A Novel DNA-binding Site for the Ferric Uptake Regulator (Fur) Protein from Bradyrhizobium japonicum*

The Fur protein is a global regulator of iron metabolism and other processes in many bacterial species. A key feature of the model of Fur function is the recognition of a DNA element within target promoters with similarity to a 19-bp AT-rich palindromic sequence called a Fur box. The irr gene from Bradyrhizobium japonicum is under the control of Fur. Here, we provide evidence that B. japonicum Fur (BjFur) binds to the irr gene promoter with high affinity despite the absence of DNA sequence similarity to the Fur box consensus. Both Escherichia coli Fur and BjFur bound a synthetic Fur box consensus DNA element in electrophoretic gel mobility shift assays, but only BjFur bound the irr promoter. BjFur maximally protected a 30-bp region in DNase I footprinting analysis that includes three imperfect direct repeat hexamers. BjFur formed a high mobility complex and a low mobility complex with DNA in electrophoretic gel mobility shift assays corresponding to occupancy by a single dimer and two dimers or a tetramer, respectively. A mutation in the downstream direct repeat DNA sequence allowed high mobility complex formation only. In vitro transcription from the wild type irr promoter or from a mutated promoter that allowed only dimer occupancy was repressed by Fur, indicating that the dimer can be a functional repressor unit. Our findings identify a novel DNA-binding element for Fur and suggest that the Fur box consensus may not completely represent the target sequences for bacterial Fur proteins as a whole. In addition, Fur binding to a target promoter is sufficient to repress transcription in vitro.

Control of iron homeostasis is essential to most living organisms. Iron is required for many cellular processes but can be unavailable because it is insoluble in its predominant ferric (Fe 3ϩ ) form. However, excess iron is toxic because it catalyzes the formation of reactive oxygen species that can damage DNA, protein, and lipids. Organisms have specific mechanisms to control iron acquisition and to store it in an inert form. In bacteria, studies on the control of iron homeostasis have focused largely on Fur (ferric uptake regulator), a regulatory protein that responds to cellular iron. Fur represses genes that are involved in high affinity iron transport under iron replete conditions and that are derepressed when the metal is scarce. In addition, Fur is involved in numerous other facets of iron metabolism and also in processes not obviously linked to iron, such as acid shock response (1), synthetic pathways (2), and the production of toxins and other virulence factors (3). Fur is the founding member of a family of regulators that also includes Zur (4,5), PerR (6), and Irr (7,8). These proteins differ in function and have different DNA-binding sites but are all involved in metal-dependent control of gene expression.
Fur homologs are found in many bacterial genomes, but the protein has been studied in relatively few of them. Structural analysis of Fur and its DNA binding properties has been most extensively studied in Escherichia coli, Pseudomonas aeruginosa, and Bacillus subtilis, whereas analyses of fur mutants and the identification of genes under Fur control have also been studied in those bacteria and in several other organisms as well. P. aeruginosa Fur was crystallized as a dimer (9), which also appears to be its oligomerization state in solution (10,11). It has been proposed that at least two Fur dimers occupy its target promoter based on the size of protected DNA in footprinting analyses (12). In addition, Fur binds two zinc atoms/ monomer; one zinc has a structural function, and the other may be the ferrous iron-binding site in vivo (9). However, the E. coli protein binds DNA when one or both metal sites is occupied by zinc (13). Furthermore, cysteines are zinc ligands in E. coli (13,14) but not in P. aeruginosa (9).
The working model for Fur function posits that, when bound by ferrous (Fe 2ϩ ) iron, Fur binds its target DNA within the promoter of the regulated gene to repress transcription. However, when iron is limiting in the cell, Fur protein is unbound by iron and no longer binds DNA with high affinity, hence gene expression is derepressed. It is generally assumed that Fur binding blocks access of RNA polymerase to the promoter to repress transcription, but this has not been demonstrated directly, nor have the possibilities that Fur excludes an activator or recruits a repressor been ruled out.
Central to the model is the so-called Fur box, a DNA-binding element for Fur that contains similarity to a 19-bp, AT-rich palindromic consensus sequence (Fig. 1). This consensus sequence was originally derived from examination of promoters of numerous Fur-regulated genes (12). Subsequent searches have yielded one promoter that matches the Fur box consensus exactly (15), with 14-or 15-bp matches out of 19 being more typical and 11 bp as a minimum match (15)(16)(17). Sequence similarity to a Fur box consensus within promoter regions of genes is taken as ab initio evidence for regulation by Fur. The assumption that Fur recognizes the two 9-bp inverted repeats of the palindrome was challenged by studies that interpret the consensus as three shorter hexameric repeats in a head-tohead-to-tail orientation (18) (Fig. 1). Yet a third interpretation of the consensus using B. subtilis Fur is a two 7-1-7 inverted repeat motif that accommodates two dimers (19). Thus, there is no discrepancy in the sequence of the Fur-binding element consensus in those studies but rather in the interpretation of the functional pattern within that sequence.
Bradyrhizobium japonicum is a Gram-negative bacterium that lives as a free-living organism or in symbiosis with soybean. The B. japonicum fur gene was identified based on functional complementation of an E. coli mutant (20). It has also been characterized in Rhizobium leguminosarum (21), and homologs are found in the genomes of other taxonomically related organisms within the ␣-proteobacterial group as well. Fur is involved in controlling iron metabolism in B. japonicum (20,22) but appears to play a lesser role in R. leguminosarum (21). In addition to Fur, the Irr protein in B. japonicum is involved in iron metabolism, where it mediates iron-dependent regulation of heme biosynthesis (7). The irr gene is controlled by iron at both transcriptional (22) and post-translational levels (8,23). Evidence for Fur-mediated transcriptional control is based on the observations that iron-dependent accumulation of irr mRNA is aberrant in a fur mutant, and extracts from E. coli cells harboring the B. japonicum fur gene bind to DNA upstream of the irr gene (22). However, the irr promoter region does not contain DNA sequence similar to the Fur box consensus, indicating an important difference between the B. japonicum Fur protein and the other proteins on which the Fur box consensus sequence is based. In this study, we provide evidence for a novel DNA binding activity for B. japonicum Fur in the irr gene promoter, suggesting that the Fur box consensus may not completely represent the target sequence for bacterial Fur proteins as a whole. Furthermore, it is shown that Fur binding to the irr gene promoter is sufficient to repress transcription in vitro.
Bacterial Strains, Plasmids, Media, and Growth-E. coli strain DH5␣ was used for propagation of plasmids and was grown at 37°C on LB media with appropriate antibiotics. Plasmid pSKiron contains the 19-mer Fur box consensus sequence 5Ј-GATAATGATAATCATTATC-3Ј cloned into the ClaI site of pBluescript SK and was a gift from Dr. M. L. Vasil (24). pET14bjFur (Novagen, Madison, WI) contains the B. japonicum fur gene cloned into the NdeI and BamHI sites of pET14b. B. japonicum fur was derived from parental strain USDA I110 (20). pET14ecFur contains the E. coli fur gene cloned into the NdeI and BamHI sites of pET14b. The resultant protein expressed from pET14b contains a histidine tag that is used for purification.
Overexpression and Purification of Fur-Fur was overexpressed in E. coli strain BL21(DE3)(plysS) cells containing fur in pET14b, initially grown on LB media containing chloramphenicol (25 g/ml) and ampicillin (200 g/ml). Cells were inoculated from an overnight culture grown in LB media containing chloramphenicol and ampicillin into 1 liter of fresh 2ϫ YT containing antibiotics, 50 M FeCl 3 , and 1 mM ZnCl 2 . Overexpression was induced in cells at mid-log phase by adding 20 ml/liter of 95% ethanol and a final concentration of 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside, and then incubating at 20°C for 4 h with shaking. Cells were harvested by centrifugation at 4000 ϫ g, washed in TNG (50 mM Tris-HCl, 50 mM NaCl, 5% glycerol, pH 7.4), and resuspended in 15 ml of phosphate binding buffer (5 mM imidazole, 300 mM NaCl, 10% glycerol, 50 mM NaPO 4 , pH 8.0), 10 mM phenylmethylsulfonyl fluoride, and 25.5 g of aprotinin/5 g of cells. Cells were disrupted by passage twice through a French pressure cell at 1200 p.s.i. and clarified by centrifugation at 37,000 ϫ g for 45 min. 1 ml of 50% Ni-NTA 1 slurry (Qiagen Inc., Valencia, CA) was added to 4 ml of clear lysate and rocked for 60 min at 4°C. The Ni-NTA slurry-protein mixture was poured into a column and washed three times with 25 ml of phosphate wash buffer (20 mM imidazole, 300 mM NaCl, 10% glycerol, 50 mM NaPO 4 , pH 8). Purified His-Fur was eluted with phosphate elution buffer (250 mM imidazole, 300 mM NaCl, 10% glycerol, 50 mM NaPO 4 , pH 8). The His tag was cleaved by adding 3 units of thrombin (Novagen, Madison, WI) per 2 mg of His-Fur and incubating at room temperature for 2 h. To remove imidazole, Fur was further purified by retention through a 3-kDa nominal molecular weight limit Ultrafree column (Millipore, Billercia, MA). The His tag was removed by resuspending the retentate in 1.2 ml of 50% Ni-NTA/300 l of purified Fur and loading on a column. The eluate was collected in fractions, and fractions containing purified Fur were combined. Protein was diluted to a 1 M working stock solution in electrophoretic gel mobility shift (EMSA) binding buffer (10 mM bis-Tris borate, 1 mM MgCl 2 , 40 mM KCl, 5% glycerol, 0.1% Nonidet P-40, 1 mM dithiothreitol, pH 7.5).
Electrophoretic Mobility Shift Assay-EMSAs were used to determine Fur binding to DNA. The presence of manganese was necessary for DNA binding activity of Fur. Mn 2ϩ presumably substitutes for Fe 2ϩ , but the latter is readily oxidized to Fe 3ϩ in air. Using a protocol modified from de Lorenzo et al. (25), Fur was incubated for 30 min at 4°C in a 20-l volume of EMSA binding buffer supplemented with 50 ng of herring sperm DNA, 2 g of bovine serum albumin, 100 M MnCl 2 , and radiolabeled DNA probe. 100 pM DNA probe was used for K d determination and competition experiments, 1 nM DNA probe was used for all other experiments. Double-stranded DNA probes were produced by boiling and slowly cooling synthetic DNA oligonucleotides (Integrated DNA Technologies, Coralville, IA) in annealing buffer (150 mM NaCl 2 , 10 mM Tris-HCl, 1 mM EDTA, pH 8.0) and filled in with [␣-32 P]dCTP (3000 Ci/mmol) with the klenow fragment of DNA polymerase (Promega, Madison, WI). Following incubation, EMSA reactions were analyzed on 5% nondenaturing polyacrylamide gels in electrophoresis buffer (20 mM bis-Tris borate, pH 7.5) that were prerun for 30 min at 200 V of constant voltage. After electrophoresis at 4°C for 45-90 min at 200 V, gels were dried and autoradiographed. The Fur box probe was isolated as an 83-bp fragment from a NotI and XhoI digestion of pSKIron. A BanII and HpaI digest of pSKSBIrr was used to isolate a 63-bp fragment containing the irr promoter. A NotI and XhoI digestion of pBluescript SK (Stratagene, La Jolla, CA) was used to isolate a 69-bp fragment used as nonspecific DNA. Autoradiograms were developed on BioMax film (Eastman Kodak Co.) and scanned using a GS-700 densitometer (Bio-Rad), and signal intensities were detected and quantified using Quantity One (Bio-Rad). To determine the dissociation binding constant (K d ), binding reactions were titrated with various concentrations of Fur. Bound and unbound DNA was quantified by comparing relative signal intensities and analyzed using Graphpad Prism (Graphpad Software Inc., San Diego, CA). DNA probes used to delimit sequence sufficient for Fur binding in the irr promoter were initially synthesized as overlapping fragments and compared by EMSA. After identifying a 39-bp probe that was sufficient for both low and high mobility complex formation, base substitutions using nonspecific sequence were used to progressively alter the 5Ј and 3Ј ends of the probe, further refining small DNA regions sufficient for formation of each individual complex and both complexes.
Competition Assays-The relative affinity of B. japonicum Fur for Fur box and irr promoter DNA was determined by competition of one of the radiolabeled DNAs with the other unlabeled DNA in EMSA analysis. In the first case, 100 pM 32 P-labeled irr promoter DNA was incubated with 50 nM BjFur in the binding reaction along with 1-100-fold excess of irr promoter DNA, Fur box DNA, or nonspecific DNA. In the second case, the experiment was carried out in the same way except that the Fur box was radiolabeled.
DNase I Footprint Analysis-DNase I footprint analyses examined the DNA region protected by Fur binding. Fur was incubated for 30 min at 4°C in a 50-l volume EMSA binding buffer containing 125 ng of herring sperm DNA, 5 g of bovine serum albumin, 100 M MnCl 2 , and 1 nM radiolabeled irr promoter DNA. The DNA was labeled at one end with [␣-32 P]dCTP and the klenow fragment of DNA polymerase. 50 l of room temperature solution of 5 mM CaCl 2 and 10 mM MgCl 2 was added after 1 min followed by the addition of 0.45 units of RQ1 RNase-free DNase (Promega) in 18 l of 40 mM Tris-HCl (pH 7.0). The reaction was incubated for 2 min at room temperature. Reactions were stopped by addition of 90 l of stop solution (200 mM NaCl, 30 mM EDTA, 1% SDS, 100 g/ml yeast RNA). DNA was extracted with phenol:choloroform (1:1) followed by ethanol precipitation. G ϩ A ladders of the labeled DNA were produced as described (26). Digested probe products were separated on a 15% denaturing polyacrylamide gel containing 7 M urea in Tris borate EDTA electrophoresis buffer. Autoradiograms were developed on BioMax film. The DNA probe used for DNase I footprint analysis was radiolabeled at the 3Ј end; therefore, the top of the gel represents the 5Ј end.
Determination of the Mass of Fur-DNA Complexes-An EMSA-based method for determining the molecular weight of protein-DNA complexes was carried out as described previously (27). EMSA reactions as described above were run on nondenaturing gels with protein standards. Using autoradiography of Coomassie-stained gels, logarithms of relative mobility (R m ) for EMSA complexes and native proteins were plotted against acrylamide concentration, showing the relationship of the mobility of each species and gel concentration. The negative slope of R m for each protein standard was plotted against molecular weight to generate a standard curve from which the estimated molecular weight of each EMSA complex was interpolated.
In Vitro Transcription Assays-To directly examine the influence of B. japonicum Fur on transcription from the irr promoter, in vitro transcription assays were performed on native and altered irr promoter DNA. irr promoter fragments were generated by PCR amplification of irr sequence in pSKSBIrr using primers 5Ј-TTTGAATTCGTGAC-CAAAATCGGCTAAG-3Ј and 5Ј-TTTCTGCAGGGACGTCGTCGTCGT-GAT-3Ј, producing a 272-bp DNA fragment that terminated 80 bp downstream of the irr transcription start site determined previously (22). Similar PCR reactions were performed on pSKSBIrr derivatives with four base pair alterations that independently affected formation of the high mobility complex or low mobility complex but did not mutate the putative Ϫ35 box. pSKSBIrr derivatives were produced using the QuikChange site-directed mutagenesis kit (Stratagene).

B. japonicum Fur Binds to the irr Promoter-Previous work
shows that the iron-dependent control of irr mRNA expression is aberrant in a fur mutant and that E. coli extracts that overexpress B. japonicum Fur (BjFur) bind to DNA corresponding to the irr gene upstream region (22). However, a Fur boxlike element was not found in this region. To further study this interaction, we overexpressed and purified recombinant Fur proteins from E. coli (EcFur) and B. japonicum and examined their DNA binding activities by EMSA (Fig. 2). Recombinant BjFur bound to a 63-bp DNA fragment corresponding to the irr upstream region. EcFur, however, did not bind to this DNA; 50 nM BjFur was sufficient to bind all of the DNA in the EMSA assay (Fig. 2), but up to 150 nM EcFur did not bind the irr gene promoter (data not shown). The B. japonicum fur gene complements an E. coli mutant (20). Consistent with that observation, BjFur bound to the Fur box consensus sequence, as did EcFur (Fig. 2). Thus, BjFur has a unique DNA binding activity in addition to that found in the E. coli protein. This activity presumably allows BjFur to regulate the irr gene despite the absence of a Fur box consensus element in the irr promoter.
The affinity of BjFur for the irr promoter and Fur box consensus DNA was assessed by measuring bound DNA as a function of BjFur concentration in an EMSA assay (Fig. 3A). The dissociation binding constants (K d ) for the two DNA elements were nearly identical and estimated to be 5.1 and 4.4 nM for the irr promoter and the Fur box, respectively. The relative affinities of the two DNA elements for BjFur were also assessed by competition EMSA analysis (Fig. 3, B and C). Binding of BjFur to radiolabeled irr promoter DNA was competed out by unlabeled Fur box DNA as effectively as by unlabeled irr promoter DNA (Fig. 3B). Similar results were obtained when the experiment was repeated using radiolabeled Fur box DNA (Fig.  3C). Thus, the BjFur DNA-binding regions for each element are the same or overlapping. The data indicate that binding of BjFur to the irr promoter is similar to its binding to a known Fur target element, and therefore, it is likely to be physiologically relevant. These findings suggest that control of irr gene by Fur involves direct binding of Fur to the irr promoter.

The BjFur-binding Region of the irr Promoter Contains Three Imperfect Direct Repeat Sequences Dissimilar from the Fur Box
Consensus-The BjFur-binding region in the irr promoter was further defined by DNase footprinting (Fig. 4). The promoter was maximally occupied at 10 nM BjFur, resulting in approximately a 30-bp protected region. EMSA analysis using various double-stranded DNA oligonucleotides representing different regions of the irr promoter also delimited the BjFur binding to this region (see below). The DNA binding region was found to extend from Ϫ48 to Ϫ19 with respect to the transcription start site (ϩ1) and contains three imperfect direct repeat hexameric sequences (Fig. 5). Two tandem repeats of TGCATC are preceded by a TGCGAG separated by 5 bp. The BjFur-binding site on the irr promoter has only 7 of 19 matches to the Fur box consensus (on the strand complementary to that shown in Fig.  5). The probability for a random match of 7 out of 19 is very high (0.1 based on calculations from Ref. 28). Furthermore, whereas the Fur box consensus can be interpreted as inverted repeat or direct repeat DNA, the BjFur-binding region of the irr promoter can only be interpreted as direct repeat DNA. From this, we conclude that binding of BjFur to DNA does not require the recognition of inverted repeat sequences.
Binding of Fur to the irr Promoter Yields Two Complexes-Titration of irr promoter DNA with BjFur revealed two binding species in EMSA analysis (Fig. 6). At low BjFur concentrations, a high mobility complex (HMC) was predominant, whereas a low mobility complex (LMC) was prevalent at 50 nM protein.
The footprinting analysis indicated a concentration-dependent occupancy of the irr promoter by BjFur (Fig. 4), and the two EMSA complexes are consistent with that observation.
EMSA analysis was used to delimit the BjFur site on the irr promoter by first narrowing down the binding region to a 39-bp DNA fragment (Fig. 7). Subsequently, sequences on the left or right side of the 39-bp DNA were substituted with nonspecific sequence to keep the overall length constant. In addition, 5 and 50 nM BjFur were used for each DNA tested, which allows detection of the HMC and LMC, respectively. A fragment containing 29 bp of irr promoter DNA corresponding to the protected region in footprints gave both the HMC and LMC in the EMSA analysis (Fig. 7). Substitutions from the right side that eliminated some or all of the downstream direct repeat (DR3), but which kept DR1 and DR2 intact, resulted in loss of the LMC. However, the HMC was formed with that DNA and was observed at both 5 and 50 nM BjFur (Fig. 7B). Thus, DNA containing those two direct repeats was sufficient for formation of the smaller complex. Substitutions from the left side that eliminated some or all of DR1, leaving DR2 and DR3 intact, abrogated formation of the HMC. Thus, DR2 and DR3 are not equivalent to DR1 and DR2 in terms of binding BjFur, which may be due to differences in sequence or spacing between the repeats. Surprisingly, the LMC was observed with the DR2-DR3 fragment. However, a significant amount of DNA remained unbound with 50 nM protein, indicating that the affin-  Fig. 4. The arrows denote the three direct repeat sequences. Underlined sequences indicate the putative Ϫ35 and Ϫ10 regions. The ϩ1 position is the transcription start site. HMC and LMC denote the DNA regions necessary to form a high mobility complex and low mobility complex, respectively, as described in Fig. 6 and under "Results." ity of BjFur for the mutated DNA was less than that for the wild type (Fig. 7B).
We wanted to determine the occupancy of the irr promoter by BjFur in each of the DNA-protein complexes. The acquisition of DNAs that formed only an HMC or LMC allowed the measurement of the apparent molecular weight of the protein-DNA complexes in native PAGE using gels containing different polyacrylamide concentrations (Fig. 8) (see "Experimental Procedures"). When compared with globular protein standards, the apparent molecular mass of the HMC was 56 kDa, which is close to the 60 kDa predicted for a BjFur dimer bound to a 39-bp DNA fragment. The LMC had an apparent molecular mass of 94 kDa, which is in good agreement with the 98 kDa expected of two BjFur dimers or a tetramer bound to a 39-bp DNA fragment. We suggest that BjFur can occupy the irr promoter as one or two dimers.
Binding of a BjFur Dimer to the irr Promoter Is Sufficient to Block Transcription in Vitro-Fur is presumed to repress gene expression by binding the promoter of target genes and blocking transcription (12). If binding is sufficient, then it should be possible for Fur to block transcription in vitro in the absence of additional regulatory factors. We developed an assay for transcription of a double-stranded DNA template initiated from the irr promoter using commercial E. coli RNA polymerase holoenzyme (see "Experimental Procedures"). An RNA doublet of 79 bp and 80 nucleotides in size was synthesized from the irr promoter (Fig. 9). This synthesis was abolished in the presence of 10 nM BjFur, indicating that Fur is sufficient to block transcription. The BjFur-binding region includes the putative Ϫ35 region of the irr promoter; thus, BjFur binding may preclude access by RNA polymerase.
Current models of Fur envision two dimers docked to DNA (9). Because BjFur can bind the irr promoter as one or two FIG. 6. BjFur forms high and low mobility complexes with irr promoter DNA. 32 P-labeled irr promoter DNA was titrated with various concentrations of BjFur in binding reactions, and complexes were resolved and visualized by EMSA as described in the legend for Fig. 2. HMC and LMC denote the high mobility complex and low mobility complex, respectively, formed in a concentration-dependent manner. Free, no protein. dimers, we wanted to address which complex was able to block transcription. Mutations in the irr promoter were designed that allowed only the HMC or the LMC to form but did not inactivate the promoter. A 4-bp substitution within direct repeat 3 (DR3) abrogated the LMC in the EMSA analysis, but formation of the HMC was still observed (Fig. 9, A and B). Conversely, a 4-bp substitution within DR1 eliminated the HMC, but the LMC was observed with 50 nM BjFur. Transcription was initiated from both of the mutant promoters, and similar amounts of RNA accumulated in the 20-min reaction when compared with the wild type (Fig. 9C). Addition of 10 nM BjFur completely inhibited transcription from the HMC promoter, similar to what was observed for the unmutated promoter. This concentration of protein did not substantially affect transcription from the LMC promoter, but it was inhibited by 25 nM protein and abolished completely at 100 nM Fur (Fig. 9C). This observation is consistent with the lower affinity of BjFur for the LMC promoter. We suggest that binding of the BjFur dimer to the irr promoter is sufficient to repress transcription. DISCUSSION Fur homologs are found throughout the eubacterial kingdom. A generalized model for Fur function posits that the protein binds ferrous iron when the metal is available, conferring the ability to recognize and bind to a Fur box DNA element to repress transcription of target genes (12). This model is based largely on work with Fur proteins from the taxonomically related organisms E. coli, P. aeruginosa, and Vibrio sp., but its general applicability is reinforced to some extent by work with the dissimilar organism B. subtilis. With respect to DNA recognition, the recent controversy has not been with the Fur element sequence but rather with the interpretation of the consensus as inverted repeat or direct repeat sequence (18,19). In the present study, we demonstrate that B. japonicum Fur recognizes a DNA element within the irr gene promoter that is dissimilar from the Fur box consensus sequence. The affinity of BjFur for the irr promoter was similar to that for the Fur box consensus, and irr mRNA accumulation is controlled by Fur (22). Thus, the DNA-protein interaction defined herein is physiologically relevant. The best alignment of the Fur box consensus to the 30-bp protected region of the irr promoter is only 7 of 19 residues and, accordingly, the element is not recognized by Fur from E. coli. This low match is predicted to occur with very high frequency (ϳ9 ϫ 10 5 sites/strand for a genome of 9 ϫ 10 6 bp), and thus, the Fur box cannot be the basis of this binding site. Genes whose expression is affected in a fur mutant, but lack a Fur box element in their putative promoters, have been identified by microarray and proteomic analyses (29), but in those cases, Fur has not been shown to bind those promoters, and the effect of Fur may be indirect. To our knowledge, this is the first example of a DNA binding activity for a Fur protein that is disparate from recognition of the Fur box consensus. The findings suggest that a generalized model for Fur with respect to its target DNA needs to be expanded or modified. In addition, searches for Fur-regulated genes in genomes based on homology to an upstream Fur box consensus may exclude desired genes in organisms where Fur has not been characterized.
Whereas the Fur box consensus can be interpreted as inverted repeat or direct repeat sequence, the BjFur DNA-binding element in the irr promoter contains only direct repeat sequence. Therefore, a requirement for inverted repeat DNA for recognition by BjFur can be unequivocally ruled out. BjFur binds to both the Fur box consensus and the irr promoter element, and both elements contain three imperfect direct repeat hexamers. The simplest extrapolation from these findings is that BjFur recognizes direct repeat sequences in the Fur box as well as in the irr promoter element.
Although the novel DNA binding activity of BjFur must have a structural basis, there are no obvious differences between the B. japonicum protein and Fur from other organisms. Phylogenetic analysis (ClustalW; (30)) of numerous Fur homologs shows that the proteins cluster according to taxonomy, and the B. japonicum protein is not an outlier (data not shown). It is likely that the structural features of BjFur conferring the novel DNA binding are subtle and may be elucidated by mutational analysis and identification of other Fur proteins with that activity.
Fur proteins are likely to be dimeric in solution (10,11), and the P. aeruginosa protein was crystallized as a dimer (9). A model for Fur based on the crystal structure of the unbound protein depicts two dimers binding to target DNA. This model accommodates the observation that Fur protects at least 30 bp of DNA in DNase I footprinting experiments, whereas a dimer should protect only ϳ20 bp (9). In that model, Fur recognizes three hexamers, with the central hexamer region in contact with both dimers. In the present study, BjFur maximally protected 30 bp of the irr promoter at a protein concentration that formed a low mobility complex comprising two dimers or a tetramer. These findings are in agreement with the proposed model based on the P. aeruginosa Fur structure. In addition, a high mobility complex comprising a BjFur dimer was predom- FIG. 9. Repression of transcription from wild type (wt) and mutant irr promoters by BjFur. As shown in A, the underlined sequence CGAG within direct repeat 1 was replaced with TCTA. That altered DNA yielded only the LMC in EMSA analysis. The underlined sequence GCAT within direct repeat 3 was changed to AACA, resulting in formation of only the HMC. B, EMSA analysis of mutants described in panel A. 0 (free), 5 nM, or 50 nM BjFur was used to detect the HMC and LMC. C, in vitro transcription from the wild type and mutant promoters in the presence of 0, 25 nM, or 100 nM BjFur. Transcription of the DNA templates was carried out using 0.5 units of commercial E. coli RNA polymerase holoenzyme as described under ''Experimental Procedures.'' The products were 79 and 80 bp in size. inant in EMSA analysis at low protein concentrations and was the only complex observed at any concentration when DR3 was mutated (Figs. 7 and 9B). Thus, the irr promoter can be occupied by a dimer, and DR1 and DR2 appear to be sufficient for high affinity binding. However, DNA in which DR1 was altered, leaving DR2 an DR3 intact, did not yield a dimer, and therefore, the sequence or spacing of the repeats may be important since DR1 has a different sequence than the other repeats and is separated from DR2 by 5 bp. We suggest that the wild type promoter is occupied by a dimer at low BjFur concentrations and makes contact with DR1 and DR2. At higher BjFur concentrations, a second dimer binds to DR2 and DR3, and perhaps the second dimer is also stabilized by interaction with the first dimer. Thus, titration of DNA with Fur results in loss of the dimer in EMSA (Fig. 6) and additional protection of DNA in footprinting analysis (Fig. 4). Mutation of DR1 abrogates high affinity binding, but there may be sufficient DNAprotein and protein-protein interactions to form the low mobility complex at higher BjFur concentrations.
Fur is presumed to repress gene expression by binding to the target promoter and blocking transcription initiation. However, to our knowledge, the possibility that Fur repression involves the recruitment or exclusion of other regulatory factors has not been addressed. Here, we show that Fur was sufficient to repress transcription from the irr promoter in vitro. Furthermore, transcription from a mutant promoter that allowed only dimer occupancy was repressed by Fur as well, suggesting that a single dimer may be functional as a repressor and that the DNA containing two direct repeats can be a negative regulatory element.