Architecture of a Fur Binding Site: a Comparative Analysis

Abstract

Fur is an iron-binding transcriptional repressor that recognizes a 19-bp consensus site of the sequence 5′-GATAATGATAATCATTATC-3′. This site can be defined as three adjacent hexamers of the sequence 5′-GATAAT-3′, with the third being slightly imperfect (an F-F-F configuration), or as two hexamers in the forward orientation separated by one base pair from a third hexamer in the reverse orientation (an F-F-x-R configuration). Although Fur can bind synthetic DNA sequences containing the F-F-F arrangement, most natural binding sites are variations of the F-F-x-R arrangement. The studies presented here compared the ability of Fur to recognize synthetic DNA sequences containing two to four adjacent hexamers with binding to sequences containing variations of the F-F-x-R arrangement (including natural operator sequences from the entS and fepB promoter regions of Escherichia coli). Gel retardation assays showed that the F-F-x-R architecture was necessary for high-affinity Fur-DNA interactions and that contiguous hexamers were not recognized as effectively. In addition, the stoichiometry of Fur at each binding site was determined, showing that Fur interacted with its minimal 19-bp binding site as two overlapping dimers. These data confirm the proposed overlapping-dimer binding model, where the unit of interaction with a single Fur dimer is two inverted hexamers separated by a C:G base pair, with two overlapping units comprising the 19-bp consensus binding site required for the high-affinity interaction with two Fur dimers.

Under high-iron conditions, the ferric uptake regulator (Fur) of Escherichia coli represses the transcription of many genes, including those for iron acquisition systems (4), oxidative stress responses (31, 42), metabolic pathways (40), and various virulence factors (17, 27). Fur is a dimer in solution (11, 29), with each 17-kDa monomer containing two metal binding sites: one for a structural Zn(II) ion, and the other for the reversible association of Fe(II) under high-iron conditions (1, 3, 24). This association with Fe(II) repositions the N-terminal domain in an open configuration that associates with DNA at a 19-bp consensus site with the sequence 5′-GATAATGATAATCATTATC-3′ (8, 19, 40). The consensus was derived from the alignment of multiple Fur-regulated operators (12, 20), many selected by Fur titration assays (40, 44) or cycle selection (32). Fur binding has been examined in detail with a variety of footprinting techniques (9, 12, 14, 15, 20, 23, 25), spectroscopic analyses (1, 11, 19, 24, 29, 36), and atomic force microscopy (17).

Originally, it was proposed that Fur recognized this sequence as 9-bp inverted repeats separated by a single base pair (classical model; Fig. 1A) (4); however, more recent data suggested that the consensus is recognized as three adjacent hexameric units of the sequence 5′-GATAAT-3′ (hexamer model; Fig. 1A) (14, 15). The ability of Fur to bind synthetic DNA fragments containing one to five adjacent hexamers in the same orientation was tested with several in vitro assays, including gel retardation, DNase I and hydroxyl radical footprinting, and “missing T” contacts. These studies showed that a minimum of three adjacent hexamers were required for initial binding and that additional hexamers increased the affinity of Fur for the sequence (14). The consensus Fur binding sequence actually contains two hexamers in the forward orientation separated by a single base pair from a third hexamer in the reverse orientation; this arrangement was also tested in the hexamer studies, though not as thoroughly.

FIG. 1.

Fur binding site. A shows the 19-bp consensus Fur binding site for E. coli and various models of recognition. The top sequence shows the consensus Fur binding site. The lower left sequence demonstrates the classical model of binding, as described in the introduction. The classical model shows each monomer binding a 9-bp inverted repeat (shown as arrows) of the consensus, with an A:T base pair in between. The lower right sequence depicts the hexamer model, also described in the introduction, with the unit of recognition being the sequence 5′-GATAAT-3′ (shown as arrows). It is uncertain how Fur would bind this sequence; some have suggested that each hexamer is recognized by a single dimer. B shows the binding sites tested in this study.

A new model for Fur-DNA interactions was recently proposed, based on in vitro binding studies between purified Fur and wild-type and mutated forms of the fepD-entS promoter region in E. coli. These data indicated that the 19-bp consensus is organized as overlapping 13-mer sequences, with each 13-mer containing inverted 5′-GATAAT-3′ hexamers interrupted by a single nucleotide (25). In this overlapping-dimer binding model, Fur is predicted to interact with this site as two overlapping dimers positioned on opposite faces of the DNA helix. This interaction is similar to that seen with DtxR, an iron-responsive transcriptional regulator in gram-positive bacteria with a function homologous to that of Fur (6, 38, 39). DtxR and Fur do not have significant amino acid sequence homology, but their predicted secondary structures are very similar, especially in the N-terminal DNA binding domains (19).

Fur, like DtxR, contains a winged-helix motif for DNA binding, a motif found in over 50 transcriptional regulators in both eukaryotes and prokaryotes (28). This motif consists of three α-helices followed by an antiparallel β-sheet, with the helices interacting with specific DNA bases in the major groove of the binding site and the β-sheet (forming a “wing”) making secondary contacts with DNA in the minor groove. This family of regulators includes MarR, OmpR, catabolite activator protein, and LexA in bacteria and yeast heat shock transcription factor, hepatocyte nuclear factor 3γ, and E2F in eukaryotes (2, 10, 18, 22, 28). Additional evidence suggested that the DNA binding domains of Fur and DtxR undergo similar conformational changes upon metal binding, causing the N-terminal region to be repositioned to allow association with its DNA target (19). The proteins showed similar footprint patterns, with protection of a 30- to 35-bp sequence centered around 19-bp consensus sequences (26, 41). X-ray crystallography of DtxR bound to DNA showed that it binds its 19-bp consensus as two overlapping dimers on opposite faces of the helix (34, 45). The overlapping-dimer binding model also explains how Fur, a rigid protein (36), was able to make contacts on both sides of the DNA helix (25).

A comparative analysis is needed to distinguish Fur-DNA interactions with respect to the distinct architectures suggested by the hexamer model, which predicts that three tandem direct hexamers form the initial binding site, and most natural operators, which contain two forward hexamers separated by a single base pair from a third hexamer in the reverse orientation. In this study, gel retardation assays showed that the natural configuration is necessary for high-affinity Fur-DNA interactions, since the affinity of Fur for three or four adjacent hexamers in the same orientation is significantly less than that of both synthetic and natural sequences with hexamers arranged as two in the forward orientation separated by one base pair from one or two hexamers in the reverse orientation. The single-base-pair spacer (usually a C:G pair) is shown to be critical, as its absence led to significantly reduced affinity for the sequence. These data, as well as a recent study that examined binding properties of the Fur species from Bacillus subtilis (5), confirm various aspects of the overlapping-dimer model, in which two inverted hexamers separated by a C:G spacer are the unit of recognition by one Fur dimer and confirm that two overlapping units are required for stable interaction of two Fur dimers with the consensus 19-bp sequence. This model is further confirmed by Ferguson analysis, which demonstrated the stoichiometry of the Fur-DNA interaction.

MATERIALS AND METHODS

Plasmids and strains.

E. coli DH5α from Invitrogen Corp. (Carlsbad, Calif.) was used as the host for all plasmids. Plasmid pUC19 was purchased from Promega Corp. (Madison, Wis.).

Media, growth conditions, and enzymes.

Luria-Bertani (LB) medium was used for growth of all strains, and strains containing pUC19 derivatives were grown in LB with 50 μg of ampicillin per ml (30). All restriction enzymes and T4 polynucleotide kinase were purchased from New England Biolabs (Beverly, Mass.); Taq polymerase was purchased from Sigma Chemical Co. (St. Louis, Mo.); and [γ-³²P]dATP (3,000 Ci/mmol) was purchased from New England Nuclear (Boston, Mass.). Fur protein was purified as described previously (25).

General genetic methods.

Plasmids were isolated with the GenElute plasmid miniprep kit from Sigma, and PCR products were cleaned with the QIAquick PCR purification kit (Qiagen, Inc., Valencia, Calif.). Cloning procedures and transformations were carried out as previously described (37). Resulting clones were sequenced by the University of Missouri DNA Core Facility with an ABI 377 sequencer from Applied Biosystems, Inc. (Foster City, Calif.).

Creation of Fur binding sites.

The Fur binding sites used in this study were created by annealing complementary primers containing sequences of natural or synthetic Fur binding sites flanked by EcoRI and BamHI restriction sites. These primers are listed in Table 1. Equimolar amounts of the complementary primers were incubated at 94°C for 5 min and then allowed to cool to room temperature over a 2-h period. The resulting double-stranded DNA was then cleaved with EcoRI and BamHI and ligated into pUC19 cut with the same restriction enzymes, and the plasmids were transformed into DH5α. The pUC19 derivatives containing the various Fur binding sites were then used as a template for PCR to create DNA fragments for use in gel retardation assays and Ferguson assays.

TABLE 1.

Primers used for creation of binding sites

Binding site	Primer	Sequencea
F-F	2hexF	5′-GGAATTCTGAGATAATGATAATCAGGATCCCG-3′
	2hexR	5′-CGGGATCCTGATTATCATTATCTCAGAATTCC-3′
F-F-F	3hexF	5′-CGGAATTCGATAATGATAATGATAATGGATCCGCG-3′
	3hexR	5′-CGCGGATCCATTATCATTATCATTATCGAATTCCG-3′
F-F-F-F	4hexF	5′-GGAATTCGATAATGATAATGATAATGATAATGGATCCGCG-3′
	4hexR	5′-CGCGGATCCATTATCATTATCATTATCATTATCGAATTCC-3′
F-F-x-R	HHTF	5′-CGGAATTCGATAATGATAATCATTATCGGATCCGCG-3′
	HHTR	5′-CGCGGATCCGATAATGATTATCATTATCGAATTCCG-3′
F-F-x-R-R	HHTTF	5′-GGAATTCGATAATGATAATCATTATCATTATCGGATCCCG-3′
	HHTTR	5′-CGGGATCCGATAATGATAATGATTATCATTATCGAATTCC-3′
fepB	fepBF	5′-GGAATTCGAAAATGAGAAGCATTATTGGATCCCG-3′
	fepBR	5′-CGGGATCCAATAATGCTTCTCATTTTCGAATTCC-3′
entS	entSF	5′-GGAATTCGATAATGAAATTAATTATCGTTATCGGATCCCG-3′
	entSR	5′-CGGGATCCGATAACGATAATTAATTTCATTATCGAATTCC-3′

^aThe sequence of the actual binding site is shown in bold.

Gel retardation assays.

Templates for gel retardation assays were created by PCR with primers M13F (5′-CAGGGTTTTCCCAGTCACGA-3′) and M13R (5′-TCACACAGGAAACAGCTATGAC-3′) on the pUC19-derived plasmids described above. PCR products were cleaned and end labeled with [γ-³²P]dATP (3,000 Ci/mmol) and T4 polynucleotide kinase. Gel retardation assays were performed as previously described (12). Briefly, 0.1 nM DNA was incubated with increasing amounts of Fur in 1× gel binding buffer [10 mM Bis-Tris (pH 7.5), 5 μg of sonicated salmon sperm DNA per ml, 5% (vol/vol) glycerol, 100 μM MnCl₂, 100 μg of bovine serum albumin per ml, 1 mM MgCl₂, and 40 mM KCl] in a total volume of 20 μl for 10 min at 37°C. The entire reaction was then loaded on a 7.5% nondenaturing polyacrylamide gel (75:1, acrylamide-bisacrylamide), and the gel was run at 200 V for 2.5 h. The banding pattern was examined by exposing the gel to Kodak XRP-5 film (Eastman Kodak, Rochester, N.Y.).

To quantitate the apparent dissociation constant (K_D) for each sequence, gels were exposed to Imaging Screen-K from Kodak. The image was scanned into a Molecular Imager FX (Bio-Rad Laboratories, Hercules, Calif.), and the intensity of the individual bands was measured with the Discovery Series Quantity One software (Bio-Rad). Percent DNA bound by Fur versus unbound fragments was calculated at each Fur concentration. One hundred percent represents the association of all DNA in the sample with Fur. Apparent K_D is defined as the concentration of protein at which 50% of the DNA is bound. The values listed are the averages of three separate experiments (3).

Stoichiometry determination (Ferguson analysis).

The molecular weight of protein-DNA complexes formed in gel retardation assays was determined with a modification (33) of a previously published procedure (16) for determining the molecular weight of proteins of unknown size. In this method, gel retardation assays were performed with DNA created from the pUC19-derived plasmids via PCR with primers ferg-1 (5′-GGCCAGTGAATTCGATAATGA-3′) and ferg-2 (5′-AGGTCGACTCTAGAGGATCC-3′). The PCR fragments were cleaned, cut with EcoRI and XbaI to generate smaller fragments ranging from 32 to 40 bp in length, and end labeled with [γ-³²P]dATP (3,000 Ci/mmol) with T4 polynucleotide kinase. For each binding site tested, reaction mixes contained no protein or 200 nM protein. Samples of binding site with and without protein, along with protein markers (nondenatured protein molecular weight marker kit; Sigma), were analyzed on a set of seven gels with polyacrylamide concentrations of 6.5, 7.5, 8.5, 9.5, 10.5, 11.5, and 12.5%. Gels were run as described above except that bromophenol blue marker dye was added to each sample before loading. The gels were stained in Coomassie blue for 1 h, destained for 2 h, dried, and exposed to Kodak XRP-5 film.

To calculate the relative mobilities (R_f) of the protein-DNA complexes and the protein markers, the distance from the top of the well to the complex or the protein marker was divided by the distance from the top of the well to the bromophenol blue band. The logarithm of R_f for each sample was plotted against the percent gel concentration, and the negative slopes for each sample were plotted against the molecular weights of the known protein markers. The molecular weight of the various protein-DNA complexes was then obtained with this graph. Since the molecular weight of the DNA is known, this value was subtracted from the total estimated molecular weight, giving the molecular weight of the protein in each complex. This value was then divided by 17,000, the approximate molecular weight of the Fur monomer, thus giving the number of Fur monomers per gel-shifted species. The Ferguson procedure is outlined more thoroughly in Technical Bulletin MKR-137 from Sigma.

RESULTS

Strategy.

To compare the ability of Fur to recognize various binding site sequences (Fig. 1B) in vitro, oligonucleotides containing the sites were annealed, and the resulting double-stranded DNA was cloned into pUC19. The pUC19-derived plasmids were then used to create DNA fragments for gel shift experiments. In addition, these sequences were used to determine how many Fur molecules recognized each sequence in Ferguson assays. The purpose of this study was to compare binding to three types of sites: those containing adjacent hexamers (F-F, F-F-F, and F-F-F-F), those containing two hexamers in the forward orientation followed by one or two hexamers in the reverse orientation, with a base in between (F-F-x-R and F-F-x-R-R), and natural Fur binding sites (fepB, with an F-F-x-R orientation, and entS, with an F-F-x-R-R orientation). In the case of the natural binding sites, only the immediate binding site was used to examine the initial binding events; flanking sequences were not included, as these have been shown to affect the ability of Fur to polymerize on DNA (15).

Gel retardation assays comparing synthetic binding sites.

Previous studies (14) demonstrated that Fur binding sites consist of three adjacent hexamers of the sequence 5′-GATAAT-3′, which could be arranged as three hexamers all in the forward orientation (F-F-F) or as two in the forward orientation followed by one in the reverse orientation, with a single base pair in between (F-F-x-R). It was suggested that binding sites consisting of the F-F-F configuration were recognized just as well as F-F-x-R sites. A more thorough comparative analysis was undertaken in a series of gel retardation assays with labeled DNA containing synthetic initial binding sites flanked by unrelated sequences and increasing amounts of Fur protein.

Gel retardation assays were first performed with DNA containing two to four tandem direct hexamers (F-F, F-F-F, and F-F-F-F; see Fig. 2A). With F-F DNA, high concentrations of Fur were required for binding, with very weak complexes detectable at 25 nM Fur. At the highest concentrations tested (1000 nM; data not shown), unbound DNA still remained. These results were expected, based on previously published work (14). For F-F-F, Fur-DNA complexes were detected at 25 nM Fur, with most of the DNA bound by 250 nM Fur. Finally, with F-F-F-F, the affinity of Fur for DNA was significantly greater, with complexes forming at 5 nM Fur and all the DNA bound between 10 and 25 nM Fur. Multiple Fur-DNA complexes formed with F-F-F-F, demonstrating that more Fur dimers associate with this sequence as the concentration increases. The formation of four distinct gel-shifted species is highly reminiscent of gel shifts with wild-type entS (25).

FIG. 2.

Gel retardation assays with synthetic DNA. DNA was labeled and gel shifts were performed as described in Materials and Methods. Images were created from autoradiographs, and Fur concentrations for each lane are given in nanomolar values. The site tested is shown above each image, and the actual sequence being tested for each gel shift is listed below the image, with arrows depicting the hexamer sequences. A depicts gel shifts for contiguous hexamer sites F-F, F-F-F, and F-F-F-F. B shows gel shifts for synthetic sites F-F-x-R and F-F-x-R-R.

The ability of Fur to bind sequences containing two forward hexamers followed by one or two reverse hexamers with a base in between (F-F-x-R and F-F-x-R-R) was also tested (Fig. 2B). In the case of F-F-x-R, the consensus Fur binding site, very little Fur was needed for association to occur, with initial binding at concentrations as low as 0.25 nM (data not shown). By 5 nM, all of the DNA was bound by the protein, showing that the F-F-x-R site is much better recognized than the F-F-F site. For F-F-x-R-R, affinity was also high, with all DNA being bound at 5 nM as well, showing that it was also better recognized than its four-hexamer counterpart, F-F-F-F. Like F-F-F-F, multiple Fur-DNA complexes formed with increasing concentrations of protein.

The apparent K_D for all sites tested, a value indicative of affinity for each site, is shown in Table 2. Apparent K_D is defined as the Fur concentrations at which 50% of the DNA is associated with the protein. The apparent K_D was much lower for the sites containing a single base pair in between hexamers in opposite orientations, suggesting that Fur has a greater affinity for these sequences than for those containing contiguous hexamers in the same orientation.

TABLE 2.

Apparent dissociation constants (K_D) from gel shifts^a

Binding site	Mean apparent KD (nM) ± SD
F-F	250.00 ± 18.0
F-F-F	36.10 ± 7.1
F-F-F-F	6.30 ± 0.6
F-F-x-R	0.95 ± 0.1
F-F-x-R-R	0.78 ± 0.2
fepB	8.60 ± 1.3
entS	3.01 ± 0.8

^aK_D is the value at which 50% of the DNA is bound by Fur and is given as the Fur concentration at which this occurred.

Gel retardation assays comparing natural binding sites.

The above studies confirmed previous data (14) that increasing the number of hexamers in a binding site increases the affinity of Fur for the sequence. Contradictory to previous data, however, these results also showed that Fur preferred to bind sequences containing two forward hexamers separated by one base pair from one or more hexamers in the reverse orientation. These results were then compared to natural Fur binding sites in the E. coli genome, most of which have an F-F-x-R or F-F-x-R-R architecture. In the enterobactin gene island, the fepB site has an F-F-x-R arrangement (7), and the entS site is of the F-F-x-R-R variety (9, 25).

At 5 nM, Fur was associated with the fepB site, and by 100 nM Fur, all DNA was bound (Fig. 3). Although Fur had a high affinity for this sequence (with an apparent K_D of 8.6; Table 2), Fur had an even higher affinity for the consensus sequence F-F-x-R, as expected. For entS, Fur complexes were detected at low concentrations (1 nM) (Fig. 3), though binding was not as strong as seen with F-F-x-R-R. Once again, it was expected that F-F-x-R-R would be recognized with a higher affinity than entS, as the sequence of entS diverges from that of the consensus. Regardless, Fur recognizes this natural binding site much better than the contiguous four-hexamer arrangement (F-F-F-F). As seen with the other four-hexamer arrangements, multiple complexes formed as the Fur concentration increased, with two species predominating and polymerization occurring at high Fur concentrations (50 to 100 nM). Interestingly, previous studies have shown that, in gel shifts, Fur recognizes the wild-type entS sequence (the immediate binding site and surrounding sequences) at lower concentrations than what is seen here (25). This suggests that surrounding sequences may play some role in the initial binding event.

FIG. 3.

Gel retardation assays with natural binding sites. DNA was labeled and gel shifts were performed as described in Materials and Methods. Images were created from autoradiographs. Fur concentrations for each lane are given in nanomolar amounts. The site tested is shown above each image, and the actual sequence being tested for each gel shift is shown below the image, with arrows depicting the hexamer sequences. In this figure, natural binding sites at the fepB and entS promoter regions were tested.

Ferguson analysis.

The gel shift experiments suggest that Fur formed a single complex with DNA containing three hexamers in either orientation (F-F-F, F-F-x-R, and fepB sites) and formed multiple complexes at sites containing four hexamers (F-F-F-F, F-F-x-R-R, and entS). To determine the number of Fur dimers in the various gel-shifted species, the Ferguson method was used (16, 33). This method allows one to determine the molecular weight of protein-DNA complexes by running a series of nondenaturing gels of various acrylamide concentrations and comparing the mobility of protein-DNA complexes in these gels to that of molecular weight standards (Fig. 4B). For each of the binding sites tested, a smaller DNA fragment (ranging from 32 to 40 bp in length, containing only the binding site itself surrounded by unrelated sequence) was used than in previous gel shift experiments. Gel retardation assays were performed with 0.1 nM end-labeled DNA alone and mixed with 200 nM Fur protein, and representative gel shifts are presented in Fig. 4A.

FIG. 4.

Ferguson analysis of binding sites. Ferguson analysis was performed as described in Materials and Methods. A shows representative gel shifts with DNA containing F-F, F-F-F, F-F-F-F, and F-F-x-R-R binding sites, as indicated above each image. In each case, the left lane represents DNA with no protein added (−), and the right lane represents reactions with the addition of 200 nM Fur (+). B represents the plot of relative mobility (R_f) versus acrylamide concentration, as described in Materials and Methods. The legend is shown below. C is the double-log plot generated to determine the molecular weights of the various Fur-DNA complexes. The y axis depicts the negative slope calculated for molecular weight standards (as described in Materials and Methods), whereas the x axis represents the molecular mass of these standards (in kilodaltons). The legend is shown below. BSA, bovine serum albumin.

Generally, as more hexamers were added to the binding sequence, the mobility of the Fur-DNA complex decreased, suggesting that more Fur dimers associated with the sequence. For sites containing three hexamers in any orientation, migration was very similar. Both the F-F-x-R-R and its natural counterpart, entS, showed two major gel-shifted species of similar size, whereas the other four-hexamer sequence, F-F-F-F, formed only one major complex. These results differed dramatically from the gel shift results presented in Fig. 2 and 3. For F-F-x-R-R and entS, it is likely that only two species formed because more Fur cannot associate with the much smaller 40-bp DNA fragment. For F-F-F-F, it was expected that the gel shift results would be similar to those seen with F-F-x-R-R and entS; however, only one major species of a higher molecular weight was formed.

When the F-F site was used, a weak complex of 52 kDa formed (Fig. 4C), though its appearance was somewhat inconsistent. This molecular mass is close to the predicted value of 55.1 kDa for the F-F sequence complexed with one dimer of Fur, since F-F DNA is 21.1 kDa (a 32-bp fragment) and one Fur dimer is 34 kDa. However, this complex did not show up in some experiments, suggesting that the association of one dimer with DNA is unstable. For F-F-F, a 95.7-kDa complex formed, close to the predicted 90.4 kDa for a 36-bp DNA fragment complexed with two Fur dimers. Similar results were seen with the F-F-x-R site and the fepB site, having molecular masses of 96.9 and 94.2 kDa, respectively. For both the F-F-x-R-R and the entS site, the faster-migrating complex corresponded to two dimers bound to the DNA (molecular masses of 98.1 and 90.5 kDa, compared to the predicted 94.4 kDa for two dimers bound to 40-bp DNA), whereas the slower-migrating complex corresponded to three dimers associated with the DNA (132.5 and 128.7 kDa, compared to 128.4 kDa). For F-F-F-F, only one major complex with a molecular mass of 163.8 kDa formed, indicating that four dimers bound the sequence (predicted molecular mass of 162.4 kDa for four dimers at a 40-bp DNA fragment). It was expected that, like the other four-hexamer sites, two major complexes would form. These results suggest that earlier intermediates were passed through quickly to generate a complex with four Fur dimers. Gel shifts were performed with lower Fur concentrations, but intermediates were not present (data not shown).

DISCUSSION

Fur recognition of its 19-bp consensus 5′-GATAATGATAATCATTATC-3′ is not well understood. Although most evidence suggests that Fur does not bind this site as 9-bp inverted repeats, as originally proposed (8, 40), the alternative hexamer model, defining the unit of recognition as 5′-GATAAT-3′ (14), is also insufficient in explaining many observations about the Fur-DNA interaction. In the work presented here, the hexamer model was tested more thoroughly. As in earlier studies (14), three hexamers were required for stable Fur-DNA interactions, with additional hexamers increasing the affinity of Fur for the sequence. However, the arrangement of these hexamers was shown to be important for high-affinity recognition. Sites containing two hexamers in the forward orientation separated by one base pair from a third hexamer in the reverse orientation exhibited higher affinities in gel retardation assays than sites with three or even four direct hexamers with no separation (Fig. 2). Natural binding sites are of the F-F-x-R configuration, with many containing overlapping binding sites, giving an F-F-x-R-R configuration. In a recent Fur titration assay selection (44), almost all Fur binding sequences selected were overlapping.

The experiments presented here compare these hexamer arrangements as well as the natural sites fepB (an F-F-x-R site) and entS (an F-F-x-R-R site). In the case of the seven sequences tested here, the hierarchy of affinity was F-F-x-R-R > F-F-x-R > entS > F-F-F-F > fepB > F-F-F ≫ F-F. Clearly, the natural architectures are more efficiently recognized. The exception was fepB, a natural F-F-x-R site that was not as well recognized as a four-hexamer site. The reduced affinity for fepB merely reflects a greater divergence from the consensus as opposed to the architecture of the site.

Gel retardation assays produced distinctive banding patterns with these arrangements. The F-F site showed only one Fur-DNA complex, and sites with three hexamers in any arrangement (F-F-F, F-F-x-R, and fepB) formed a single complex with higher mobility than the F-F site. For four-hexamer sites, multiple complexes were formed. In the case of F-F-F-F, at least four major complexes were seen, with the central complex appearing the most prevalent and intermediates and extensions occurring at lower and higher Fur concentrations, respectively. With F-F-x-R-R, four species formed in a pattern similar to that seen in previous studies with entS (25); however, in the entS studies presented here, only three major species were produced, two forming at low concentrations and the third representing polymerization of the protein on DNA at high concentrations. The absence of intervening species (as seen in previous experiments) is likely due to the use of only the immediate 25-bp binding site flanked by unrelated pUC19 sequence and suggests that the formation of additional species requires the natural entS surrounding sequences. The consensus F-F-x-R-R site may be so strongly recognized that multiple Fur-DNA complexes are able to form more readily beyond those with the initial 25-bp binding site no matter what the surrounding sequence, but at entS, the initial complexes formed are somewhat weaker and thus more reliant on surrounding sequences for subsequent Fur binding events. Polymerization still occurred in both cases if high enough protein concentrations were used. These results may indicate that additional binding events are not necessarily sequence dependent, as previously suggested (15), especially if the initial binding site is highly conserved.

Ferguson analysis determined the number of Fur dimers in each gel-shifted species. One dimer bound to the two-hexamer site F-F, though very weakly, and two dimers bound to sites containing three hexamers. These results support the proposed overlapping-dimer binding model (25), suggesting that the 19-bp consensus consists of overlapping 13-mer sequences, with each dimer binding a 13-mer on opposite faces of the helix (Fig. 5A). This model states that Fur binds DNA in a manner similar to DtxR, an iron-responsive transcriptional repressor in gram-positive bacteria. Ferguson analysis with F-F-x-R-R and entS showed that the first two Fur-DNA complexes formed consisted of two and three Fur dimers, respectively. The overlapping-dimer binding model predicts that two dimers would associate with one 19-bp sequence and a third dimer would be added to the overlapping 19-bp sequence. For F-F-F-F, the results were unexpected. Only one species which contained four dimers formed. It is possible that once dimers begin binding this sequence, additional dimers are attracted to the site and bind cooperatively. Alternatively, it is possible that the Fur-DNA complex actually contains more than one DNA molecule per complex.

FIG. 5.

Models for Fur-DNA interactions. A represents the overlapping-dimer binding model, as described in the introduction and Discussion. In this model, each monomer (shown as an oval) binds an inverted hexamer, shown as an arrow, with two dimers required for binding the 19-bp consensus, so F-F-x-R refers to the overall architecture of the 19-bp site. C:G base pair spacers are shown in bold. B shows the 7-1-7 model, as recently described (5). The arrows represent the inverted 7-mer recognized by each monomer of the dimer. The bold bases represent the base separating each 7-mer in a unit. C is an application of the overlapping-dimer binding model to an extended binding site, as described in the text. C:G base pair spacers are shown in bold. Numbers 1, 2, and 3 refer to dimers 1, 2, and 3, as referred to in the Discussion.

Recent studies with Fur from Bacillus subtilis suggested that Fur recognizes its consensus as two 7-1-7 inverted repeats that overlap by 9 bp (Fig. 5B), so that a 21-bp site is required for high-affinity interactions (5). The 7-1-7 model accounts for the C:G base pair spacer in the consensus and suggests that another C:G pair spacer, located at position 8 of the sequence, is also important. In addition, most bacterial regulators are dimeric in nature and bind operators consisting of inverted repeats, with each monomer recognizing one of the repeats (21, 35). The 7-1-7 model suggests that the minimal binding site required for two dimers to bind is the 19-bp consensus flanked by a T on the 5′ end and an A on the 3′ end, and these bases are actually present at many Fur binding sites in E. coli.

Both the synthetic and natural sequences used in the experiments presented here did not include the flanking T's and A's, so the full requirement for these bases is still unknown. However, Fur did still bind all sites tested (except for F-F) with high affinity. Therefore, F-x-R is the minimal unit recognized by one Fur dimer, with two overlapping F-x-R sequences required for high-affinity Fur binding (Fig. 5A), suggesting that each monomer binds a 6-bp inverted hexamer. Two dimers are required for stable Fur-DNA interactions, as the association of a single dimer was only weakly detected even at low Fur concentrations (Fig. 2A). Therefore, the preferred architecture for binding is a minimum 19-bp sequence, with an overall F-F-x-R arrangement.

Many Fur binding sites consist of overlapping 19-bp sequences, and several have been examined experimentally, including operators at the iucA, cir, fepA-fes, sodA, and entS-fepD promoter regions (9, 12, 20, 23, 25, 43). An examination of putative Fur binding sites throughout the E. coli genome showed that a majority of sites could contain overlapping binding regions (44). Although the results presented here and previous results (7, 13, 14) show that a single 19-bp site is sufficient for binding, the overlapping binding sites may allow a greater range for Fur regulation, as such sites would allow the association of additional dimers (Fig. 5C). Overlapping sites would allow the association of at least one additional dimer, though other dimers could easily associate upstream of the initial binding site, depending on the conservation of the initial binding site itself and the AT-richness of the upstream sequence. The binding site for dimer 3 overlaps that for dimer 1, which could indicate that protein-protein interactions are occurring between dimers as more associate with the sequence. These protein-protein interactions would allow easy polymerization of protein on the DNA, as additional dimers could associate with less-conserved DNA binding sequences by relying on these protein-protein interactions.

The overlapping-dimer binding model (25) and the 7-1-7 model (5) explain how a dimeric protein would recognize its binding sequence as inverted hexamers as opposed to direct hexamers. Although these models explain the initial binding event, the ability of Fur to polymerize directionally on DNA is still not understood. Further studies will be needed to characterize subsequent Fur-DNA binding events.