Roles of the EnvZ/OmpR Two-Component System and Porins in Iron Acquisition in Escherichia coli

The work presented here solved a long-standing paradox of the negative effects of certain missense alleles of envZ, which codes for kinase of the EnvZ/OmpR two-component system, on the expression of ferric uptake genes. The data revealed that the constitutive envZ alleles activate the Feo- and OmpC-mediated ferrous uptake pathway to flood the cytoplasm with accessible ferrous iron. This activates the ferric uptake regulator, Fur, which inhibits ferric uptake system but cannot inhibit the feo operon due to the positive effect of activated EnvZ/OmpR. The data also revealed the importance of porins in iron homeostasis.

iron, bacteria have developed high-affinity ferric iron chelators called siderophores to capture, solubilize, and deliver insoluble iron into the cell (4). Unlike the Fe 3ϩ transport system, which requires a number of proteins involved in siderophore synthesis and Fe 3ϩ siderophore acquisition, the Fe 2ϩ transport system appears to consist of mainly one protein, FeoB (5). The FeoB protein is synthesized from the feoABC operon, whose expression is activated by Fnr, an anaerobic transcriptional regulator (5). FeoB is a highly conserved, 773-residue inner membrane protein that contains several GTPbinding motifs (6)(7)(8). In the absence of FeoB or FeoA, Fe 2ϩ uptake is either virtually abolished (ΔfeoB) or mildly reduced (ΔfeoA) (5). The function of FeoC, which is present only in members of the Enterobacteriaceae family, is unknown (7,8). FeoB and its homologs are required for full virulence in many bacteria, including Escherichia coli (9), Salmonella enterica serovar Typhimurium (10,11), and Helicobacter pylori (12).
Fur (ferric uptake regulator) in E. coli and its orthologs in many Gram-negative and Gram-positive bacteria are the master regulator of genes encoding both ferric and ferrous iron acquisition functions, as well as siderophore synthesis and uptake (13,14). Cells lacking Fur experience iron overload that causes oxidative damage and mutagenesis (15). Fur-regulated genes contain one or more Fur-binding sites around the Ϫ35 and Ϫ10 regions of the promoter, often referred to as the Fur boxes (16,17). Fur uses Fe 2ϩ as a cofactor: when the level of available Fe 2ϩ increases in the cell, it binds to Fur and enhances its affinity for DNA by almost 1,000-fold (2). The active Fur-Fe 2ϩ complex then binds to a Fur box and represses transcription of the iron acquisition gene. RyhB is a small regulatory RNA whose transcription is also repressed by Fur-Fe 2ϩ (18). Consequently, when Fur is active, the levels of RyhB are low, resulting in stabilization and translation of over a dozen mRNAs encoding nonessential iron utilization proteins, including those that store iron (Bfr), detoxify superoxide (SodB), and catalyze steps of the tricarboxylic acid cycle (AcnA and SdhCDAB) (19). Thus, excess Fe 2ϩ activates Fur to halt further iron uptake and at the same time, promotes the utilization of Fe 2ϩ , and inversely, low intracellular iron level induces iron uptake and utilization (20). Recent genome-wide analyses revealed a more comprehensive profile of Fur and RyhB regulons (21,22).
Whereas Fur and RyhB are the principal determinants of iron homeostasis in E. coli, evidence exists supporting the involvement of some two-component signal transduction systems (TCS) in iron homeostasis. EnvZ and OmpR are the archetypal TCS in E. coli, where EnvZ serves as a sensor kinase and OmpR as a response regulator (23). They respond to medium osmolarity and influence the expression of OmpC and OmpF, the two major porins that facilitate the diffusion of small hydrophilic solutes (ϳ600 Da) across the outer membrane (24). OmpC is preferentially expressed in high osmolarity, whereas OmpF expression is favored in low osmolarity (25). Microarray data from an ΔompR ΔenvZ background showed a significant increase in the expression of a number of Fur-regulated genes, particularly those involved in enterobactin siderophore synthesis and transport (26). Over 3 decades ago, Lundrigan and Earhart (27) reported that in a perA (envZ) mutant background, the levels of three iron-regulated proteins were significantly reduced. These authors suggested that this could be due to a posttranscriptional defect. Later, it was speculated that this inhibition could be due to the indirect effects of envZ/ompR, leading to alterations in the structure and diffusion properties of the outer membrane (28). While characterizing revertants of an E. coli mutant defective in outer membrane biogenesis, we discovered several pleiotropic envZ alleles conferring an OmpC ϩ OmpF Ϫ LamB Ϫ phenotype (29). These alleles were hypothesized to biochemically lock EnvZ into a conformation that causes increased OmpR phosphorylation. This activated EnvZ/OmpR state is thought to enable OmpR to bind to promoters with weak OmpR-binding affinities. One such pleiotropic envZ allele, envZ R397L , was characterized in detail (29). The preliminary whole genomic microarray analysis of the envZ R397L mutant carried out in our laboratory found that the largest group of genes (Ͼ50) affected by the activated EnvZ R397L /OmpR ϩ background belonged to the Fur regulon (30; unpublished data).
In this study, we show that EnvZ R397L exerts its effect on the Fur regulon in part by Expression of OmpC is activated constitutively in the envZ R397L background, while that of OmpF and LamB is severely inhibited (29). To determine whether OmpC is somehow involved in the envZ R397L -mediated downregulation of fecA, fepA, fhuA, and fhuF, we examined their transcript levels in the ΔompC and ΔompC envZ R397L backgrounds. Remarkably, without OmpC, envZ R397L was unable to exert any significantly negative effect on fepA, fhuA, and fhuF expression, while the effect on fecA diminished from 10-fold in the presence of OmpC to Ͻ2-fold without OmpC (Fig. 1). Interestingly, the levels of all four transcripts went up in ΔompC cells (Fig. 1). We theorize that without OmpC, diffusion of Fe 2ϩ into the cell is decreased and the less active Fur fails to fully repress fecA, fepA, fhuA, and fhuF expression.
If the intake of Fe 2ϩ by OmpC porin increases active Fur-Fe 2ϩ levels, then the absence of FeoB, the Fe 2ϩ -specific iron transporter, should also interfere with this activation and abrogate the Fur-mediated effects of envZ R397L on fecA, fepA, fhuA, and fhuF. Indeed, just like in the ΔompC background, fecA, fepA, fhuA, and fhuF transcript levels went up in the ΔfeoB background, and envZ R397L could either no longer impose a significant negative effect (fepA, fhuA, and fhuF) or the effect was significantly reduced (fecA).
Effects of envZ R397L on the ferrous transport system. The data presented in Fig. 1 show the involvement of the FeoB ferrous iron transporter and OmpC porin in envZ R397L -mediated downregulation of the ferric iron transport system. While ompC expression increases in the envZ R397L background (29), the status of the feo operon in this background is unknown. The feo operon is under the negative control of Fur (5). Consequently, if higher Fur-Fe 2ϩ activity is present in the envZ R397L background, as we have suggested above, then the expression of the feo operon, like that of fecA, fepA, fhuA, and fhuF, should also be inhibited. This, however, will be incongruent with our data showing envZ R397L 's dependence on feoB for its effects. We therefore hypothesized that feo expression, like that of ompC, is activated by envZ R397L to such a degree that it more than compensated for the feo downregulation by increased Fur-Fe 2ϩ activity.
To test these possibilities, we analyzed feoA and feoB transcript levels in different genetic backgrounds by RT-qPCR (Fig. 2). Note that feoABC are part of a contiguous FIG 2 Determination of the relative gene expression of feoA and feoB by RT-qPCR. RNA was isolated from bacterial cultures grown to mid-log phase. Relative quantification of transcripts in various genetic backgrounds was performed using the 2 -ΔΔCT method, with ftsL and purC serving as reference genes. The relative fold changes in gene expression and error bars representing standard deviations are shown. The bacterial strains used included RAM1292 (wild type), RAM1541 (envZ R397L ), RAM2697 (Δfur), RAM2698 (envZ R397L Δfur), RAM2699 (ΔompC), RAM2700 (envZ R397L ΔompC), RAM2701 (ΔfeoB), and RAM2702 (envZ R397L ΔfeoB). operon and therefore likely expressed from a polycistronic message. Consequently, feoA and feoB transcript analysis probes their respective coding regions in a polycistronic message. In the EnvZ R397L background, feoA and feoB transcript levels went up dramatically over those in the envZ ϩ control strain. As expected, their levels also went up in the Δfur background. Interestingly, in the envZ R397L Δfur background feoA and feoB transcript levels increased well above those in the individual mutation backgrounds, indicating that envZ R397L and Δfur act independently and synergistically to enhance feo expression. Again, these observations were recapitulated using the chromosomally integrated feo::lacZ fusion (Fig. S1). These data support our hypothesis that envZ R397L activates feo expression in a fashion that counteracts repression by higher levels of Fur-Fe 2ϩ .
We then examined the effects of envZ R397L on feoA and feoB transcript levels in the absence of OmpC or FeoB. Without OmpC or FeoB, a modest 2-fold increase in feoA and feoB (ΔompC) or feoA (ΔfeoB) transcripts was observed (Fig. 2). We interpret this to reflect a modest relief in the Fur-mediated repression of the feo operon, since we have already implicated OmpC and FeoB in the ferrous iron transport and increase in Fur-Fe 2ϩ levels ( Fig. 1). The presence of envZ R397L in the ΔompC or ΔfeoB background led to an increase in feoA and feoB, or feoA transcripts, respectively, in a synergistic fashion, which is likely due to the simultaneous activation of feo expression by envZ R397L and a modest decrease in the Fur-mediated repression of feo from the absence of OmpC and FeoB. These data showed that envZ R397L inhibits ferric transport pathway but activates ferrous transport pathway.
Intracellular iron levels in the envZ R497L mutant. The OmpC/Feo-mediated increase in Fur-Fe 2ϩ activity in the envZ R397L background implies that the cytoplasm of the envZ R397L mutant contains higher levels of accessible iron than that in the cytoplasm of the EnvZ ϩ cell. To test this directly, we measured the intracellular pool of accessible iron by whole-cell electron paramagnetic resonance (EPR) spectroscopy, a method established in the Imlay laboratory (32). The data presented in Fig. 3 show that the wild-type (EnvZ ϩ ) strain had 32 M of accessible intracellular iron. Expectedly, this level rose 4-fold to 120 M in the Δfur mutant. Remarkably, the level of accessible iron in the envZ R397L was also very high (135 M) and remained high in the Δfur envZ R397L double mutant (105 M), thus supporting the notion that a higher pool of accessible iron in the envZ R397L background is responsible for the higher levels of active Fur-Fe 2ϩ .
Next, we tested whether the FeoB-mediated ferrous transport pathway is responsible for the elevated level of accessible iron in the envZ R397L mutant. The accessible iron level in the ΔfeoB mutant was 20 M or 35% less than the parental feoB ϩ strain (Fig. 3), explaining the observed deregulation of the Fur regulon in the ΔfeoB mutant ( Fig. 1 and 2). Strikingly, without feoB, envZ R397L failed to increase intracellular iron levels (Fig. 3), thus confirming the involvement of the FeoB-mediated ferrous transport in elevating the intracellular pool of iron, which, in turn, would increase Fur-Fe 2ϩ levels and repress expression of fecA, fepA, fhuA, and fhuF. As described below, EnvZ/OmpR play a more direct role in activating feo expression to overcome the Fur-mediated downregulation.
Effects of envZ R397L on fepA and feo requires phosphorylated OmpR. Previously, it was shown that the pleiotropic effects of the mutant envZ allele, envZ473 with its V241G substitution, is mediated through OmpR (33). In that study, the authors did not analyze the iron regulon. Here, we sought to test whether the effect of envZ R397L on iron regulon requires functional OmpR. We used a missense allele of ompR with a D55Y substitution, which confers a null phenotype with respect to ompC and ompF expression, presumably due to the inability of the mutant OmpR to be phosphorylated. The conserved D55 residue of OmpR is the site of phosphorylation (34). The ompR D55Y allele was isolated in a fepA::lacZ envZ R397L background among Lac ϩ revertants (R. Misra, unpublished data). Using a linked chloramphenicol resistance (Cm r ) marker, we transduced the ompR D55Y envZ R397L mutations into a feo::lacZ background so that the effects of the mutant ompR and envZ alleles on feo expression can be determined. It is worth noting that although ompR/envZ are highly linked to the feo operon, we were able to construct the above strain since feo::lacZ is marked by the kanamycin resistance (Km r ) gene and the mutant ompR/envZ alleles produce a distinct porin phenotype.
Data presented in Fig. 4 show that envZ R397L reduced fepA::lacZ expression ϳ4-fold, whereas ompR D55Y abolished this effect of envZ R397L . Likewise, the presence of envZ R397L elevated feo::lacZ expression 5-fold and again ompR D55Y abolished this increase in feo expression. Curiously, feo::lacZ expression in the ompR D55Y envZ R397L background was slightly lower than that seen in the wild-type background, suggesting a role for functional OmpR in the expression of the feo operon. Together, these data show unambiguously that the negative and positive effects of envZ R397L on fepA and feo, respectively, require functional OmpR.
Direct regulation of feoABC operon by EnvZ/OmpR. The data in Fig. 2 and 4 showed a dramatic increase in the feo transcript/transcription levels in the envZ R397L / ompR ϩ background. This could be due to the direct regulation of feo by OmpR or an effect of an OmpR-controlled factor on the feo promoter or feo transcript. We took cues from an earlier publication that showed overexpression of RstA, the response regulator of the RstB/RstA TCS, upregulated feoB expression and repressed the Fur regulon in Salmonella Typhimurium (35). Electrophoretic mobility shift assays (EMSAs) showed direct binding of RstA to the feo promoter sequence (35). Moreover, these authors identified the "RstA motif" (TACA-N 6 -TACA) upstream of the S. Typhimurium feoA gene of the feo operon (35). Although OmpR recognition sequences are quite degenerate (36,37), one of the motifs-GTTACANNNN-resembles that of RstA (Fig. 5A). Indeed, both RstA and OmpR regulate some of the same genes by binding to overlapping promoter sequences (38). Our initial assessment detected two potential sequences (Ϫ294)-TTAT CAtttcaTTAACA-(-278) and (Ϫ165)-CCAACAttcgCACACA-(-150) upstream of the feoA ATG codon that might contain both RstA and OmpR binding motifs (Fig. 5A).
EMSA was carried out to test whether OmpR can bind directly to the feo promoter region. The coding region of ompR was cloned into an expression vector, pET24d(ϩ). To aid in protein purification, six consecutive histidine codons were included at the 3= end of the gene during cloning, and the protein was purified to near homogeneity by metal affinity chromatography (Fig. S3). The purified protein was used directly without in vitro phosphorylation. Using biotinylated primers, two DNA templates of the feo regulatory region, encompassing the predicted OmpR binding motifs, were amplified by PCR (Fig. 5A). As a positive control for OmpR binding, two ompC DNA fragments were also included for EMSA (Fig. 5B). No DNA gel shift occurred with the smaller feoB DNA fragment containing one of the predicted OmpR binding motifs (Fig. 5C). However, the larger feo promoter fragment, containing the upstream predicted OmpR binding motif, displayed shifts after incubation with purified OmpR 6His (Fig. 5C). Consistent with these in vitro data, we found that overexpression of OmpR 6His from a pBAD24 replicon increased feo::lacZ expression 2-fold (from 140 Ϯ 8 Miller units in the pBAD24 vector containing strain to 296 Ϯ 25 Miller units in the strain containing pBAD24-ompR 6His ). OmpR bound to the ompC promoter fragment containing the high-affinity OmpR-binding motif C1 (39), but not with the one containing the partial C2 and the entire C3 motif (Fig. 5C). Incidentally, only the ompC fragment, containing all three OmpR motifs, expressed the promoterless lacZ gene in an OmpR-dependent manner (Fig. S4), thus corroborating the EMSA data. Together, these data indicated that OmpR positively regulates feo expression by directly binding to the feo promoter region.
Role of porins in iron homeostasis. The data in Fig. 1 and 4 revealed a possible mechanism by which a pleiotropic allele of envZ downregulates the ferric transport systems by employing the OmpC/FeoB-mediated ferrous transport pathway. While these data implicated EnvZ/OmpR and OmpC in iron transport, the use of a pleiotropic envZ allele may have created an unnatural genetic environment in which EnvZ/OmpR and porins become involved in iron homeostasis. To eliminate this possibility, we determined the roles EnvZ/OmpR and porins in iron transport using the null alleles of Roles of EnvZ/OmpR and Porins in Iron Acquisition ® ompR and the porin genes. Before testing their roles, we disabled the high-affinity ferric transport system, since porins likely mediate iron transport by simple diffusion of ferrous or small iron-chelated compounds, and this passive activity of porins will likely be masked by the high-affinity iron transport system. In E. coli, the high-affinity iron transport principally involves a ferric chelator, enterobactin, and TonB that interacts with the outer membrane iron receptors for the release of chelator-Fe 3ϩ complexes bound to the receptor. Consequently, we disabled the ferric iron transport by deleting aroB, tonB or both. The aroB gene encodes 3-dehydroquinate synthase, which is required for the second step of the chorismate pathway in the synthesis of enterobactin, aromatic amino acids, and other important compounds (40).
We first determined the iron dependency of wild-type, ΔaroB, ΔtonB, and ΔaroB ΔtonB strains by growing them on Lysogeny broth agar (LBA), LBA supplemented with 40 M FeCl 3 and LBA containing 200 M 2,2=-dipyridyl (DP), a synthetic iron chelator (Fig. 6). Bacterial growth in the absence of aroB was unaffected on LBAϩFeCl 3 or LBA ( Fig. 6A and B). However, significant growth impairment of the ΔaroB strain occurred on LBAϩDP plates (Fig. 6C), reflecting the loss of a major, enterobactin-mediated iron Plus and minus signs denote the presence and absence, respectively, of OmpR in the reaction mixture prior to gel electrophoresis. Gels were electroblotted, and DNA bands were detected by treating membranes with stabilized streptavidin-HRP conjugate, followed by luminol/enhancer and stable peroxide. Arrows point to positions of unshifted DNA fragments. transport system. In contrast to ΔaroB, the deletion of tonB impaired bacterial growth even on LBA (Fig. 6B), which contains around 6 M iron, and completely prevented growth on LBAϩDP medium (Fig. 6C). The ΔtonB strain grew like WT on LBAϩFeCl 3 , showing that the growth impairment of this strain on LBA was due to low accessibility to iron. Interestingly, growth of the ΔaroB ΔtonB double mutant improved slightly on LBA compared to the ΔtonB strain ( Fig. 6B) but ceased again on LBAϩDP (Fig. 6C). An improvement in growth of the double mutant compared to the ΔtonB strain on LBA may be due to the absence of extracellular enterobactin-Fe 3ϩ complexes, which, when allowed to accumulate outside the ΔtonB cells, would sequester iron from the medium and further exacerbate growth defects (41). Because of the greater growth dependence of the ΔtonB and ΔtonB ΔaroB strains on external iron sources than the ΔaroB strain, we selected the former two genetic backgrounds to examine the effects of EnvZ/OmpR and porins in iron transport. It is worth noting that we did not determine bacterial growth rates by monitoring growth of liquid cultures because the ΔtonB strain frequently reverts without supplemented iron, and these faster-growing revertants take over the population to artificially display better-than-expected growth.
We employed two different null ompR alleles, ompR101 and ΔompR::Km r , both of which produce the OmpC -OmpFphenotype. The ompR101 allele was transduced into a ΔtonB background using a linked tetracycline resistance (Tc r ) marker, malPQ::Tn10, while ΔompR::Km r was transduced directly using the Km r gene that replaced the deleted ompR gene. Although both ompR alleles could be transduced in the ΔtonB strain when transductants were selected on LBAϩFeCl 3 containing appropriate antibiotics, the resulting null ompR ΔtonB transductants grew poorly compared to the ompR ϩ ΔtonB strain (Fig. 7A, sectors 4 and 5). In contrast, ompR101 and ΔompR::Km r severely compromised growth of the ΔtonB strain on LBA not supplemented with FeCl 3 (Fig. 7B, sectors 4 and 5). Similar to the ΔtonB ompR101 strain, we were able to construct the ΔaroB ΔtonB ompR101 strain on LBAϩFeCl 3 medium, where it grew poorly (Fig. 7A,  sector 8) but not as poorly as on LBA without FeCl 3 , where the strain failed to form Roles of EnvZ/OmpR and Porins in Iron Acquisition ® single colonies (Fig. 7B, sector 8). These observations pointed to a critical role for the EnvZ/OmpR TCS in iron transport in the absence of the high-affinity iron transport system.
Although the porin genes are the main targets of the EnvZ/OmpR regulatory system, transcription of other genes is also affected either directly or indirectly in the ompR-null mutant (26). Therefore, to establish unambiguously the importance of porins in iron transport, we attempted to delete the porin genes in a background devoid of the high-affinity transport system. In the ΔtonB background, the deletion of ompC or ompF individually did not significantly influence growth on LBA ( Fig. 7C and D, compare sectors 3 and 4 to sector 2). Strikingly, however, we failed to delete ompC and ompF simultaneously, via P1 transduction of ΔompF::Km r and ΔompC::Cm r alleles, in the ΔtonB background even when transductants were selected on LBAϩFeCl 3 plates carrying appropriate antibiotics. In contrast, when the ΔtonB ΔompC or ΔtonB ΔompF double mutant was first complemented with a plasmid expressing one of the porin genes, the uncomplemented porin gene from the chromosome could be readily deleted by P1 transduction. The plasmid-complemented triple mutants displayed growth behavior similar to the uncomplemented ΔtonB ΔompC and ΔtonB ΔompF double mutants on LBAϩFeCl 3 or LBA ( Fig. 7C and D, compare sectors 3 and 4 with sectors 7 and 8). It is worth noting that the ΔtonB ΔompC and ΔtonB ΔompF strains were not defective in P1 transduction, since drug resistant markers not associated with the porin genes or their regulators could be transduced readily into these strains. Moreover, unlike the ΔtonB strain, in the wild-type and ΔaroB backgrounds the ompC and ompF genes could be deleted simultaneously without causing iron dependency or a significant growth defect (Fig. S5). These data indicated for the first time that the OmpC and OmpF porins play a critical role in iron intake when the high-affinity iron transport system is blocked.

DISCUSSION
Although the EnvZ/OmpR TCS is classically associated with the regulation of the OmpC and OmpF porins in response to medium osmolarity (23), recent transcriptomics and chromatin immunoprecipitation analyses showed that it is a global regulatory system (26,37,42). Indeed, missense alleles of envZ, called perA and tpo, isolated over 3 decades ago, were shown to also influence nonporin regulons, including pho and mal (43)(44)(45). A separate study revealed that perA lowered the expression of three ironregulated proteins without an apparent reduction in the rate of enterobactin secretion (27). This led these authors to suggest that the effect of the perA (envZ) allele on the expression of iron-regulated proteins is most likely posttranscriptional (27).
In the present study, we sought to resolve the mechanism by which the activated EnvZ/OmpR TCS reduces expression of genes involved in iron homeostasis and determine the role of porins in iron acquisition. We used the envZ R397L allele, which is phenotypically similar to the pleotropic perA and tpo alleles of envZ, i.e., in the envZ R397L background, OmpC levels go up, while those of OmpF and LamB go down dramatically (29). The RT-qPCR (this work) and the whole-genome microarray data (30; unpublished data) showed that in the presence of envZ R397L the transcript levels of several Furcontrolled genes, including fecA, fepA, fhuA, and fhuF, decreased significantly. In the case of fhuA and fhuF, the effects of envZ R397L required Fur, while expression of fecA and fepA was still reduced by envZ R397L in the absence of Fur. These observations indicated the involvement of at least two different mechanisms by which envZ R397L affected iron regulon. In support of the Fur-dependent mechanism, the whole-cell EPR data confirmed the presence of significantly elevated levels of accessible iron in the envZ R397L strain. Several observations supported the hypothesis that in the envZ R397L mutant, FeoB and OmpC are responsible for increased intracellular Fur-Fe 2ϩ level (Fig. 8). First, unlike the expression of genes involved in ferric iron transport or metabolism, expression of the feoAB genes involved in ferrous iron transport went up dramatically in the envZ R397L background. This increase in the expression of the ferrous iron transport system had an adverse effect on the ferric iron transport system, since the absence of FeoB, the ferrous permease, abolishes or significantly reduces the negative effects of envZ R397L on ferric transport/metabolic genes. Second, like FeoB, the absence of OmpC (envZ R397L already severely represses ompF expression [29]) largely negated the inhibitory effects of envZ R397L on ferric transport/metabolic genes. Because the single deletion of feoB or ompC and the simultaneous deletion of feoB and ompC reversed the effects of envZ R397L on fecA, fepA, fhuA, and fhuF to the same extent, it indicated that FeoB and OmpC must act in the same pathway to transport ferrous iron into the cell and elevate Fur-Fe 2ϩ levels. Third, the absence of FeoB or OmpC in an EnvZ ϩ background caused derepression of six Fur-controlled genes, indicating that the ferrous iron transport pathway is active under our experimental conditions and that envZ R397L enhances this pathway to achieve its inhibitory effects on the ferric transport system. Lastly, we provided direct evidence of excessive iron inside the envZ R397L mutant by whole-cell EPR spectroscopy measurements, which showed that, as in the Δfur mutant, the level of accessible iron in the envZ mutant rose 4-fold over that present in the parental strain. Moreover, this increase in the intracellular free pool of iron in the envZ R397L mutant was dependent on FeoB. From these observations, we conclude that the upregulation of the OmpC-FeoB ferrous iron transport pathway by envZ R397L elevates the intracellular Fur-Fe 2ϩ level, which, in turn, represses the expression of iron-regulated genes (Fig. 8). These effects of envZ R397L required functional OmpR since the presence of ompR D55Y , which confers a null phenotype, neutralized all envZ R397L phenotypes.
Whereas envZ R397L -mediated reduction in fhuA and fhuF transcript levels required Fur, the effects of envZ R397L on fepA and fecA transcripts did not. This suggested the existence of another regulatory mechanism responsible for the envZ R397L -mediated downregulation of fepA and fecA that did not involve Fur. Previous studies showed that the plasmid-mediated overexpression of OmrA and OmrB small RNAs, whose expression is under the EnvZ/OmpR control, can downregulate fepA and fecA transcript levels (31,46). We have shown that envZ R397L increases OmrA expression almost 10-fold (29). This increase in OmrA expression could contribute to the downregulation of fepA and fecA. However, the fact that deleting ompC or feoB in a Fur ϩ background abolishes envZ R397L -mediated downregulation of the ferric iron transport genes suggests that envZ R397L -mediated increase in OmrA and OmrB levels contributes little, if any, to fepA and fecA repression. Consistent with this notion, we found that deletions of ΔomrA and ΔomrB failed to reverse the negative effect of envZ R397L on fepA or fecA (Fig. S2). We conclude that a mechanism independent of Fur and OmrA and OmrB must also exist for the envZ R397L -mediated downregulation of fepA and fecA. A direct role of EnvZ/ OmpR has not been ruled out.
As stated above, unlike the ferric transport genes, expression of the ferrous transport genes feoAB, which are also under the control of Fur, went up in the envZ R397L background. At first glance, this appears inconsistent with the notion that an increase in the Fur-Fe 2ϩ level by envZ R397L should also decrease feoAB expression. Our data suggest that envZ R397L overcomes the repressive effect of Fur-Fe 2ϩ on feoAB by activating their expression. Moreover, because feoAB expression in the envZ R397L background increases dramatically without Fur, it shows that Fur-Fe 2ϩ does repress feoAB in the envZ R397L background, but the positive effect of envZ R397L on feoAB expression overwhelms the negative effect of Fur on these genes. The EMSA data showed that purified OmpR binds to the feo regulatory region containing a putative OmpR binding site, indicating that OmpR directly activates feoAB expression. Interestingly, the wholegenome microarray data from an ΔompR ΔenvZ (porin-minus) strain showed that feoB transcript levels decreased 2-fold, whereas those of fepA and fecA increased 2-fold (26). These observations are consistent with our proposal that the EnvZ/OmpR TCS directly stimulates the FeoB-OmpC pathway to increase intracellular Fe 2ϩ levels and thus active Fur-Fe 2ϩ complexes, which then downregulate the expression of the ferric iron transport genes.
The proposed role of EnvZ/OmpR in iron homeostasis is similar to that suggested for RstA in S. Typhimurium (35). The authors found that overexpression of RstA increased feoAB expression and repressed fhuA and fhuF expression. A RstA binding site was identified in the feo promoter and the EMSA data confirmed that RstA bound there (35). The RstA binding motif "TACAtntngtTACA" resembles that of OmpR's "GTTACAnnnnG TTACA" and not surprisingly, both proteins regulate overlapping genes by binding to the similar sequences (38). Our EMSA data showed that OmpR binds to the feo promoter region. Specifically, it binds to a DNA fragment containing the sequence "ttATCAtttcattAACA" located 278 bp upstream of the start codon of feoA. The OmpR binding studies carried out here involved the purified protein not modified by in vitro phosphorylation. Therefore, it is possible that stronger binding and/or additional binding sites may be discovered with phosphorylated OmpR. It is worth noting that in previous EMSAs, unphosphorylated RstA from S. Typhimurium and E. coli was shown to bind to their target promoter sequences (35,47). Further work will be required to identify the exact binding sequences and to determine whether OmpR and RstA bind to the same, overlapping, or distinct regulatory sequences of the feo operon.
Our work also revealed for the first time the essential role of OmpC and OmpF porins in iron acquisition when the TonB-dependent ferric transport pathways are inoperative. The absence of OmpC or OmpF produced no growth defects in the ΔtonB background on LBA supplemented with iron, but the construction of a triple-knockout mutant (ΔtonB ΔompC ΔompF) required the expression of at least one of the porin genes from a plasmid replicon. Interestingly, unlike the porin-devoid triple-knockout mutant, we were able to construct ΔtonB ΔompR and ΔtonB ΔaroB ΔompR mutants, albeit only on LBA supplemented with iron. In the ΔompR background, ompC and ompF porin expression is extremely low but presumably not zero, which is the case in the ΔtonB ΔompC ΔompF mutant. We think that this extremely low porin expression permits the construction of the ΔtonB ΔompR strain, which can form single, albeit very small colonies, but only on LBA supplemented with iron. Whereas the diffusion of ferrous iron across the outer membrane occurs via OmpC or OmpF channels, at least three proteins-FeoB, MntH, and ZupT-can transport ferrous iron across the inner membrane of E. coli cells (48). Consistent with this, a ΔtonB ΔfeoB double mutant is viable and grows like the ΔtonB mutant (data not shown).
Although the essential role of porins in iron acquisition becomes apparent without the TonB-dependent, high-affinity ferric iron transport systems, their derepression without OmpC or FeoB indicate that the porin-mediated iron transport is active even in the presence of the TonB-dependent high-affinity iron transport systems. The importance of the porin-FeoB pathway for bacterial growth should further increase as E. coli cells enter microaerobic or anaerobic environments where the ferrous species predominates. The involvement of porin and FeoB in iron-dependent growth and/or virulence has been reported for several bacteria, including E. coli (9), S. Typhimurium (10; 11), Helicobacter pylori (12), Vibrio cholerae (49), and Mycobacterium smegmatis (50). Interestingly, M. smegmatis porins increase ferric citrate uptake (50). Similarly, a study reported liganded iron uptake via the OprF porin in Pseudomonas aeruginosa (51). There are no definitive reports in E. coli showing the involvement of porins in liganded iron transport, even for ferric citrate, whose size is below the diffusion limits of the porins (52). Regardless of these ambiguities, published reports and the work carried out acid (DTPA) and desferrioxamine were obtained from Sigma-Aldrich. All other chemicals were of analytical grade. The growth medium was supplemented with ampicillin (50 g/ml), chloramphenicol (12.5 g/ml), kanamycin (25 g/ml), or tetracycline (10 g/ml) when necessary. To induce plasmid-borne gene expression, 0.2% L-arabinose or 0.4 mM IPTG (isopropyl-␤-D-thiogalactopyranoside) was added to the medium.
Genetic and DNA methods. Standard bacterial genetic methods, including P1 transduction and plasmid transformation, were carried out as described by Silhavy et al. (54). To clone the ompR and rstA genes into pBAD24 (55) and pET24d(ϩ) (Novagen), DNA corresponding to their open reading frames (ORFs) were amplified by PCR using primers that carried appropriate restriction enzyme sites for cloning. The reverse primers used for cloning into pBAD24 additionally contained nucleotides encoding six consecutive histidine codons. (Primer sequences are available upon request.) Deletion of the fepA, feoA, feoB, and feoAB genes from their chromosomal locations and subsequent scaring of the antibiotic-resistant marker at the deletion sites were done using the -Red-mediated gene recombination method (56). Deletions were confirmed by PCR and DNA sequence analyses. In some instances, promoterless lacZY genes were recombined at the deletion scar site by the method of Ellermeier et al. (57).
RNA isolation, real-time quantitative PCR, and microarray analyses. Total RNA was extracted from 5 ml cells grown to log phase (OD 600 ϳ0.6) at 37°C using TRIzol Max bacterial RNA isolation kit (Invitrogen). RNA was further purified using the RNeasy kit (Qiagen), and the quality of RNA was assessed by using an Agilent 2100 bioanalyzer (Agilent Technologies). The purified RNA was then converted to either single-stranded cDNA for use in RT-qPCR or double-stranded cDNA for use in DNA microarray analysis.
For RT-qPCR, single-stranded cDNA was synthesized from 10 g of RNA using 100 pM random hexamer primer (Integrated DNA Technologies) and M-MuLV reverse transcriptase (New England Biolabs). After reverse transcription, cDNA was treated with 5 U of RNase H (New England Biolabs) for 20 min at 37°C, followed by purification with a QIAquick PCR purification kit (Qiagen). To quantify the RNA transcripts, 300 nM concentrations of primer specific to the gene of interest and 20 ng of cDNA was added to SYBR green PCR master mix (Applied Biosystems) in a 20-l reaction. Primers were designed according to the manufacturer's protocol included with the SYBR green PCR master mix and RT-qPCR reagents. Critical threshold (C T ) values were determined by using an ABI Prism 7900HT sequence detection system (Applied Biosystems). The relative quantification of target transcripts was calculated according to the 2 -ΔΔCT method (58) using ftsL and purC as the endogenous control genes. Briefly, changes in C T value (ΔC T ) for the gene of interest were calculated by subtracting that gene's average C T from the average C T for the endogenous control gene. The ΔC T for the mutant was then subtracted from the wild-type strain's ΔC T value to give the ΔΔC T value. Each PCR was performed in triplicate and fold changes in transcript levels, along with the standard deviations, were calculated from at least two experiments (n Ն 2).
For microarray analysis, an Invitrogen superscript double-stranded cDNA synthesis kit was used to generate double-stranded cDNA according to the manufacturer's instructions. Single-stranded cDNA was synthesized from 10 g of RNA using a 100 pM concentration of random hexamer primer (Integrated DNA Technologies) and Superscript II reverse transcriptase. Second strand synthesis was performed according to the manufacturer's instructions, and the reaction was stopped with 0.5 M EDTA. RNA was then digested using RNase A (25 g/ml final concentration), followed by treatment with phenol-chloroform and precipitation with ethanol. Double-stranded cDNA was further purified with a QIAquick PCR purification kit (Qiagen) and quality tested by using an Agilent bioanalyzer. Cy3 fluorescently labeled cDNA was used to probe array slides printed with 4,254 E. coli ORFs. Array slides contained 8 probes per gene (in duplicate) corresponding to roughly 72,000 probes per sample. Sample labeling with Cy3 fluorescent dye, hybridization to the 4-plex array (0771112 E. coli K-12 EXP X4, catalog number A6697-00-01), washing, and one-color scanning were performed by Roche Nimblegen in accordance with their standard protocol. Analysis of gene expression profiles was performed using ArrayStar 2.0 software (DNAStar) with a focus on genes with a Ն2-fold change in gene expression. P values were generated with the Student t test, and false positives were minimized using false discovery rate analysis (59).
Enzymatic assays. A ␤-galactosidase assay was performed according to the Miller method (60). Assays were carried out with at least two independent cultures. The ␤-galactosidase activity was expressed as Miller units (60). In some instances, kinetic analysis of enzyme activity was carried out using a VersaMax (Molecular Dynamics) microtiter plate reader in quadruplicate, and the activity was measured as the rate of ONPG cleavage divided by the cell density in each well.
Electron paramagnetic resonance spectroscopy. Free iron concentration in whole cells was determined by EPR spectroscopic analysis (32) with some modifications. Briefly, overnight grown bacterial cultures were diluted 1:100 in 200 ml of LB and grown shaking at 37°C until the optical density at 600 nm (OD 600 ) reached 0.8. Cells were pelleted by centrifugation in a GSA rotor (Sorvall) for 10 min at 6,000 ϫ g. Pellets were resuspended in 10 ml of LB containing 10 mM DTPA (to chelate extracellular iron) and 20 mM desferrioxamine (to chelate intracellular free or accessible ferric iron) and incubated with shaking for 37°C for 15 min. Cells were pelleted as described above and washed twice with 5 ml of ice-cold 20 mM Tris-HCl (pH 7.4). The final cell pellet was resuspended in 0.3 ml of ice-cold 20 mM Tris-HCl (pH 7.4) containing 30% glycerol. A 250-l aliquot of this cell suspension was placed in a quartz EPR tube (length, 250 mm; external diameter, 4 mm [Wilmad-Labglass]). Tubes were frozen in loosely packed dry ice and then transferred to -80°C until the EPR analysis. The remaining cells were diluted 10 3 -fold to determine the OD 600 . Iron standards were prepared from a freshly prepared 10 mM FeCl 3 ·6H 2 O stock in a buffer containing 20 mM Tris-HCl (pH 7.4) and 1 mM desferrioxamine. Theoretical concentrations of iron standards were 100, 50, 25, 10, 5, and 0 M. The actual iron concentrations were determined by measuring the OD 420 of each standard and using the formula following: molar concentration ϭ A 420 /, where [Fe] is 2.865 mM Ϫ1 Cm Ϫ1 . A 250-l aliquot of each standard was placed in separate EPR tubes that were then frozen. The standard curve was generated by plotting EPR signals against actual iron concentrations (Fig. S6). The free iron concentration for each strain was determined from the EPR data and the standard curve. The intracellular free iron concentration was then deduced by integrating the intracellular volume of the cell (1 ml of 1.0 OD 600 cells has an intracellular volume of 0.00052 ml; Jim Imlay, unpublished data) and using the following formula: intracellular free iron concentration ϭ [Fe] from standard curve/cell paste OD 600 ϫ 0.00052 ml.
EPR measurements were carried out at the EPR Facility at Arizona State University. Continuous wave EPR spectra were recorded using an ELEXSYS E580 CW X-band spectrometer (Bruker, Rheinstetten, Germany) equipped with a model 900 EPL liquid helium cryostat (Oxford Instruments, Oxfordshire, UK). For all measurements, the magnetic field modulation frequency was 100 kHz, the amplitude was 1.25 mT, the microwave power was 10 mW, the microwave frequency was 9.44 GHz, the sweep time was 42 s, and the temperature was 20 K.
OmpR purification. OmpR was purified from BL21(DE3) cultures carrying a pET24-ompR 6His plasmid. Overnight cultures, grown without IPTG, were diluted 1:100 in 1 liter of LB, grown with vigorous shaking for 90 min, and then supplemented with IPTG and grown for another 2 h. The cells were pelleted, washed with 10 mM Tris-HCl (pH 7.5), resuspended in lysis buffer (10 mM Tris-HCl [pH 7.5], 1 mM EDTA, 100 g/ml lysozyme), and incubated on ice for 30 min. MgCl 2 (10 mM final), phenylmethylsulfonyl fluoride (2 mM final), and DNase I (40 g/ml final) were then added to the cell suspension. Cells were lysed by passage through a French pressure cell three times, and the lysate was centrifuged at low speed to remove unlysed cells. Envelopes were removed from the lysate by ultracentrifugation at 105,000 ϫ g for an hour at 4°C. Supernatant was filtered through a 0.45-m syringe filter, and the filtrate was subjected to nickel affinity column chromatography using buffers for protein binding (20 mM sodium phosphate [pH 7.4], 20 mM imidazole, and 50 mM NaCl), washing (20 mM sodium phosphate [pH 7.4], 50 mM imidazole, and 300 mM NaCl), and elution (20 mM sodium phosphate [pH 7.4], 300 mM imidazole, and 300 mM NaCl). Samples from eluted fractions were analyzed by SDS-PAGE, and protein bands were visualized after Coomassie blue staining (Fig. S3). Fractions representing OmpR 6His peaks were pooled and dialyzed against a buffer containing 20 mM sodium phosphate (pH 7.4) and 300 mM NaCl. Purified proteins were stored at 4°C in the dialysis buffer supplemented with glycerol (5% final concentration), EDTA (0.1 mM, final concentration), and dithiothreitol (0.1 mM, final concentration).
Electrophoretic mobility gel shift assays. EMSAs were carried out using a LightShift chemiluminescent EMSA kit (Thermo Scientific). ompC and feoABC promoter fragments were generated by PCR using primers specific to the region of interest, with one of the primers biotinylated. Biotin-labeled DNA probes (20 fmol), purified OmpR 6His (100 pmol), and other relevant reagents provided with the kit were incubated for 20 min at room temperature, and the reaction was stopped by adding 5ϫ loading buffer. The mixture was analyzed by 5% acrylamide gel electroblotted onto polyvinylidene difluoride Immobilon-P membrane (Millipore) using a Mini Trans-Blot cell (Bio-Rad). After transfer, DNA was cross-linked to the membrane using Hoefer UV Crosslinker and incubated with stabilized streptavidin-HRP conjugate for an hour. DNA was detected by a molecular imager ChemiDoc XRS system (Bio-Rad) after the membrane was incubated for 5 min with freshly mixed luminol/enhancer and stable peroxide solutions.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
This study was supported in part by grants from the National Institutes of Health (GM048167 and AI117150, both now completed) and the School of Life Sciences, Arizona State University (ASU).
We thank Ananya Sen and Yidan Zhou in Jim Imlay's lab for training R.M. in EPR analysis. The initial EPR work at the University of Illinois Urbana-Champaign facility was supported by GM49640 to Jim Inlay. The final EPR analysis was conducted at the ASU EPR facility. We thank Marco Flores for assistance in EPR analysis at ASU and in constructing Fig. 3.