-C. analysed ITC results; M.L.P. and M.F.F. planned experiments and wrote the manuscript.
This work has been supported by grants B18 from Gobierno de Aragón, BFU2012-31458/FEDER and BFU2016-77671-P/FEDER from MINECO and UZ2016-BIO-02 from the University of Zaragoza. V.C.S.-E. was recipient of a fellowship from Gobierno de Aragón.
Authors thank Drs Andrés González, Ignacio Luque and Teresa Bes for the helpful discussion and critical reading of the manuscript. They also thank Sandra Sampériz for the assistance during protein purification, Dr Olga Abián for help during ITC data analysis and to Iñigo Echániz and José Luis Diez for their technical help.
The Authors declare that there are no competing interests associated with the manuscript.