Evidence of an Unusually Long Operator for the Fur Repressor in the Aerobactin Promoter of Escherichia coli *

Production of the siderophore aerobactin inEscherichia coli is transcriptionally metalloregulated through the iron-dependent binding of the Fur (ferric uptake regulator) to a large region (>100 base pairs) within the cognate promoter in the pColV-K30 plasmid. We show in this article that such an unusually long operator results from the specific addition of degenerate repeats 5′-NAT(A/T)AT-3′ and not from a fortuitous occupation of the DNA adjacent to the primary binding sites by an excess of the repressor. Furthermore, the protection pattern revealed by DNase I and hydroxyl radical footprinting reflected a side-by-side oligomerization of the protein along an extended DNA stretch. This type of DNA-protein interactions is more like those observed in some eukaryotic factors and nucleoid-associated proteins than typical of specific prokaryotic regulators.

Production of the siderophore aerobactin in Escherichia coli is transcriptionally metalloregulated through the iron-dependent binding of the Fur (ferric uptake regulator) to a large region (>100 base pairs) within the cognate promoter in the pColV-K30 plasmid. We show in this article that such an unusually long operator results from the specific addition of degenerate repeats 5-NAT(A/T)AT-3 and not from a fortuitous occupation of the DNA adjacent to the primary binding sites by an excess of the repressor. Furthermore, the protection pattern revealed by DNase I and hydroxyl radical footprinting reflected a side-by-side oligomerization of the protein along an extended DNA stretch. This type of DNA-protein interactions is more like those observed in some eukaryotic factors and nucleoid-associated proteins than typical of specific prokaryotic regulators.
Prokaryotic transcriptional regulators bind DNA to repress or activate expression of specific genes or groups of genes (1). Although the sequences recognized can be extremely diverse, most regulatory proteins naturally bind discrete target sites within the bacterial genome. However, some regulators (typically the nucleoid-associated proteins; Ref. 2) are also known to bind somewhat degenerated sequences or structural motifs, thus spreading DNA-protein interactions along extended nucleotide sequences. This feature is shared with a variety of eukaryotic regulators, typically those containing zinc fingers such as the transcription factor TFIIIA (3). In this respect, the Fur protein of Escherichia coli displays both properties found in specific transcriptional factors and those in more global regulators. Fur is the product of the fur (ferric uptake regulation) gene (4 -7), which controls transcription of iron-dependent promoters in many prokaryotes. This regulator is a zinc-containing, Fe 2ϩ -binding protein (8) that inhibits transcription of distinct genes implicated in the response to iron starvation when the metal is in excess in the medium (9 -12). But, in addition, Fur appears to play an important role also in a variety of cell functions unrelated to iron acquisition, such as the production of several virulence determinants (13), the defense against oxygen radicals (14,15), the acid shock response (16), chemotaxis (17), metabolic pathways, and others (18 -23).
The interaction of the Fur protein-Fe 2ϩ complex with its operators has been characterized with diverse techniques in several promoters of E. coli and other genera (14, 24 -30). These studies have revealed that every iron-dependent promoter contains a target DNA sequence with different degrees of similarity to a palindromic 5Ј-GATAATGATAATCATTATC-3Ј, 19-bp 1 consensus box (26,31,32). More recently, we have reinterpreted such a consensus as the combination of three repeats of the simpler motif 5Ј-NAT(A/T)AT-3Ј (33), in which the thymines would be the bases determining the type of contact of the Fur protein with such a minimal unit of interaction. The corollary of this interpretation is that extended sites for Fur binding could be naturally or artificially assembled by simply adding multiple adjacent 5Ј-NAT(A/T)AT-3Ј hexamers to a minimum of three repeats. This is a very attractive possibility, because it would permit the generation of repertoires of binding sites of varying extensions and affinities, which would allow Fur to act in some promoters as a very specific regulator and in others as a more general co-regulator (12). Although this notion has been substantiated using synthetic DNA sequences consisting of synthesized 5Ј-NAT(A/T)AT-3Ј hexamers (33), it is unclear whether long sequence-dependent Fur operators are operative in natural iron-regulated promoters. The promoter of the operon responsible for the biosynthesis of the aerobactin siderophore (referred hereafter as Paer) is particularly interesting in this respect (26,32,33,34). Unlike other promoters controlled by Fur in which the operator involves a clear-cut target sequence (13,24,(27)(28)(29), Paer is bound by the repressor to three distinct extents depending on the concentration of the protein ( Fig. 1 and see below). Although Fur binding to the adjacent sites named I and II can be justified by their similarity to the consensus, the massive protection of the further upstream sequences (the so called polymerization region; Fig. 1) is intriguing, because it does not contain clear Fur consensus boxes. Such an extensive occupation of the promoter by the repressor spreading over 100 bp has been revealed not only by DNase I and hydroxyl radical footprinting (32, see below) but also visualized directly through electron microcoscopy (35). Other iron-regulated promoters appear to undergo such an ample occupation as well (25,36), so it might be a genuine phenomenon and not an unspecific protection caused by an excess of the protein.
In this work, we show that the as yet unaccounted binding of the Fur protein to the 5Ј upstream region of the aerobactin promoter is due to the functionality of a long operator composed of nine adjacent 5Ј-NAT(A/T)AT-3Ј hexamers. This operator, which is entirely sequence-dependent, becomes effective only following the occupation of the other two sites. These results support the notion that the binding of Fur to DNA is mediated by the recognition of hexameric repeats and that an increasing number of adjacent repeats allow a co-operative binding of the repressor mediated by lateral protein-protein interactions. Furthermore, we argue that this type of interaction, which has features reminiscent of some transcription factors (37), endows the protein with the ability to behave both as a very specific repressor and as a more general regulator.

EXPERIMENTAL PROCEDURES
General Procedures-The Fur protein used in all the assays was purified to homogenity following the metallo-affinity purification protocol of Wee et al. (7). According to Ref. 8, such purification protocol yields a Fur protein containing 1 atom of zinc/repressor monomer, whose DNA binding ability is responsive to Mn 2ϩ in our assays system (see below). Protein Fur concentrations indicated through this work refer to the protein monomer. DNA techniques were run according to published protocols (38).
DNA Templates for Footprinting Assays-The organization of the DNA fragments used in footprinting assays is shown in Fig. 2. The fragment wt1 is a 368-bp EcoRI-PvuII segment from plasmid pUC-LE15 that contains the region spanning positions Ϫ128 to ϩ32 of the aerobactin promoter region (using as a reference the transcription start site of the main promoter P1) as an EcoRI-BamHI plus a vector-born unrelated BamHI-PvuII extension of 208 bp. The strategy for creating the promoter termed ⌬50 is sketched in Fig. 2A as well. Primers were devised for amplification of the sequence Ϫ50 to ϩ32 (thus excluding the P2 promoter) and the upstream extension region. This fragment was recloned in pUC19 using the EcoRI and BamHI sites present in the amplified fragments (BamHI already present and EcoRI entered with the rightwards primer). To get a template of identical size to wt1 for the footprint assay, a primer was engineered that contained a terminal NcoI site located at exactly the same distance that the EcoRI site of the wt1 promoter (Fig. 2B). Such a segment was then entered at the single HincII site of the previous pUC19 derivative. This new plasmid contained an insert in the vector that spans the new promoter construct ⌬50. The fragment generated after restriction with NcoI-PvuII allowed a base-wise comparison of its footprint with the wild type fragment because the end-sites, EcoRI or NcoI, were located exactly at the same point.
For the second series of templates shown in Fig. 2B, modified variants of the aerobactin promoter with increasing distance between the Fur box I (Fig. 1) and the downstream protected region were constructed as follows. In the EcoRI-BamHI insert of plasmid pUC-LE15 a novel ClaI restriction site was introduced by site-directed mutagenesis at the boundary between Fur boxes I and II ( Fig. 2B) with the method of Kunkel et al. (39). Digestion of the resulting construct with ClaI, filling-in of the cohesive ends, and religation originated a novel NruI as well as ϩ2 bp insertion between the boxes. The same ClaI-digested plasmid was ligated to the linker 5Ј-CGACCATGGT-3Ј, which entered a novel NcoI site as well as a ϩ10 insertion. Finally, NcoI digestion of the resulting construct, filling-in the cohesive ends and religation generated the ϩ14 bp along with a new NsiI site. The mutated segments were cloned back to pUC19 and used as the source of the end-labeled restriction fragments employed in the footprinting assays. To this end, they were excised from these pUC19 derivatives as EcoRI-PvuII or NcoI-PvuII segments (for labeling of the bottom strand) and purified by electrophoresis on nondenaturing 5% polyacrylamide gels. The overhanging ends of the restriction fragments were then filled-in with [␣-32 P]dATP and Klenow polymerase, after which they were further purified from nonincorporated nucleotides on small Sephadex G-25 columns.
Footprinting with DNase I and Hydroxyl Radicals-DNA-protein interactions were probed with DNase I as described in Refs. 32 and 33. Samples were preincubated for 5 min at 37°C with the amounts of the Fur protein indicated in each case. Each tube was then added with 2.5 FIG. 2. Paer promoter variants used as templates for DNA footprinting analysis. A, the fragment wt1 is a 368-bp EcoRI-PvuII segment from plasmid pUC-LE15 spanning positions Ϫ128 to ϩ32 of the aerobactin promoter region as an EcoRI-BamHI plus a vector-born unrelated BamHI-PvuII extension. The promoter variant termed ⌬50 was created by amplifying the sequence Ϫ50 to ϩ32 as an EcoRI and BamHI fragment, combining it with an NcoI-EcoRI extension of identical size of that of the wt1 segment and cloning the whole in pUC19 (see text for explanation). The NcoI-PvuII segment present in the resulting plasmid has its NcoI end located at exactly the same distance from the Fur boxes as the EcoRI site of the wt1 fragment, thus allowing a faithful comparison of its footprint with the wild type fragment. The relative position of each fragment in respect to the functional motifs of the aerobactin promoter (led by the iucA gene) are indicated below, as well as the location of the radioactive label (asterisk) in the DNA fragments assayed. B, modified variants of the aerobactin promoter with increasing distance between the Fur boxes I and II. The boundary between the two boxes was entered with a novel ClaI site, which was further employed for addition of 2, 10, or 14 bp (new bases in bold type) as explained under "Experimental Procedures." The mutated segments were cloned back to pUC19 and used as the source of the end-labeled EcoRI-PvuII restriction fragments employed in the footprints.
FIG. 1. Organization of the aerobactin promoter region. The overall arrangement of functional elements within the DNA segment placed at 5Ј in respect to the aerobactin gene cluster is shown. The promoter region includes two Ϫ10/Ϫ35 hexamers that define promoters P1 (proximal) and P2 (distal). The primary target DNA sequences for the Fur protein (sites I and II) and the upstream extension are pointed as defined by DNase I footprinting (Ref. 32 and Fig. 2), with an indication of the two segments with a maximal coincidence with the 19-bp consensus Fur binding sequences (5Ј-GATAATGATAATCATTATC-3Ј, Fur boxes). The transcription start sites of each of the promoters is indicated as well.
ng of DNase I and further incubated for 2 min. Reactions were stopped, nucleic acids precipitated, dried, and directly resuspended in 7 M urea (with tracking dyes). Samples were loaded on 7-10% polyacrylamyde sequencing gels with 7 M urea. AϩG reactions (40) with the same labeled DNA fragments or sequencing size markers were loaded onto the gels together with the treated samples. Footprinting of DNA with hydroxyl radicals generated in situ with Fe/EDTA/ascorbate were carried out (32) on Fur-DNA mixtures prepared and preincubated in the same conditions as before.

RESULTS AND DISCUSSION
Visualization of a Continuous Pattern of Fur-DNA Interactions through the Aerobactin Promoter Region-To match faithfully the extensions of each binding site for the Fur protein along the Paer promoter with the specific bases involved in protein-DNA contacts, we conducted the experiment shown in Fig. 3. In it, we compared directly the protections caused by increasing concentration of Fur-Mn 2ϩ to either DNase I nicking or hydroxyl radical cleavage of sugar-phosphate bonds (Mn 2ϩ was used instead of Fe 2ϩ because of its superior stability under the aerobic conditions of the experiment; Refs. 11, 12, and 26). The reference DNase I footprint to the left of the gel shown in Fig. 3 revealed the position of each of the known three regions sequentially protected by growing repressor concentra-tions. This includes first a 31-bp sequence (site I) spanning the Ϫ35 hexamer of the P1 promoter, an additional 18 -19 bp downstream protection (site II) overlapping the Ϫ10 box, and the less defined further upstream region protected toward 5Ј (26). Although weaker than those of sites I and II, this last protection involves exactly 60 bp, so that the addition of all sites covered by the protein at the higher protein concentration comes to 110 bp. When the same protein-DNA contacts are inspected in strict parallel with hydroxyl radical footprinting, some salient features become apparent. First, that the distinction between the three binding regions revealed by DNase becomes less clear-cut. At the higher Fur concentration, the whole of the 110-bp DNA sequence displays a continuous and repetitive pattern of two protected residues/four nonprotected bases. The frame of such a regular pattern is shifted only once and by one base at the very boundary between the DNase I site I and the upstream protected region, to then resume the previous two protected/four nonprotected mold through the further upstream region. Furthermore, the data of Fig. 3 suggest that the occupation of such an upstream region does not commence until the site II is fully bound by the repressor. Because we could not distinguish discrete binding sites within the 110-bp region but rather a continuum of repetitive interactions, we wondered whether the entire DNA stretch actually functions as a natural, extended operator of the type predicted by the reinterpretation of the Fur consensus sequences presented before (33).

The Primary Fur Binding Site in the Paer Promoter Nucleates the Occupation of the Adjacent Downstream Sequence-
Because the 50-bp sequence of the Paer promoter spanning DNase sites I and II interacts invariably with the Fur protein with the 6-bp periodicity discussed above, we first addressed whether such sites are independent (as they should be by the 19-bp consensus criterium) or site II is a sequence-dependent extension of site I. Such secondary sites are protected in most E. coli Fur-regulated promoters, although the sequences can be very variable (24,28,29,14). To address this issue, we engineered a ClaI site next to the Fur box of site I (primary binding site of the protein). This site was employed to insert extra bases that changed the relative orientation of the downstream site II by 2, 10, or 14 bp (Fig. 2B). The expected result of such insertions was to either offset moderately the two target sequences (ϩ2) or to separate them but keep the phase of the DNA helix (ϩ10) or to entirely offset and separate sites I and II (ϩ14). The resulting promoter variants were then footprinted with DNase I in the presence of growing concentrations of Fur-Mn 2ϩ , with the results shown in Fig. 4.
The 2-bp insertion (Fig. 4) between the sites entered a change in the extension and strength of the occupation of site II. The first 31 bp (site I) were protected to the same extension and the same protein concentration as the wild type promoter. However, occupation of site II (which was displaced further downstream: upwards in Fig. 4) required a significantly higher Fur-Mn 2ϩ level. Interestingly, the protection of the 5Ј extension region occurred at the very same protein concentration as in the wild type promoter, thus suggesting that such an extension is entirely independent of the presence of protein bound to site II. Although these results indicated that occupation of site II is co-operative with that of site I, they do not rule by themselves that both sites are indeed independent. This issue, however, was unequivocally ascertained by the results of the ϩ10 and ϩ14 promoters. Regardless of the maintenance of the DNA helix phase (ϩ10) or its full disruption (ϩ14), the increased distance between sites I and II resulted in the inability of the downstream site to bind any protein. In both cases, the protection of site I was in all comparable with the wild type Paer After treatment with DNase I or hydroxyl radicals, the mixtures were processed as described previously (32). Size markers from a sequencing reaction were loaded to the right lane to identify the extent of the footprinted sequences. The boundaries of the three regions sequentially protected by Fur are indicated.
promoter. These data favor the notion that site II is an extension of site I rather than a separate target sequence. As in the case of the ϩ2 template, the separation of the sites by longer insertions did not affect at all the upstream 5Ј extensions, which were detected to the same extent and apparent intensity as in the wild type promoter.

Extensive Binding of Fur to the DNA Adjacent to the Primary Binding Site in the Aerobactin Promoter Is Sequence-specific-
The results above gave a preliminary hint on whether the lateral enlargements of the protection caused by Fur on most iron-regulated promoters of E. coli (24,28,29,14) is sequencespecific or whether they just reflect an artifactual occupation caused by a high protein concentration in vitro. This is a reasonable doubt, because such prolongation not always matches the 19-bp consensus Fur box (24,28,29,14). The data of Fig. 4 show that not any sequence adjacent to site I within the aerobactin promoter is suitable to become protected by a high concentration of the repressor. Furthermore, extensions require a certain frame and distance in respect to the primary site. However, the secondary site does include a sequence stretch similar to the 19-bp consensus, thereby suggesting that frame, distance, and specific sequence are all necessary for the enlargement toward the site II. This is, however, less clear at the third protected region, where the extension phenomenon is far more dramatic that in site II (Fig. 3). In this case, the upstream protection observed spans 60 additional base pairs. Although such a protection has a clear directionality and defined boundaries, the sequence involved does not show any significant homology with the reference 19-bp Fur consensus box (26). It is thus conceivable that such a massive protection is not specific and therefore irrelevant to understand the metalloregulation of the promoter. To ascertain this question, we simply prepared a new DNA template ( Fig. 2A) in which we faithfully replaced the upstream DNA by an unrelated sequence. The substitution was such that a DNA fragment of a size identical to that bearing the wild type Paer promoter could be examined in parallel in DNase I footprinting assays. The results shown in Fig. 5 indicated that the unrelated sequence failed to bear any visible extension of the footprint, even at the higher protein concentrations. Furthermore, the 5Ј boundary of the protection was located exactly at the point where the heterologous sequence started (marked with an arrow in Fig. 5). We thus conclude that the binding of Fur to the third region is indeed sequence-dependent. We argue below that this cannot be explained with the generally accepted 19-bp consensus model, but it is perfectly compatible with the notion that a shorter 5Ј-NAT(A/T)AT-3Ј motif is the basic unit of Fur binding.
Reinterpretation of the Fur Operator within the Aerobactin Promoter-The data presented in this work support the hypothesis (33) that Fur binding sites do not follow the standard palindromic organization of target sequences for regulators in prokaryotic promoters (1). Instead, Fur operators of different extensions can be formed by addition, in any orientation, of a minimum of three NAT(A/T)AT hexamers. Although this was shown to be true for artificially assembled NAT(A/T)AT mul- promoter was compared in each gel with its derivatives added with 2, 10, or 14 bases (see sketches in Fig. 2B). The introduction of the ClaI site did not affect Fur binding (not shown). The end-labeled restriction fragments were preincubated for 5 min at 37°C with Fur protein (monomer) concentrations of 30, 60, 120, and 240 nM. After DNase I treatment the reactions were processed as described in (26). AϩG reactions (40) made on the same fragments were loaded in parallel with each sample. The limits of the three sequential regions protected by Fur are indicated. timers (33), this report shows that this is the case also for a natural promoter such as that of the aerobactin operon, whose extended binding sites for the repressor cannot be easily explained with the generally accepted 19-bp consensus model. Fig. 6 shows a reinterpretation of the pattern of Fur-DNA interactions in the promoter as the result of a single, enlarged operator that is formed by additions of up to 18 boxes whose frame give a maximum match to the reference AT(A/T)AT pentamer. Such boxes are separated in all cases by one intervening extra base. The one exception is at the boundary between the site I and the protected upstream region, which lacks such an additional base, a fact that is faithfully reflected in the hydroxyl radical footprint of the region (Fig. 3). It seems that either the deletion of 1 base or the addition of two bases between boxes (as in the artificial promoter ϩ2, see Fig. 4) flaws the co-operative occupation of adjacent hexamers but does not inhibit it. In fact, it is revealing that such a naturally existing deletion between site I and the upstream extension is required to frame maximally the further upstream sequence to the reference NAT(A/T)AT motif. But how does this hypothesis equate the actual data?
The sequence that is protected by the lowest concentrations of Fur-Mn 2ϩ includes 31 bp and, according to the hydroxyl radical footprinting of Fig. 3, consists of a whole of five adjacent hexamers, three of them with a nearly perfect match to NAT(A/ T)AT. The side repeats contain less conserved T residues, and thus their occupation requires a higher repressor concentration, which establishes the pause in the protection that is clearly revealed by DNase I footprint (Fig. 3) and which defines site II. Such second site would include three additional repeats. This extension certainly requires protein-protein interactions with the repressor already bound to site I, to compensate the divergence in the sequence. In fact, some hexamers have only a limited match with the consensus. Thus, the downstream sequence may not bind by itself to the Fur-Mn 2ϩ complex, but it does in the context of the whole promoter. Finally, the long upstream extension can also be sorted out as an array of adjacent Fur-binding hexamers frameshifted by one base in respect to the sequence of boxes included in sites I and II. Although such a shift may explain the lower affinity, the hydroxyl radical data of Fig. 3 shows that the shift resettles the pattern of protein-DNA interactions to the maximum match with the NAT(A/T)AT array. It thus appears that the suboptimal alignment with the primary sites and the considerable sequence divergence of the upstream region are balanced by a higher number of boxes that, as a result, produce an unusually long operator.
Conclusion-Although not to the same dramatic extent as the aerobactin system, many if not all iron-regulated promoters FIG. 5. The extended binding of Fur-Mn 2؉ to the upstream promoter region is sequence-dependent. The gel compares the ability of Fur-Mn 2ϩ to bind to the wild type aerobactin promoter and to an equivalent variant (⌬50; Fig. 2A) in which the region upstream of the Fur site I has been replaced by an unrelated sequence of the same size. Both wt1 and ⌬50 fragments were preincubated for 5 min at 37°C with increasing amounts of the Fur protein (monomer): 30, 60, 120, and 240 nM and processed as before (32). AϩG reactions (40) were carried out with the same labeled DNA fragment and loaded onto the gels together with the treated samples. The start of the substitution is indicated with an arrow to the right of the figure. Note normal occupation of sites I and II but total lack of upstream extension in the ⌬50 template.
FIG. 6. Reinterpretation of the Fur-protected sites in the aerobactin promoter. The figure shows the array of NAT(A/T)AT hexamers at the Paer promoter region that account for the data presented in this work. The boundaries of the primary and secondary operators defined with DNase I footprint (sites I and sites II, respectively), as well as the extension toward adjacent upstream sequences (e.g. the polimerization region), are indicated. The sequences are boxed in hexamers on the basis of maximal similarity to the proposed minimal unit of interaction, which is coincident with the pattern found in the OH radical assays (Fig. 3). Note the unique frameshift of the array at the boundary between site I and the upstream extension. The location and orientation of the three hexamers that determine the primary binding of the repressor to the whole promoter and nucleate the subsequent upstream and downstream extensions are pointed (for a discussion on the orientation of the hexamers, see Ref. 33. of E. coli (14,22,24,28,29,36) contain Fur target sequences that spread beyond the core iron box. No natural Fur binding sites have been found to give less than a 31-bp footprint with DNase I, although the minimal operator is only 19 bp. It thus looks likely that such adjacent sequences are not casual but are indeed arrayed in a configuration of various 6-bp repeats with a potential to interact specifically with the Fur protein as a whole. Extended sites might tolerate a degree of divergence in the sequences involved, which could be compensated by the higher overall affinity. These additional contacts might strengthen the overall binding of the DNA segment to the regulator and do explain why the protection is not limited to the consensus Fur box. The 6-bp box criteria accounts for the variability and extension of the sequences protected by Fur in most iron-regulated promoters and is also compatible with the relatively high amount of Fur molecules (approximately 5000) found inside the cell (16,41). The published DNase I footprinting assays on several promoters (14,24,28,29) can be consistently reinterpreted as arrays of hexameric sequences akin to those of the aerobactin promoter, in which the key T residues are conserved to various degrees. This mode of Fur-DNA interaction, in which new Fur molecules must necessarily bind adjacent hexamers through side-by-side oligomerization, explains the gradual physiological response observed in Fe 2ϩresponsive systems, because it would make possible an entire range of repression levels of iron-controlled promoters (12). The affinity for specific promoters would vary depending on the number of repeats present on each operator and the conservation of their sequences, thus generating a hierarchy of transcriptional responses depending on small changes in the iron status of the cell. Such an ability of Fur to control promoters through extensive DNA-protein interactions makes this protein to be mechanistically closer to general regulators than to specific transcriptional factors. In fact, because Fur is a Zncontaining protein (8), it is curious that the type of DNAprotein interactions reported here have certain reminiscence to the occupation of adjacent DNA sites by individual zinc fingers within eukaryotic transcription factors such as transcription factor TFIIIA (3,37).