Characterization of the interaction between Fur and the iron transport promoter of the virulence plasmid in Vibrio anguillarum.

The expression of iron transport genes fatDCBA in Vibrio anguillarum strain 775 is negatively regulated by two iron-responsive repressors, the Fur protein and the antisense RNA, RNAalpha. Here we report the identification of the promoter for the iron transport genes and studied the interaction between the V. anguillarum Fur protein and this promoter. The iron transport promoter was localized in a region approximately 300 base pairs upstream of fatD by both primer extension and S1 mapping analysis. High activity of the promoter was measured in response to iron depletion in the wild-type strain when a promoter-lacZ fusion was examined, whereas the promoter was constitutive in the Fur-deficient strain. Gel retardation and DNase I footprint analysis showed that Fur binds specifically to two contiguous sites comprising the promoter region and the region downstream of the transcription start site. The identified Fur binding sites showed a low degree of homology to each other as well as to the consensus sequence for the Escherichia coli Fur protein. DNase I footprints pattern suggested a sequential interaction of Fur with these two sites that renders a protection in the template strand and a hypersensitivity to the nuclease in the nontemplate strand. The periodicity of the hypersensitive sites suggested that the promoter DNA undergoes a structural change upon binding to Fur, which might play a role in the repression of gene expression.

The expression of iron transport genes fatDCBA in Vibrio anguillarum strain 775 is negatively regulated by two iron-responsive repressors, the Fur protein and the antisense RNA, RNA␣. Here we report the identification of the promoter for the iron transport genes and studied the interaction between the V. anguillarum Fur protein and this promoter. The iron transport promoter was localized in a region approximately 300 base pairs upstream of fatD by both primer extension and S1 mapping analysis. High activity of the promoter was measured in response to iron depletion in the wild-type strain when a promoter-lacZ fusion was examined, whereas the promoter was constitutive in the Fur-deficient strain. Gel retardation and DNase I footprint analysis showed that Fur binds specifically to two contiguous sites comprising the promoter region and the region downstream of the transcription start site. The identified Fur binding sites showed a low degree of homology to each other as well as to the consensus sequence for the Escherichia coli Fur protein. DNase I footprints pattern suggested a sequential interaction of Fur with these two sites that renders a protection in the template strand and a hypersensitivity to the nuclease in the nontemplate strand. The periodicity of the hypersensitive sites suggested that the promoter DNA undergoes a structural change upon binding to Fur, which might play a role in the repression of gene expression.
The marine pathogen Vibrio anguillarum strain 775 possesses a highly efficient plasmid-mediated iron uptake system that competes with the vertebrate host fish for the iron bound to high affinity iron binding proteins, such as transferrin and lactoferrin, leading to the establishment of infection (1). The pJM1 plasmid-mediated iron uptake system includes an iron scavenger, the siderophore anguibactin, and an energy-coupled iron transport system that internalizes the ferric anguibactin complex (2). The iron transport system includes an 86-kDa outer membrane receptor protein encoded by the fatA gene (3,4), a 40-kDa cytoplasmic membrane-embedded lipoprotein that possesses periplasmic binding domains encoded by the fatB gene (5), and two 37-kDa integral membrane proteins encoded by the fatD and fatC genes (6,7). These genes are located contiguous in the pJM1 plasmid in the order fatD, -C, -B, and -A. (Fig. 1). Genetic studies utilizing transposon (Tn3::HoHo1) insertion mutagenesis have demonstrated that mutations in fatD, -C, or -B affected FatA expression, suggesting an operon organization of these genes (6,8). A further regulatory linkage between iron transport genes and anguibactin biosynthetic genes has also been proposed since insertion mutations in the iron transport genes led to lowering or complete shutoff of anguibactin biosynthesis (8).
Expression of both, siderophore biosynthesis and iron transport genes, is regulated by positive and negative factors which function at low and high iron level, respectively. The positive regulation is mediated by the plasmid-encoded 110-kDa AngR protein whose gene lies immediately downstream of fatA and by trans acting factors encoded in a region located noncontiguous to the iron uptake region in the pJM1 plasmid (5,8,9,10). The negative regulation is afforded by the chromosomally encoded Fur protein (11,12) and by an antisense RNA (RNA␣) (13). RNA␣ is transcribed as a counter-transcript of the fatB mRNA under the control of Fur (15). While the repression mediated by RNA␣ occurs at the post-transcriptional level, the Fur protein seems to regulate the expression of the iron transport genes at the transcription level. Using an RNase protection assay it has been shown that the fatB and fatA messages are no longer negatively regulated at high iron level; and therefore, their expression is constitutive when a null mutation is generated in the fur gene of the 775 strain (14).
The function of Fur in the repression of iron-regulated genes has been described in many bacteria (16); however, the interaction of Fur with its cognate operator has been characterized predominantly in the Escherichia coli plasmid aerobactin system (17)(18)(19) in which the Fur protein, in the presence of divalent metal ions, binds to a 19-bp 1 operator sequence in the promoter region. The primary structure of the Fur protein (about 17 kDa) is highly conserved in a variety of bacteria. The comparison of predicted amino acid sequence showed that V. anguillarum Fur shares a high degree of similarity to its homolog in Vibrio species, Vibrio cholerae (94%) and Vibrio vulnificus (92%), and a lower degree of similarity to E. coli Fur (76%) (11).
We report here the identification of the promoter that drives the expression of the polycistronic transcript of the fatDCBA genes and the interaction of the promoter with the Fur protein in vitro, to understand the regulatory mechanism of Fur in V. anguillarum. Analyses by gel retardation and DNase I footprints showed that the V. anguillarum Fur protein interacts with the promoter region in a different manner from its coun-terpart of E. coli, suggesting the existence of a different repression mechanism in this organism.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-V. anguillarum strain 775 is a wild-type strain harboring the plasmid pJM1 (1), and its isogenic strain 775MET11 is Fur-deficient (12). E. coli strain MM294 harboring pRK2013 was used as a helper for conjugation between V. anguillarum and E. coli HB101 (20) strains carrying recombinant plasmids, E. coli XL1-blue (21) for cloning. The plasmids pJHC-SC85 and pJHC-SC87 are derivatives of pBluescript SKϩ (Stratagene) containing either the 1.4-kb BstEII-HindIII fragment from the fatD upstream region between the sites HindIII and SmaI in the vector (pJHC-SC85) or the 0.3-kb XmnI-EcoRI fragment from the fatD upstream region between the sites EcoRV and EcoRI in the vector (pJHC-SC87). The plasmid pJHC-TW95Z was constructed using the following cloning steps: the 1.4-kb BstEII-HindIII fragment containing the fatD promoter region was cloned between the sites SalI and HindIII in a derivative of pBR325 (22), which carries a kanamycin resistance gene inside the ampicillin resistance gene (PstI site) in the opposite orientation to the disrupted gene. The construct was then used as a vehicle to clone the 9-kb HindIII-BamHI fragment containing the lacZYA operon, isolated from the plasmid pPD104 (23), between the HindIII and NcoI sites of the vector.
Cell Growth-V. anguillarum cells were grown at 26°C in Trypticase soy broth (Difco) supplemented with 1% (w/v) NaCl and antibiotics, when necessary, until stationary growth phase. The cultures were then diluted 100-fold in M9 minimal medium (1) supplemented with 0.2% casamino acids and antibiotics, when necessary. The cells were grown repeatedly until stationary growth phase and subsequently transferred in 100-fold dilution in the minimal medium described above without antibiotics. To generate various iron levels for the cells, the minimal medium in this step was supplemented with either 2 g/ml ferric ammonium citrate for iron-rich condition or ethylenediamine-di-(o-hydroxyphenylacetic) acid (EDDA) from 0 to 15 M for iron limitation. The cells were grown to optical density of 0.5-1 at 600 nm and harvested to isolate total RNA or to measure the ␤-galactosidase activity. The amounts of the antibiotics applied in V. anguillarum strains were: 60 g/ml kanamycin. The Fur-deficient strain was grown in the presence of 2 mM MnCl 2 .
General Methods-Conjugation to V. anguillarum (24), ␤-galactosidase assay (25), and RNA isolation (26) were followed as described previously. The primer extension experiment was carried out with the synthetic primers 5Ј-GTAGCGCACAAAGTAAAGAACGCC-3Ј (primer 1) and 5Ј-GGTCACAGCAGGGATTTAGCAGAGC-3Ј (primer 2), which are complementary to the 5Ј-end region of fatD. The primer was endlabeled with T4-polynucleotide kinase (Life Technologies, Inc.) in the presence of [␥-32 P]ATP and annealed to total RNA (50 g). Reverse transcription from the primer by avian myeloblastosis virus reverse transcriptase (Promega) and urea-PAGE (6%) were followed as described (21). The S1 mapping analyses were carried out using riboprobes as described (21). The riboprobes, 690 nucleotides (nt) containing the XmnI-HindIII segment from the plasmid pJHC-SC85, and 378 nt containing the XmnI-EcoRI segment from the plasmid pJHC-SC87 were prepared by linearization of the plasmid with XmnI and HindIII, respectively, and following transcription with T3 RNA polymerase. RNA-RNA hybridization was carried out at 42°C overnight and followed by digestion with S1 (Life Technologies, Inc.) at 37°C for 2 h. The resulting S1 digested samples were analyzed in parallel with a known sequence marker by urea-PAGE (6%).
Protein-DNA Interaction Assays-V. anguillarum Fur, purified to homogeneity, was obtained from E. Zhelenova from the laboratory of R. Brennan. Gel mobility retardation assay was carried out with the 310-bp HindIII-EcoRI isolated from the plasmid pJHC-SC87 and labeled with [␣-32 P]dATP by the filling in reaction of the Klenow enzyme. The labeled DNA (24 pM) was incubated with various concentrations of Fur protein (30 nM to 60 M), 1 g of poly(dI-dC), and 2 g of bovine serum albumin in the final volume of 20 l of reaction buffer (10 mM Bistris/boric acid (pH 7.5), 100 M MnCl 2 , 1 mM MgCl 2 , and 40 mM KCl) at 37°C for 15 min. The reaction mixture was separated by nondenaturing (nd) PAGE (5%) in buffer (10 mM Bistris/boric acid (pH 7.5), 100 M MnCl 2 ) at 90 V for 3 h. DNase I footprint assay was performed according to the method of Galas and Schmitz (27). The 315-bp HindIII-PstI fragment and 346-bp EcoRI-KpnI fragments were isolated from the plasmid pJHC-SC87 and labeled at recessive 3Ј-end of the fragments by the filling in reaction of the Klenow enzyme in the presence of [␣-32 P]dATP. The labeled DNA preparations were incubated with various concentrations of Fur (480 nM to 60 M) under the same reaction conditions described in the gel retardation assay at 37°C for 20 min. The reaction mixture was exposed to a partial digestion with DNase I (Boehringer Mannheim) at 37°C for 2 min. After stopping the reaction with 25 mM EDTA, the DNA was precipitated and analyzed by urea-PAGE (6%).

Determination of the Transcription Start Site for fatDCBA mRNA and Identification of the Iron Transport Promoter-To
determine the transcription start of the fatDCBA mRNA, primer extension analysis was carried out with a primer (primer 1) complementary to the 5Ј-end region of the fatD transcript (Fig. 2C). As shown in Fig. 2A, a number of primer extension products were observed only with RNA samples harvested under iron limitation. The pattern of the molecular species, two larger and a number of smaller bands, is identical at all three levels of iron limitation. Using molecular standards the two largest bands were estimated as 340 Ϯ 10 nt, which maps the transcription start site(s) about 300 nt upstream of the translational initiation start site of fatD. To identify the precise location of these two start sites, a further primer extension experiment was performed with a primer (Primer 2, Fig. 2A) that hybridizes at the region closer to the start site(s). The molecular weight of the two largest primer extension products was compared with a sequence obtained using the same primer ( Fig. 2A) and located at 302 and 287 nt upstream of the start codon of fatD.
The multiple bands detected could be generated as a result of multiple promoters and therefore multiple transcriptional start sites or processing products from larger molecules in vivo or in vitro. Alternatively they could be artifacts of the detection method, generated by the stalling of reverse transcriptase at certain stem-loops in the RNA structure, which might lead to earlier stops during the reverse transcription. To clarify the presence of multiple bands in primer extension analysis, the S1 mapping method was applied. Radiolabeled riboprobes containing either a 302-nt XmnI-EcoRI segment (obtained as a 378-nt fragment using plasmid pJHC-SC87) or a 535-nt XmnI-HindIII segment (obtained as a 609-nt fragment using plasmid pJHC-SC85) were employed in this analysis. RNA-RNA hybrids were subjected to S1 nuclease activity. Fig. 2B shows S1-mapped transcriptional start sites. As it was the case with the primer extension approach the S1 mapping method detects two major transcripts whose start sites correspond to the two major start sites identified by primer extension analysis, confirming that these two mRNA species are indeed present in vivo. The S1 mapping method also shows a number of minor bands, all of which are smaller than the two major bands. The pattern of the minor bands detected with two different riboprobes, however, are not identical and distinguishable from bands detected by primer extension experiments. This result suggested that the minor bands observed in the two mapping methods applied here are artifacts of the detection methods. The search of the RNA structure in the fatD upstream region, by the computer program GCG, reveals at least four stem-loop structures (data not shown). When the location of the stem-loops was compared with that of the early stops detected in primer extension analysis, there was no consistency between them. The DNA sequence of the fatD upstream region is highly rich in A ϩ T (78%). Therefore, it is likely that those minor products obtained with both mapping methods are generated by breathing of two strands that leads to early termination of reverse transcription or nicking by S1. It still remains to be identified whether the two major bands are both products of independent transcription initiation at different sites or if the smaller band is a specific degradation product of the larger one. According to the transcription start site mapped by the larger band (95-nt species; Fig. 2A), we mapped a putative Ϫ10 (TAGCAT) and the Ϫ35 (CTTACA) promoter region. These sequences, however, share a low homology to the consensus sequence of the E. coli promoter. The diversity of the sequence in the iron transport promoter implies that this promoter might require a transcriptional activator that locates RNA polymerase in the promoter region and enhances its binding.
Analysis of the Iron Transport Promoter-lacZ Fusion-Since the expression of fatDCBA mRNA is regulated by the iron level, it is expected that the putative promoter, identified by primer extension, is also regulated by the iron level in the cell. To assess whether this was the case the putative promoter, called iron transport promoter, was examined using the lacZ-reporter gene expression to verify that fatDCBA genes are indeed driven from this promoter. The 1.4-kb BstEII-HindIII fragment containing the putative promoter region (Fig. 1) was coupled transcriptionally to the lacZ-reporter gene in a pBR325 derivative plasmid, resulting in plasmid pJHC-TW95Z. The fusion construct was subsequently conjugated into the wild-type strain 775 as well as the Fur mutant strain 775MET11. Cells harboring pJHC-TW95Z were grown under various concentrations of iron. The ␤-galactosidase activity of the cells was measured when cells were in the early stationary growth phase, and the . Lanes G, A, C, and T represent the sequence of the fatD upstream region obtained from the primer used in the respective experiment. B, S1 nuclease mapping of the 5Ј-end. Total RNA preparations from strain 775 grown in high iron (lanes 2, 2 g/ml ferric ammonium citrate) or low iron (lane 3, minimal medium alone; lane 4, 3 M EDDA) medium were hybridized with two different riboprobes containing either the XmnI-EcoRI or XmnI-HindIII segment (see C and Fig. 1). Lanes 1, riboprobes without S1 treatment were loaded. The gel electrophoresis was carried out as described above. The size of the S1-generated products were measured by comparison with known sequence marker (not shown). C, DNA sequence of fatD 5Ј-end region. Only the sequence of the nontemplate strands is shown. The translational start of the fatD gene ATG and the location of the primer are indicated by bold letters. The first (ϩ1) and second transcription start sites are indicated by small arrows and the Ϫ10 and Ϫ35 sequences are denoted. results are presented in Fig. 3. As the diagram shows, the ␤-galactosidase activity of the wild-type cells measured was negligible when the cells were grown in high iron, indicating no promoter activity. However, the overall ␤-galactosidase activity increased significantly under iron depletion with EDDA. The maximum increase was ϳ200-fold at 3 M EDDA, and the activity decreased slightly under high iron stress (up to 15 M EDDA). Such iron-responsive activity of the promoter was not observed in the Fur mutant cells (Fig. 3). Here, the ␤-galactosidase activity was high under both high and low iron conditions, indicating that the promoter is constitutive in the absence of Fur in vivo. Taken together, these results strongly suggested that the promoter within the tested area is activated by iron limitation and repressed in high iron levels. Such iron regulation in a promoter region may be achieved by a complex of the repressor Fur protein with iron, as characterized in many other iron-regulated genes in Gram-negative bacteria.
Fur Binding Assay with the Iron Transport Promoter-The possibility that the iron regulation of the iron transport promoter is mediated by Fur-iron complex was examined in vitro.
To determine the binding activity of the V. anguillarum Fur protein to the promoter, a gel retardation assay was performed under the binding reaction conditions employed for E. coli Fur binding to the aerobactin promoter (19) with few minor modifications. The detection of Fur-DNA complex was carried out in the presence of metal ion Mn 2ϩ , as a substitution of the natural Fur co-repressor Fe 2ϩ , both in the binding reaction and gel electrophoresis (nondenaturing PAGE). The 302-bp XmnI-EcoRI promoter fragment (24 pM), obtained as a 310-bp Hin-dIII-EcoRI fragment from pJHC-SC87, was incubated with increasing concentrations of V. anguillarum Fur protein (30 nM to 60.8 M) in the presence of 1 ϫ 10 4 -fold weight excess of nonspecific competitor DNA poly(dI-dC) in the reaction prior to the nondenaturing PAGE. As shown in Fig. 4, Fur protein specifically binds to the promoter DNA fragment. The varying amounts of Fur protein in the reaction mixture led to the detection of Fur-promoter complexes with four different gel migrations, designated as complexes a, b, c, and d in Fig. 4. At lower protein concentrations, 30 -120 nM monomer, only a single complex band (a) was detected, and in higher concentrations 240 -960 nM two additional complex bands (b and c) appeared. While the complexes a and c seem to be formed in the same high abundance in this protein titration assay, the complex b appeared to be of lower abundance as compared with the other two complexes, a and c. From the gel retardation assay we were not able to estimate the number of Fur protein binding units on the tested DNA fragment. However, it is evident that the complexes b and c result from Fur binding at more than two different sites in the DNA fragment. The less abundant complex b implies an unstable nature of this complex, at least as assessed by the gel retardation method. This could be due to either an unstable conformation of the complex or a Fur binding site with low affinity. At protein concentrations 960 nM and higher, the protein-DNA interaction shows a plateau, indicating that all available Fur binding sites are occupied by Fur. The highest retarded complex d occurred at a 60.8 M protein concentration. It is likely that this complex was formed as a result of nonspecific binding of Fur aggregates in the tested conditions. Although this gel retardation assay was not a detailed kinetic assay, we estimated the apparent binding constant K app of Fur to the iron transport promoter between 30 and 60 nM. This value is relatively high as compared with that of E. coli Fur to the iucA promoter (5 nM; Ref. 28) and the sodA promoter (10 -20 nM; Ref. 29).
DNA Footprinting Analysis of Fur-Iron Transport Promoter Complex-To locate the Fur binding sites in the 302-bp XmnI-EcoRI fragment containing the promoter region, DNase I footprinting assays were performed. The DNA fragment was radiolabeled at the 3Ј-end of either the template or nontemplate strand relative to the fatDCBA transcripts. The binding reaction was carried out with 1.5 nM DNA fragment and varying concentrations of the V. anguillarum Fur protein (480 nM to 60 M) under the same binding conditions applied in the gel retardation assay. The autoradiograms of DNase I footprint patterns on both strands are shown in Fig. 5 and the summary of the Fur-DNA interaction is represented in Fig. 6. Both template and nontemplate strands showed a large interaction region with Fur that comprised about 83 and 56 nt, respectively. At low protein concentrations (480 -960 nM) the template strand reveals a region of 49 nt protected from DNase I cleavage by binding to Fur (Fig. 5, lanes 8 and 9). This primary interaction site with Fur, designated as site I, is located directly in the promoter region, spanning from position ϩ10 to Ϫ32 relative to the transcription start site, overlapping with the Ϫ10 and Ϫ35 region of the promoter. In the region between position ϩ10 and Ϫ7 the protection was observed in only a few nucleotides. It is possible that protected nucleotides in this region are the result of the paucity of the DNase I footprints. At higher concentrations of protein (Ͼ1.9 nM) a protected region appeared in the downstream region of the transcription start site from position ϩ17 up to ϩ50. This secondary binding site (site II, 36 nt) is a few nucleotides smaller than the primary binding site I (42 nt), indicating that Fur binds these two sites either with different multimeric forms or in a different manner. The protection pattern in sites I and II (82 nt) was maintained in the presence of higher protein amounts (Fig. 5, lanes 2-6).
The footprint pattern in the nontemplate strand was detected in the region spanning from position Ϫ3 up to position ϩ49. The footprints within this region contain three areas that show hypersensitivity to DNase I cleavage. In each case, two hypersensitive nucleotides are flanked by protected nucleotides. These three hypersensitive sites appeared periodically spaced by about 15-16 nucleotides, which means a periodicity of longer than one helical turn of B-DNA. As is the case of the template strand, the footprints of the nontemplate strand revealed two Fur interaction sites (IЈ and IIЈ) that are distinguishable in affinity for Fur. The primary interaction at site IЈ observed at protein concentrations 480 -960 nM reveals protection and hypersensitive sites from position Ϫ6 to position ϩ13 (Fig. 5, lanes 8 and 9) within the site I region in the opposite strand. The secondary interaction site (site IIЈ) with two hypersensitive sites was detectable at 1.9 M Fur in the downstream region of site IЈ. The site IIЈ is located in the complementary region of site II. The temporal and spatial coincidences of interaction sites suggested that the footprints at sites I and IЈ or at site II and site IIЈ feature the same event with different consequences. The sensitivity of this strand was enhanced at higher concentrations of protein. The upstream region of the site IЈ revealed extremely high sensitivity in a large area of about 80 nt. The hypersensitive sites appeared only in the nontemplate strand, and the main protection was observed in the template strand. These data suggested that the mutimeric forms of Fur bind mainly to the template strand and change the conformation of the strand upon binding that causes a stronger exposure of the less interactive nontemplate strand to the DNase I nucleolytic attack. The enhanced sensitivity in the upstream area could be generated likewise by a conformational change of the DNA, e.g. DNA bending, presumably constrained by protein-protein interaction. Experiments to explore these possibilities are currently being carried out. DISCUSSION The regulatory function of the Fur protein in the pJM1encoded iron uptake system has been demonstrated previously by isolating a Fur-deficient isogenic strain, a null mutant that is constitutive in both, expression of the FatB and FatA proteins, as well as in the production of dihydroxybenzoic acid, an intermediary in the biosynthesis of the siderophore anguibactin (5,11). The aim of the present study was to characterize the repression mechanism mediated by the Fur protein in the expression of the iron transport genes, in conjunction with the identification of the promoter(s) that drive the transcription of these genes.
Primer extension analysis has identified, in accordance with S1 mapping experiments, two distinct RNA species starting from the position 302 and 287 nucleotides upstream of the fatD translation start, when the wild-type cells were grown under iron limitation. It still remains to be determined whether these two transcripts are a result of two independent transcription starts or the smaller transcript is a product of a partial degradation of the larger one at a preferential site in vivo. We presumed that, independent of the smaller RNA species, the larger species presents the transcription start site (ϩ1), and according to this ϩ1 site we mapped the Ϫ10 and Ϫ35 promoter region. This putative promoter, however, shows a high divergence in the Ϫ10 and Ϫ35 sequence from the promoter consensus sequence. Since the genes fatDCBA are expressed at high level under iron stress, we presumed that a transcriptional activator functioning under iron stress is involved in the stimulation on the binding of RNA polymerase to the Ϫ35 and Ϫ10 promoter region. The activity of the promoter was monitored by using a promoter fusion to the lacZ reporter gene. The ␤-galactosidase activity representing the promoter activity was seen in high level in response to iron depletion in the wild-type cells (Fig. 3). This corresponds to the appearance of the fatDCBA transcripts in the early stage of the iron limitation when low iron stress was applied (Fig. 2). In addition to the high activity, the promoter identified seems to possess high strength by itself, although it lacks sequence homology to the Ϫ10 and Ϫ35 consensus sequence. In the absence of positive regulators, this promoter was still able to drive the transcription at a low level The 302-bp XmnI-EcoRI fragment containing the fatDCBA promoter was obtained as a 315-bp HindIII-PstI fragment or as a 346-bp EcoRI-KpnI fragment from pJHC-SC87 for 32 P labeling at the 3Ј-end of template or nontemplate strand, respectively. In each case, 1.5 nM labeled DNA fragment were incubated with different concentrations of Fur in the presence of 1 g of nonspecific competitor poly(dI-dC) in the binding reaction for 10 min at 37°C and were subjected to partial digestion with DNase I and subsequent urea-PAGE (6%). The Fur interaction sites were mapped by comparison with a known sequence marker and the size of the fragments are shown on the left-hand side of each autoradiogram in base pairs. The location of the Fur interaction sites (I, II under iron stress. 2 Moreover, as shown in the Fur-deficient strain, the promoter is active under high iron in the absence of the negative regulator. All these data imply a crucial role of Fur as a repressor acting at the promoter to control the expression of the genes fatDCBA at the transcription initiation level. Gel retardation assays showed that the V. anguillarum Fur protein specifically binds to the 310-bp DNA fragment containing the iron transport promoter in the presence of high amount (1 ϫ 10 4 -fold weight excess) of noncompetitor DNA poly(dI-dC). The Fur-DNA complexes were formed, resulting in an early complex, complex a, and the later complexes, complexes b and c. The successive binding of Fur to the promoter DNA was corroborated by DNase I footprinting analysis. The Fur interaction with DNA was observed in both strands, template and nontemplate, resulting in different sensitivities to the DNase I cleavage. While the template strand showed a large protected region by Fur, the nontemplate strand exhibited enhanced sensitivity to the nuclease activity. These results imply that the template strand may be providing the main contact sites to the Fur protein, making DNase I inaccessible to those sites. Furthermore, the contact of Fur in one strand leads to the exposure of the other with enhanced sensitivity to DNase I. The template strand showed two contiguous regions protected by Fur with different affinities. The primary binding site (site I) of 42 nucleotides in size is located in the promoter, overlapping with the Ϫ10 and Ϫ35 region. The preferential protection sequences were observed predominantly in the Ϫ35 region. Unexpectedly, the secondary Fur binding region was mapped in the downstream region of the transcriptional start site, protecting 36 nucleotides.
The estimated molecular mass of the V. anguillarum Fur is ϳ16.8 kDa, similar to its homolog in E. coli. Based on in vitro data it has been suggested that the E. coli Fur protein is active as a dimer or as a multimer (17,18,28) that binds to its 19-bp dyad symmetrical operator sequence, so called Fur consensus sequence or iron box (19, 28; Fig. 7) in the promoter region. The Fur consensus sequence has been initially identified and proposed in the aerobactin promoter that comprises two contiguous Fur binding sites designated as primary (site I; 31 nt) and secondary site (site II; 19 nt) due to their successive binding to the protein. In that promoter, the primary binding site overlaps with the Ϫ35 region and the spacer and the secondary site with the Ϫ10 region. As reported by de Lorenzo et al. (19), a 19-bp symmetry dyad has been found in a number of promoters of iron-regulated genes, e.g. promoters for iucA, fhuA, and fepA. However, no significant degree of sequence homology was found among them, except that base pair A:T in few positions of this 19-base pair operator is conserved. Our results show that V. anguillarum Fur interacts with a much larger DNA segment than its homolog in E. coli. The active polymeric state of V. anguillarum Fur has not been identified. However, as proposed for E. coli Fur (17), a dimer of the about 17-kDa (monomer) protein could interact with a 19-bp DNA segment. Therefore, it is plausible that two dimers (or a tetramer) of V. anguillarum Fur form a complex with each site, site I (42 nt) and site II (36 nt), as an active unit in DNA binding. To learn more about the DNA element involved in Fur binding, a search for an analog of E. coli Fur consensus has been attempted with the sequence of the V. anguillarum Fur binding sites I and II. A comparison of Fur binding sequences is presented in Fig. 7. The highest similarity to the Fur consensus sequence was observed in a region of 20 nucleotides around the Ϫ10 region in site I. The other portions of the binding sites share a low similarity to the E. coli Fur consensus sequence. Furthermore, the characteristic of the Fur consensus sequence, a dyad symmetry, was found, although imperfect, only in the region with the highest similarity. In site I, the Fur protection was observed preferentially upstream of the Ϫ10 region, but rather weak in the Ϫ10 region. It is unlikely, therefore, that the analog to the E. coli consensus sequence actively participates in Fur binding in the V. anguillarum system. As depicted in Fig. 7, the V. anguillarum Fur binding sites both in the promoter region and downstream of the transcription start are highly rich in A ϩ T. The sequence comparison among those sites shows no significant homology. However, the base pair A:T was found in a number 12). Although there was no helix-turn-helix DNA binding motif found, it has been suggested that ␣-helices in the N-terminal region of E. coli Fur may interact with the major groove of the DNA upon sequence-specific recognition (30,31). However, it is difficult to employ the binding mode of E. coli Fur for the V. anguillarum Fur, since the sequence-specific protein-DNA interactions through the major groove generally require a high conservation of DNA sequence, but the DNA sequences involved in the binding to V. anguillarum Fur show a high diversity without any pattern. Therefore, an alternative binding mode could be considered for the V. anguillarum Fur system that is based on low sequence specificity, e.g. recognition and binding to a specific DNA structure in the primary binding sequence.
The DNase I footprinting pattern shows different characteristics in the template (site I and II) and nontemplate strand (site IЈ and IIЈ). While the template strand with sites I and II features protection by Fur, the complementary strand with site IЈ and IIЈ contains at least two types of hypersensitive sites: periodic hypersensitive sites spaced by 15-16 nucleotides appeared in site IIЈ and a cluster of hypersensitive sites in the upstream region of site IЈ (Fig. 6). Such hypersensitivity to DNase I is observed when protein-DNA interactions induce a conformational change of DNA, e.g. a curve in the DNA axis whereby the hypersensitive sites are at the outer face of the helical groove (32,33). Hypersensitivity with periodicity larger than one B-helical repeat was detected when a right-handed superhelix DNA was formed by a multimeric protein, as is the case for the prokaryotic histone-like protein, H-NS (H1; Ref. 34). It is noteworthy that many proteins forming such multimeric complexes with DNA are categorized as histone-like and interact with DNA with low sequence specificity (34,35). Interestingly, the H-NS protein binds to certain promoters with low sequence specificity and deforms the DNA structure, which results in shutting down of the expression of certain metabolic genes (33). Taken together, the binding mode of V. anguillarum Fur resembles those suggested for the prokaryotic DNA-binding protein II with histone-like properties (34,35).
We interpret the results from DNase I footprints and gel retardation assay as follows. V. anguillarum Fur recognizes the DNA structure given by A ϩ T-rich sequences and binds as two dimers (tetramer) primarily to the promoter region (site I), the major interaction occurring with the template strand. This step could correspond to complex a, detected in the gel retardation assay. At higher concentrations of Fur, a further tetramer binds to the secondary binding site II, the downstream region of the transcription start site. The complex formed here seems to be an intermediate and unstable (complex b) and, therefore, undergoes further conformational changes caused by protein-protein interaction between the sites I and II. The protein-protein interaction seems to accompany the formation of the superhelix in the whole Fur binding segment (80 bp), and such a distortion may enhance sensitivity to the DNase I in the outer face of the DNA helix. This complex with superhelicity (complex c) seems to render the repression mechanism by Fur in which the distortion of the DNA helix in the promoter region leads to a deteriorated positioning of the Ϫ10 and Ϫ35 region. Such a structural change is probably no longer recognizable by RNA polymerase, as shown in the mercury resistance operon when the MerR protein was bound to its operator located in the promoter in the absence of metal ion Hg 2ϩ (36). Alternatively the presence of Fur bound in the promoter region could hinder the formation of the closed or open complex of RNA polymerase and the promoter as is the case for the H-NS mediated repression system (32).