Fur Activates the Expression of Salmonella enterica Pathogenicity Island 1 by Directly Interacting with the hilD Operator In Vivo and In Vitro

Previous studies have established that the expression of Salmonella enterica pathogenicity island 1 (SPI1), which is essential for epithelial invasion, is mainly regulated by the HilD protein. The ferric uptake regulator, Fur, in turn modulates the expression of the S. enterica hilD gene, albeit through an unknown mechanism. Here we report that S. enterica Fur, in its metal-bound form, specifically binds to an AT-rich region (BoxA), located upstream of the hilD promoter (PhilD), at position -191 to -163 relative to the hilD transcription start site. Furthermore, in a PhilD variant with mutations in BoxA, PhilD*, Fur·Mn2+ binding is impaired. In vivo experiments using S. enterica strains carrying wild-type PhilD or the mutant variant PhilD* showed that Fur activates hilD expression, while in vitro experiments revealed that the Fur·Mn2+ protein is sufficient to increase hilD transcription. Together, these results present the first evidence that Fur·Mn2+, by binding to the upstream BoxA sequence, directly stimulates the expression of hilD in S. enterica.


Introduction
Salmonella enterica is a bacterial pathogen that causes numerous diseases, ranging from gastroenteritis to systemic infections, in several hosts including humans. Moreover, it is one of the most important pathogens associated with food-borne illness worldwide [1,2]. An early step in the pathogenesis of non-typhoidal Salmonella species involves their ability to penetrate the intestinal epithelium. Invasion is mediated by the presence of a type III secretion system (T3SS), which is encoded on the tightly regulated Salmonella pathogenicity island 1 (SPI1) [3][4][5]. The main regulator of SPI1, HilA, directly activates the expression of the invF and prgH operons, which encode the components of the T3SS apparatus [6,7]. InvF facilitates the expression of several effector genes on SPI1 and elsewhere in the S. enterica genome [8][9][10]. In turn, hilA expression is under the control of other transcriptional activators, HilD and HilC, likewise encoded within the SPI1, and RtsA, located elsewhere in the chromosome [11][12][13]. These three transcription factors independently activate not only HilA expression but also each others' and their own, thus comprising a complex feed-forward regulatory loop [14]. Moreover, HilC and HilD can directly activate invF independently of HilA [11].
T3SS expression is also modulated by several distinct environmental signals [3,15], ensuring that the system is available by the time the bacterial pathogen reaches the distal small intestine, where invasion of the epithelial monolayer takes place [5]. For example, it is well known that osmolarity and bile salt concentration influence T3SS production by controlling hilD expression through the EnvZ/ OmpR and BarA/SirA two-component systems, respectively [16][17][18]. In addition, it has been shown that phosphate or Mg 2+ and Ca 2+ concentrations, sensed by the PhoR/PhoB and PhoP/PhoQ systems, respectively, are likewise involved in SPI1 regulation [3,19]. In both cases, the sensor system controls expression of the SPI1 repressor HilE, which is encoded by a gene located outside of SPI1. It has been suggested that the negative regulatory effects of HilE are exerted by its direct interaction with HilD, such that HilDmediated activation of hilA is prevented [20]. Another protein, H-NS, has also been described to control SPI1. This nucleic-acidassociated protein, binds to the promoter region of hilA, rtsA, hilD and hilC genes diminishing their expression [21,22].
Iron concentration is also associated with T3SS expression [23]. Fe 2+ is essential for bacterial development [24][25][26] and S. enterica is confronted with different free-iron concentrations during its infectious process. Normally, free Fe 2+ is scarce inside the host due its sequestration by several different cellular mechanisms [25,27]. However, in the lumen of the small intestine, where dietary iron is mainly absorbed [28], there is abundant free Fe 2+ . Accordingly, it has been reported that several of the Salmonella genes that are expressed when iron is scarce remain silent as long as the bacterium is confined the intestinal lumen [29]. The Fe 2+ concentration is thought to act as a signal that allows the pathogen to sense its location inside the host [23].
The Fur (ferric uptake regulator) protein is the main regulator of iron homeostasis in many bacteria. As a major regulator of gene expression, it not only controls genes involved in iron homeostasis but also ultimately coordinates intracellular iron levels with many other cellular processes [30]. In transcriptional and translational gene fusions, Fur was shown to activate SPI1 expression by increasing the amount of HilD [4,23]. Recent studies described the ability of Fur to modulate hilA expression by negatively controlling the levels of the H-NS global regulator [31]. In the presence of Fur?Fe 2+ , hns expression is repressed. The resulting decrease in the H-NS concentration reduces the repression that it exerts on the hilA promoter, thus allowing a rise in the expression of this gene [31,32]. Nevertheless, the previously described Fur?Fe 2+ -mediated activation of hilD expression remains unknown [23,31].
In Escherichia coli, Fur exhibits Fe 2+ -dependent DNA-binding activity to a specific sequence, namely the Fur box, located in the promoter region of genes directly repressed by Fur [33]. The Fur box is a 19-bp consensus sequence organized either as two inverted repeats separated by 1-bp, or as at least three contiguous hexamers, 59-NATWAT-39 (where N is any nucleotide and W is an A or a T), aligned in either a direct or an inverse orientation [33][34][35]. In the Fe 2+ -bound form, E. coli Fur represses genes involved in respiration, flagellar chemotaxis, the TCA cycle, glycolysis, methionine biosynthesis, phage DNA packaging, DNA synthesis, purine metabolism, and redox stress resistance [36][37][38][39]. Moreover, E. coli Fur has also an indirect positive effect on some genes by repressing the expression of the rhyB [40]. The absence of RhyB for pairing at the ribosomal binding site of mRNAs of genes positively regulated by Fur prevents their degradation by subsequent recruitment of the RNA degradosome [41].
Fur has been characterized in several other bacterial species [42][43][44][45] and other Fur-regulated pathways not related with sRNA have been described [46]. For instance, in Neisseria meningitidis, the Fur?Fe 2+ complex has been shown to act directly as a transcriptional activator once it binds to the promoter region of several virulence-associated genes [47]. Other Fur activation pathways have been reported in Pseudomonas aeruginosa, Yersinia pestis, Helicobacter pylori, and E. coli [48][49][50][51]. In S. enterica two sRNAs, RfrA and RfrB, have been identified. Both are homologous to E. coli RyhB and participate in the Fur-mediated positive control of genes such as sodB [23,52]. Nevertheless, neither RfrA nor RfrB mediates Fur control of hilD expression [23]. Moreover, direct control by Fur of the SPI1 repressor, hilE, has been ruled out [23].
To understand the molecular mechanism(s) that modulate hilD expression by Fur?Fe 2+ , we analysed whether the Fur protein of S. enterica serovar Typhimurium directly controls hilD expression. Our results show that Fur protein, in its metal-bound form, binds to an AT-rich operator located upstream of the hilD promoter region (P hilD ), and it acts directly as a transcriptional activator of hilD. These findings help to elucidate the role of iron in the regulation of SPI1 expression and provide the first evidence of a Fur-mediated direct activation mechanism in S. enterica.

Bacterial strains, plasmids and growth conditions
All bacterial strains and plasmids used in this work are listed in Table 1. Bacterial cultures were grown at 37uC in LB. When necessary, ampicillin (100 mg/ml), chloramphenicol (34 mg/ml), or kanamycin (150 mg/ml) was added to the bacterial culture. When needed 2,2-dipyrydil (DPD) was added to the medium at a concentration of 0.2 mM [31,53].

In silico searches for Fur binding sites
The 337-bp P hilD spanning 2247 to +90 (relative to the transcription start site [12]) was used for in silico searches with the Virtual Footprint online framework program [54]. The searches were carried out using the pre-existing P. aeruginosa (16-mer) matrix [54].

Protein purification
S. enterica fur was PCR-amplified using suitable oligonucleotides (Table S1), cloned into the pET15b expression vector, and transformed into E. coli BL21(DE3) pLysE strain. The Fur protein was purified using the TalonTM Metal Affinity Resin Kit (Clontech), as reported [55], and eluted from the affinity column by thrombin cleavage in buffer A (50 mM Bis-Tris/borate pH 7.5, 5 mM MgCl 2 , 10% glycerol) containing 500 mM NaCl. Fur was then loaded onto a Q-Sepharose equilibrated with buffer A containing 100 mM NaCl and eluted with a 600-1000 mM NaCl gradient. S. enterica Fur, which is free of E. coli H-NS protein, is expressed as dimers. The activity of purified S. enterica Fur was confirmed by electrophoretic mobility shift assays (EMSAs), which tested the ability of the protein to bind the promoter region of a confirmed Fur-regulated gene, foxA [56]. Figure S1 shows the Fur binding region in P foxA and the EMSA results using purified Fur protein.

Electrophoretic mobility shift assays
Appropriate DNA probes were obtained by PCR using suitable DIG-labeled oligonucleotides (Table S1). Fur EMSAs were done as previously described, with slight modifications [57]. In each case, 100 ng of each DIG-labeled DNA probe (20 nM) was incubated with increasing concentrations of Fur in buffer B (10 mM Bis-Tris/borate pH 7.5, 5% glycerol, 1 mM MgCl 2 , 40 mM KCl, 100 mg BSA/ml, 0.2 mg salmon sperm DNA/ml) with or without 100 mM MnCl 2 . For competitive assays, at least a 200-fold excess of either specific or non-specific non-labeled DNA was added. To assay the binding ability of Fur in its apo form, EDTA chelator was included in the binding mixture at a concentration of 1 mM. The binding mixture was incubated for 10 min at 37uC, after which the samples were separated by 5.5% polyacryamide gel electrophoresis (PAGE) in Bis-Tris buffer [57]. The DIG-labeled DNA-protein complexes were detected by following the manufacturer's (Roche) protocol.

Footprinting assay
For the footprinting assay, the 375-bp [a-32 P]-NcoI-HindII P hilD DNA (10 nM) from pUA1111 was incubated with increasing concentrations of Fur (1.5-100 nM) for 15 min at 37uC in buffer C (50 mM Bis-Tris/borate pH 7.5, 5% glycerol, 10 mM MgCl 2 , 1 mM MnCl 2 ). DNase I digestion was carried out by addition of the enzyme in binding buffer containing 5 mM CaCl 2 followed by incubation for 5 min at 37uC; the reactions were stopped by the addition of 25 mM EDTA. The samples were ethanol precipitated, resuspended in 6 ml of loading buffer, and fractionated on 6% denaturing (d) PAGE [58]. As the molecular weight marker, a G + A sequence reaction [59] was carried out and run in parallel with the corresponding footprinting reactions.

Construction of S. enterica mutant derivatives
S. enterica UA1888 and UA1889 strains, containing a Kan R cassette (inserted at position -192) and either wild-type (P hilD ) or the Fur BoxA mutant variant (P hilD* ), and the hilD knock-out mutant (UA1891) were constructed using the one-step PCR-based gene replacement method as described [60] and the appropriate oligonucleotides (Table S1). The PCR products were transformed in UA1875 carrying the pKOBEGA plasmid (Table 1). In the P hilD or P hilD* mutant derivative, transcription orientation of the kan gene was opposite that of the hilD gene, thus avoiding promoter interference. Indeed, the expression of neither hilD nor the downstream prgH genes was affected by the presence of the Kan R cassette in UA1888, and the expression levels were the same as those obtained using the SV5015 wild-type strain. All constructs were transferred as described [61] into the SV5015 wild-type strain or the null fur mutant derivative (Dfur, UA1880) by transduction using the P22int7(HT) bacteriophage and the suitable constructed strain as donor. The absence of the prophage in the transductants was determined by streaking them onto green plates, as described previously [62]. The obtained mutants were verified by PCR, using the appropriate primers (Table S1), and by nucleotide sequencing.

Quantitative RT-PCR assays
Quantitative real time RT-PCR (qRT-PCR) assays of hilD or foxA expression in different genetic backgrounds were carried out. To maximize the iron effect, bacteria were grown in LB medium, which contains saturated iron concentrations. When needed and to generate an iron-limiting environment DPD was added to the medium. All bacterial strains were overnight cultured and then diluted 1/100 in the appropriate media (with or without DPD) and incubated aerobically at 37uC. Once the bacterial culture reached OD 550 = 0.8, the cells were harvested and then the RNA was extracted using the Quiagen RNeasy kit following the manufacturer's instructions. The qRT-PCR assays were performed as previously reported [63] using suitable oligonucleotides (Table S1). It should be noted that the hilD oligonucleotide pair is located upstream the Kan R cassette insertion site in the hilD knock-out mutant (UA1891), thus allowing determination of the mRNA level in this genetic background. The results were normalized with respect to recA, a standard control gene not associated with the Fur regulon [64]. A change in the recA expression pattern was not observed in any of the genetic backgrounds assayed in this study (see Figure S2, in which 16S RNA is used as the standard). The relative expression level was defined as the ratio between the expression of hilD or foxA in each mutant derivative and that observed in the UA1888 strain containing a wild-type upstream Fur promoter region (P hilD ).

In vitro transcription assays
The P hilD upstream region, spanning 2247 to +90 (relative to the transcription start site), with either a wild-type BoxA or BoxA* mutant variant was cloned into pGEMH-T, generating plasmids pUA1111 and pUA1112, respectively (Table 1). In vitro transcription assays (run-off transcription) were performed using 10 nM of HindII-cleaved pUA1111 (containing P hilD ) or pUA1112 (P hilD* ), or with pGEMH-T plasmid DNA (P n ) as an unrelated Fur control. The pUA1111 vector was also used as a supercoiled DNA RNAP template. All of the DNAs were pre-incubated with increasing concentrations of Fur protein (1.5-200 nM) for 15 min at 37uC in buffer D (50 mM Tris-HCl pH 7.5, 5% glycerol, 5 mM MgCl 2 . 1 mM MnCl 2 , 2 mM spermidine, 10 mM DTT) in a 25-ml reaction. One unit of E. coli RNAP Es 70 holoenzyme (USB, Cleveland) and 0.5 mM of each rNTP (with [a-32 P]-rUTP) were added. The reactions were incubated for 60 min at 37uC and then stopped by the addition of 15 ml of loading buffer followed by heating to 75uC for 10 min. The in vitro generated transcripts from the linear templates, which contained the cloned P hilD or P hilD* , the vector promoter (P n ), and the supercoiled pUA1111 vector, were separated in 6% dPAGE, visualized, and quantified as described [65].

Results
Fur?Mn 2+ binds P hilD DNA with high affinity A search for putative Fur cognate sites in S. enterica SL1344 P hilD , spanning 2247 to +90 (relative to the transcription start site [12]), was carried out using the Virtual Footprint online framework [54]. Accordingly, two putative Fur boxes were predicted ( Figure 1A): (i) BoxA, at position 2191 to 2163, corresponding to an AT-rich region (dG + dC content ,15%, vs. 50% for the total genome) located , 100-bp upstream of the HilD and HilC binding sites and (ii) BoxB, at position 248 to 230 (dG + dC content ,31%), overlapping the promoter 235 element and situated , 30-bp downstream from the HilD and HilC binding sites [12] ( Figure 1A).
To validate the function of these putative Fur binding sites, the S. enterica Fur protein was purified and EMSA studies were performed using the P hilD region (position 2247 to +90) as probe ( Figure 1B). All EMSAs were done in buffer containing Mn 2+ as a common substitute of Fe 2+ due to its greater stability under aerobic conditions [33,47]. As shown in Figure 1B and Figure S1, in the presence of Mn 2+ , Fur dimers bound with high specificity and affinity to P hilD , with an apparent binding constant (K Dapp ) of about 4.561.5 nM, defined as the protein concentration necessary to complex 50% of labeled DNA. In contrast, in the absence of Mn 2+ or following the addition of EDTA, Fur was unable to form a complex with P hilD DNA ( Figure 1C), suggesting its metaldependent DNA-binding activity [66]. In addition, the Fur-Mn 2+ complex displayed a higher DNA affinity than Fur-Mg 2+ ( Figure  S1). Together these observations suggest that Fur binds to the P hilD promoter with the same characteristics as those reported when it acts as a repressor [33].
To test the specificity of the reaction, competition experiments were performed. A large excess (200-fold) of non-specific DNA (pGEMH-T DNA) was unable to compete with P hilD for Fur?Mn 2+ binding, whereas an excess of non-labeled P hilD DNA fully competed for binding with labeled P hilD DNA ( Figure 1D). It is therefore likely that at least one Fur?Mn 2+ binding site is present in P hilD , supporting the in silico predictions. Fur?Mn 2+ binds to BoxA DNA To elucidate which of the two putative Fur boxes, or even both, binds Fur?Mn 2+ , serial deletions of the P hilD region were obtained, generating FrgB, FrgC, and FrgD ( Figure 2A). As shown in Figure 2B, FrgD (spanning the 2247 to 2123 interval) but not FrgB (2170 to +90) or FrgC (216 to +90) bound Fur?Mn 2+ ( Figure 2B). It is therefore likely that: (i) Fur?Mn 2+ interacts with the region spanning 2247 to 2123 (containing the putative BoxA, 2191 to 2163), and (ii) Fur?Mn 2+ interacts neither with the core P hilD region nor with the 291 to 257 interval, which includes the HilD and HilC binding sites [12,21].
Fur?Mn 2+ recognizes BoxA DNA and spreads to adjacent regions DNAse I footprinting experiments were carried out to further determine the specific Fur?Mn 2+ binding site in the P hilD region. At low Fur?Mn 2+ concentrations (0.5 Fur dimers/P hilD DNA), only the 2189 to 2170 region, including BoxA, was protected from DNase I attack (Figure 3, lane 4). The BoxA site contains three copies of the hexameric 59-NATWAT-39 Fur consensus sequence separated from each other by 4-bp. Two hexamers (subsites I and III) are in the direct (R) orientation and one (subsite II) is in the inverse (r) orientation (R4-bpr4-bpR), conforming to a typical Fur box [33]. At limiting Fur?Mn 2+ concentrations, the protected sequence included subsites I and II spaced by 4-bp, but subsite III was poorly protected (see Figure 3). At sub-saturating and saturating Fur?Mn 2+ concentrations (2.5:1 to 10:1 Fur?Mn 2+ :P hilD DNA ratios), an extended DNase I -protected interval (from 2147 to 2219) was observed (Figure 3, lanes 1-3). Sites hypersensitive to DNase I attack were not apparent, suggesting that upon Fur?Mn 2+ binding no obvious major distortion of the DNA occurred. It is likely that, by cooperative interaction, Fur?Mn 2+ showed limited spread onto the 59 (, 30-bp) and 39 (, 23-bp) regions of P hilD DNA (Figure 3). Fur?Mn 2+ , under the concentrations used, halted before reaching the HilD, HilC, RstA binding region (291 to 257). Similar results were described in E. coli, when Fur?Mn 2+ acts as a transcriptional repressor [33,67,68].
To test the contribution of the AT-rich upstream operator (BoxA) to Fur recognition of P hilD DNA, a mutant BoxA DNA (P hilD* ) was constructed and then used in EMSA experiments ( Figure 4). As expected, S. enterica Fur specifically recognized P hilD DNA with wild-type BoxA in the upstream region but failed to bind P hilD* carrying mutations in the three subsites of the Fur box region ( Figure 4B).

Fur activates hilD gene expression in vivo
To determine whether Fur increases P hilD utilization in vivo, the transcription level of hilD was assayed in several S. enterica strains. Wild-type P hilD and the P hilD* variant, each with an upstream Kan R cassette, were integrated into their native locus, leading to strains UA1888 and UA1889, respectively (Table 1). It is worth noting that the mutations in BoxA* (P hilD* , Figure 4A) were located upstream of the HilD, HilC, RstA [12], and RNA polymerase (RNAP) binding sites ( [12], Figure 1A). Also, the presence of the Kan R cassette upstream of BoxA did not modify hilD expression, since the transcription level of this gene was similar to that obtained using the SV5015 wild-type strain ( Figure 5).  [12], are shown in green. The positions refer to the +1 transcription start site of the hilD gene [12]. The hilD coding region is striped. B. EMSA experiments using FrgA (P hilD ), FrgB, FrgC, and FrgD DNA probes (20 nM) from the P hilD region in the absence (2)  The hilD mRNA levels were analyzed by qRT-PCR and normalized with respect to the recA gene, which is not associated with the Fur regulon [64]. As a control, expression of the foxA gene, previously shown to be Fur repressed [56], was also measured under the same growth conditions. There was no evidence that the P hilD* mutation affected foxA expression ( Figure 5). In the absence of Fur or when iron was scarce (DPD addition), foxA expression levels increased (Figure 5), consistent with the ability of Fur?Fe 2+ to repress P foxA utilization [56]. Under ironsaturated conditions, hilD mRNA level was ,10-fold higher than under iron-limiting conditions (DPD addition) ( Figure 5). Similar results were observed when 1.5 mM EDTA was added to the media as a chelator (data not shown). The fact that hilD expression level in the UA1888 (P hilD , DPD addition), UA1889 (P hilD* ), and UA1890 (Dfur) strains were similar suggests that inactivation of BoxA (in P hilD* ) had no effect on basal expression of the gene.
Under iron-saturated conditions, the expression of hilD mRNA in the P hilD* mutant derivative was similar to that measured under iron-limiting conditions ( Figure 5), suggesting that P hilD activation does not occur by a mechanism involving transcriptional derepression. In the P hilD* mutant derivative, hilD mRNA levels were also similar to those obtained with the wild-type P hilD in a null Fur mutant strain ( Figure 5), implying that a wild-type BoxA sequence in cis is necessary for the increased accumulation of hilD mRNA.
There is a caveat to these findings, however, as there was no decrease in hilD expression in the absence of both HilD and Fur proteins, determined using transcriptional fusions [23]. This was confirmed by qRT-PCR experiments using the UA1892 (Dfur DhilD) strain, in which, as expected, hilD expression was increased by a factor of 1.260.18 with respect to the wild-type strain. Nonetheless, even though the absence of Fur or a decrease in the iron concentration resulted in a clear reduction in hilD expression ( Figure 5), both HilD and Fur appear to be necessary for full in vivo expression of the gene. This complex response might be explained by the fact that only when the HilD protein reaches a significant threshold it is able to activate the expression of hilA and further induce its own expression [69].

Fur?Mn 2+ activates hilD expression in vitro
To address whether the presence of Fur?Mn 2+ is sufficient to activate P hilD utilization, in vitro transcription experiments with linearized P hilD (pUA111) or P hilD* (pUA112) DNA were performed ( Figure 6). A single transcript band was obtained with HindIIlinearized DNA, indicating that hilD is expressed from a single promoter ( Figure 6A, lane 1). The length of the transcript was in full agreement with transcripts initiated at P hilD , as determined by primer extension [12], confirming that the 116-nt transcript was the genuine transcript from P hilD . Minor transcripts with small molecular masses were attributed to RNAP pausing sites since no obvious promoter sequences could be predicted in the putative upstream regions. The positions are related to the transcription start site (+1) [12]. The three BoxA subsites are boxed and enlarged, with arrows denoting their relative orientation. Abbreviations: 2, absence of Fur; C, a G + A sequence ladder of the DNA probe was used as molecular size marker. doi:10.1371/journal.pone.0019711.g003 In the presence of Mn 2+ and the absence of Fur, steady-state levels of the 116-nt transcripts from P hilD and P hilD* were similar (3763 arbitrary units, AU), suggesting that the BoxA mutations did not affect P hilD* utilization ( Figure 6B). In the presence of Mn 2+ , P hilD utilization, or accumulation of the 116-nt transcript, increased with increasing Fur concentration, with an optimum reached when , 2.5 Fur dimers/DNA molecule were added (12365 AU) ( Figure 6A). Similar levels of P hilD utilization were obtained using supercoiled DNA (P hilD sc) as template, ruling out any topological requirement for transcription activation ( Figure 6B). At sub-saturating Fur?Mn 2+ concentrations, P hilD utilization increased by more than 3-fold over the control without Fur?Mn 2+ (Figure 6). However, the addition of Fur?Mn 2+ did not significantly increase the levels of transcription of an unrelated promoter (Pv) (2765 AU) ( Figure 6B). P hilD* utilization was not significantly increased at Fur?Mn 2+ concentrations equal to or higher than those required to activate this promoter ( Figure 6B). As seen in Figure 6B, Fur?Mn 2+ , at subsaturating or half-saturating concentrations, interacted with BoxA, suggesting that the protein facilitates RNAP utilization of the P hilD region. It is therefore likely that Fur?Mn 2+ acts as a transcriptional activator of S. enterica hilD expression. Since transcription activation was not observed when BoxA was inactivated by mutations (P hilD* ), half-saturating Fur?Mn 2+ concentrations are apparently necessary for transcription activation of the hilD promoter in vitro. However, in the presence of , 20 Fur dimers/DNA molecule, similar activation was not observed ( Figure 6B). Since , 20 Fur dimers/ DNA molecule similarly did not affect the expression of P hilD* or an unrelated promoter (P v ) ( Figure 6), a contaminant RNase or any other non-specific effect can be ruled out as responsible for the reduced RNA synthesis at constant Mn 2+ and higher Fur concentrations. It could be hypothesized that Fur?Mn 2+ prevents P hilD activation based on its reported cooperative spreading. Thus, nucleoprotein assembly along the promoter region may interfere with, rather than stimulate, the interaction of RNAP with P hilD , returning hilD expression to its basal level.

Discussion
We show that the S. enterica hilD gene, whose product is the most important regulator of the HilA activator and therefore of SPI1 T3SS expression, contains a Fur binding site (BoxA) in the Figure 5. Relative hilD mRNA levels in several bacterial strains and iron concentrations. Relative hilD mRNA were estimated by qRT-PCR in P hilD (UA1888), P hilD* (UA1889), Dfur P hilD (UA1890), or wild-type (SV5015) strains grown in high (LB) or low (DPD) iron concentration, as described in the Materials and Methods section. To demonstrate that P hilD utilization is not affected by the upstream Kan R cassette, the hilD mRNA level in the wild-type SV5015 strain (wt) is also shown. The expression level of foxA, which is repressed by Fur [56], in all genetic backgrounds and conditions was determined as a control. For each condition, the relative gene expression levels were calculated as the ratio of each relative mRNA concentration with respect to that obtained in the isogenic wild-type strain (P hilD ) and normalized to that of the S. enterica recA gene. The mean value from three independent experiments (each in triplicate) is shown. doi:10.1371/journal.pone.0019711.g005 upstream region of P hilD (2191 to 2163). Fur, in its metal-bound form, bound with high affinity to BoxA in P hilD DNA but not to the BoxA mutant variant (BoxA*) in P hilD* . In vivo and in vitro experiments revealed that Fur bound to the upstream element of P hilD activated hilD expression, but did not activate transcription from P hilD* . Thus, metal-bound Fur appears to be a direct transcriptional activator of the S. enterica hilD gene. Being this the first evidence of Fur acting directly as an activator in this bacterium.
In most bacterial species, Fur?Fe 2+ is a transcriptional repressor, binding to cognate sites within position 235 to +12 in the promoter region [33,47,57,[70][71][72]. In these cases, transcription of the target promoters is blocked by steric hindrance rather than by preventing transcription elongation [46,66,73]. Fur?Mn 2+ bound to BoxA in P hilD , which is situated upstream of HilD, HilC (see Figure 1A [12]), or even RtsA binding sites [21], activates RNAP utilization of P hilD . A similar upstream location of the Fur cognate site has been described for Fur-activated genes in other microorganisms [46,47,51,74,75], suggesting a close relationship between the location of the Fur binding site in the promoter of the controlled gene and its role as an activator.
The majority of transcriptional activators, upon binding to their cognate upstream element adjacent to the core RNAP sites, either drive the recruitment of the latter to the target promoter or alter the conformation of the promoter DNA to facilitate RNAP loading [76][77][78][79]. Here we provide the first evidence of a direct mechanism of S. enterica hilD transcription activation by showing that Fur?Mn 2+ binds to a distal upstream-activating sequence. We thus propose that Fur?Fe 2+ is sufficient, in vitro, to increase RNAP recruitment. It is unlikely that Fur?Fe 2+ alone alters the DNA conformation, because DNase I hypersenstive sites were not observed. There are not evidences for Fur?Mn 2+ and RNAP interaction and activation via a looping mechanism, but Fur?Mn 2+ bound to its target site are sufficient for increased P hilD utilization. It can be envisaged, however, that the HilD regulation in vivo is complex as suggested the hilD expression results obtained in the double hilD fur mutant and the fact that RtsA, HilC and HilD recognize a sequence downstream Fur and apparently antagonize H-NS-and Hha-mediated repression, suggesting that these complex control region (2120 to 257) laid within BoxA (2191 to 2163) and the core promoter region (260 to + 10) [21,23].
The expression of SPI1 is known to be modulated by environmental signals, which indirectly control hilD transcription [3]. Among these signals, a relationship between extracellular iron concentrations and SPI1 expression has been suggested [23] and an indirect association of Fur and hilA expression through H-NS described [31]. In addition, several reports have shown that lowoxygen concentrations, such as those present in the intestinal lumen, increase SPI1 expression [4,16,80]. Free iron is scarce inside the host, but Fe 2+ is abundant in the intestinal lumen, where it is efficiently absorbed by intestinal epithelial cells [28]. Fur does not sense oxygen concentrations directly but is instead able to monitor the redox signal via the equilibrium between Fe 2+ and Fe 3+ [81]. It should be noted that the Fur?Fe 3+ complex is not functional and that, as Fe 3+ is insoluble, it cannot be translocated inside the cell [33,47].
This work describes the direct activation of SPI1 by Fur through its interaction with an upstream region on P hilD and thus adds new information to Fur SPI1 regulation models [21,23,31] (Figure 7). Taken together, our data shed light on the role of Fur in SPI1 control. Specifically, Fur is able to control, either directly (in the case of HilD) or indirectly through H-NS (in hilA, hilD, hilC, and rtsA), all the main regulators of SPI1. These results strengthen the relationship between S. enterica invasiveness and both iron and oxygen concentrations inside the host. Accordingly, the expression of hilD, hilA, hilC, and rtsA, and consequently that of T3SS1, should be stimulated once S. enterica reaches the epithelial surface, where iron concentrations are high and those of oxygen low. T3SS expression has been reported as essential for epithelial invasion [5]. Once Fe 2+ levels decrease or those of O 2 increase, the number of Fur?Fe 2+ complexes should diminish markedly, as should T3SS, which is no longer needed and must remain silent in subsequent steps of the infection process [5]. This sequence of events is supported by a previously published report in which, the epithelial invasiveness of Fur-defective mutants was shown, in an acidsensitive independent manner, to be lower than that of the wildtype strain [82]. Direct regulation by Fur of hilD expression would allow rapid signal transduction once the Fe 2+ concentration increases, situating P hilD in a similar hierarchic position as sRNAs that indirectly control other genes positively regulated by Fur. The reactions were incubated for 60 min at 37uC. Transcripts generated in vitro from the P hilD template DNA were separated in 6% dPAGE. The in vitro transcripts from P hilD and their lengths are shown. The sizes of the markers are indicated. B. Linear HindII-cleaved pUA1111 or pUA1112 DNA (10 nM) containing P hilD or P hilD* was pre-incubated with increasing concentrations of Fur (1.5-200 nM). Relative mRNA synthesis from P hilD (filled circles) and P hilD* (empty triangles) in the absence or presence of increasing Fur was compared and is denoted in arbitrary units (AU). Also, the pUA1111 vector containing P hilD (P hilD sc) (open squares) and the pGEMH-T plasmid vector promoter (P v ) (empty circles) were used under the same conditions as supercoiled DNA and the non-specific control, respectively. doi:10.1371/journal.pone.0019711.g006 Figure S1 A. SDS-PAGE showing the Fur purified protein. Lanes 1 and 2 correspond to non-induced and IPTG-induced cell crude extracts of BL21(DE3)pLys containing the S. enterica fur gene cloned in the pET15b vector. Lane 3 is the purification fraction containing the Fur native protein after thrombin digestion. B. Scheme of P foxA indicating the location of the Fur binding site in blue. The ATG start codon is indicated in bold. C. EMSA performed using DIG labeled P foxA probe (20 nM) and the purified Fur protein at increasing concentrations (2.5, 12.5, 50, and 187 nM). Lane (-) indicates the mobility of the DNA probe without Fur in the binding mixture. (TIF) Figure S2 Fur-Mn 2+ binds with high affinity to P hilD DNA. The 375-bp [a-32P]-NcoI-HindII DNA (2 nM) fragment containing P hilD was incubated with increasing Fur concentrations (3-400) for 15 min at 37uC in buffer A (50 mM Bis-Tris/borate buffer pH 7.5, 5% glycerol, 10 mM MgCl 2 , 1 mM MnCl 2 ) or E (50 mM Bis-Tris/ borate buffer pH 7.5, 5% glycerol, 1 mM MgCl 2 , 0.1 mM MnCl 2 ). The absence of a component is indicated by -; FD, protein-free P hilD DNA; IC, intermediate complexes; PD, protein-DNA complexes. (TIF) Figure S3 qRT-PCR assays of recA expression in the different genetic backgrounds used in this work. For each condition, the relative recA expression levels were calculated as the ratio of its mRNA concentration with respect to that obtained in the isogenic wild-type strain (P hilD ) and normalized to that of the S. enterica 16S RNA. The mean value from three independent experiments (each in triplicate) is shown. (TIF)