Global Iron-dependent Gene Regulation in Escherichia coli A NEW MECHANISM FOR IRON HOMEOSTASIS*

Organisms generally respond to iron deficiency by increasing their capacity to take up iron and by consum-ing intracellular iron stores. Escherichia coli , in which iron metabolism is particularly well understood, contains at least 7 iron-acquisition systems encoded by 35 iron-repressed genes. This Fe-dependent repression is mediated by a transcriptional repressor, Fur (ferric uptake regulation), which also controls genes involved in other processes such as iron storage, the Tricarboxylic Acid Cycle, pathogenicity, and redox-stress resistance. Our macroarray-based global analysis of iron- and Fur-dependent gene expression in E. coli has revealed several novel Fur-repressed genes likely to specify at least three additional iron-transport pathways. Interestingly, a large group of energy metabolism genes was found to be iron and Fur induced. Many of these genes encode iron-rich respiratory complexes. This iron- and Fur-de-pendent regulation appears to represent a novel iron-homeostatic mechanism whereby the synthesis of many iron-containing proteins is repressed under iron-re-stricted conditions. This mechanism accounts for the low iron contents of fur mutants and how E. modulate its iron Analysis of

Iron is an essential minor element for most organisms, playing vital roles in many important biological processes including photosynthesis, N 2 fixation, methanogenesis, H 2 production, and consumption, respiration, the TCA 1 cycle, oxygen trans-port, and DNA biosynthesis. However, despite the indispensability of iron, it is also potentially toxic due to its tendency to catalyze free radical generation. In addition, the extremely poor solubility of the oxidized, ferric form leads to bioavailability problems (1). Organisms counter the difficulties posed by iron nutrition in a number of ways. One common mechanism involves the solubilization of extracellular iron, by reduction or chelation, followed by internalization via specific transporters. Another widespread approach is the deposition of intracellular iron stores within ferritin molecules that can be subsequently utilized to abrogate the effects of iron restriction (1,2).
Iron metabolism in Escherichia coli K-12 is particularly well studied making it a model organism for investigations on ironhomeostatic processes. Like other bacteria, as well as fungi and some plants, it utilizes high-affinity extracellular ferric-chelators, called siderophores, to solubilize iron prior to transport (3). Ferri-siderophore complexes are taken up via specific outer membrane receptors in a process that is driven by the inner membrane potential and mediated by the energy-transducing TonB-ExbB-ExbD system. Periplasmic-binding proteins shuttle ferri-siderophores from the receptors to inner membrane ABC transporters that, in turn, deliver the ferri-siderophores to the cytosol where the complexes are probably dissociated by reduction. E. coli has six known siderophore receptors (Cir, FecA, FepA, FhuA, FhuE, Fiu) providing specificity for several ferri-siderophores (and ferric dicitrate) of which only enterobactin and its derivatives are synthesized endogenously (4). It also possesses three ferri-siderophore periplasmic-binding protein-dependent ABC-transporter systems, FecBCDE, FepBC-DEFG, and FhuBCD, and, like many other bacteria, can take up ferrous iron anaerobically via FeoB. In addition, E. coli contains three iron storage proteins (Bfr, FtnA, and FtnB) of which FtnA plays the major storage role (5).
Not surprisingly, the iron acquisition and storage systems are regulated in response to iron availability. This regulation is mediated by the homodimeric repressor protein, Fur, which employs ferrous iron as co-repressor (4). There is evidence that the Fe 2ϩ -Fur complex also represses genes (cyoA, flbB, fumC, gpmA, metH, nohB, purR, and sodA) involved in various noniron functions (respiration, flagella chemotaxis, the TCA cycle, glycolysis, methionine biosynthesis, phage-DNA packaging, purine metabolism, and redox-stress resistance) so it can thus be considered to be a global regulator (6 -9). Fe 2ϩ -Fur represses transcription by binding to a 19-bp sequence, designated the "iron box," normally located near the Pribnow box of cognate promoters. Fur can also act as a transcriptional activator switching on genes encoding the iron-containing proteins acon-itase A, Bfr, FtnA, fumarases A and B, succinate dehydrogenase, and superoxide dismutase B (7,10,11). This activation appears to be indirect and seems to involve (at least in some cases) the Fe 2ϩ -Fur repressed regulatory RNA, RyhB (11).
Here we use transcriptional profiling to extend the Fur modulon of E. coli. Over 100 Fe 2ϩ -Fur-regulated genes were detected, most of which have not been previously reported. These include unknown genes potentially involved in iron acquisition. A large number of energy metabolism genes, mainly encoding Fe-containing respiratory complexes, were found to be Fe 2ϩ -Fur induced. This represents a major new functional category for inclusion within the Fur modulon. 55 Fe-labeling studies and whole-cell spectroscopy showed that fur mutants are deficient in iron-containing proteins. Together, the data provide an explanation for the low iron contents of fur mutants and reveal a new Fur-dependent mechanism for iron homeostasis.

MATERIALS AND METHODS
Bacterial Strains and Culture Conditions-E. coli strains were grown in Luria-Bertani (L broth) medium (12) at 37°C and shaken at 250 rpm in an orbital shaker. Iron limitation was induced by inclusion of the ferrous iron chelator 2,2Ј-dipyridyl (dip) at 200 M.
RNA Isolation, Preparation of Radiolabeled cDNA, and Real-Time RT-PCR-Cultures of E. coli wild-type (MC4100), wild-type with dip and an E. coli fur mutant (H1941: MC4100 fur) were grown to an OD 650 nm of 1.0 (six replicates for each condition). A 1-ml sample from each culture was harvested by centrifugation and total RNA extracted using the Qiagen RNeasy kit. RNA was treated with RNase-free DNase I (Promega). Each set of six replicate samples was pooled into two groups of three to control for slight growth and extraction variations. Pooled total RNA samples were then used as templates for production of 33 P-labeled cDNA using random hexaprimers (Promega), and the labeled cDNA probes purified using G25-Sephadex columns, as described by Sigma-Genosys. Quantitative RT-PCR was performed using an ABI 5700, the Sybr Green RT-PCR kit (Qiagen) and primers designed to amplify 50 -80-bp fragments. Specificity was confirmed by electrophoretic analysis of the reaction products and by inclusion of template or reverse-transcriptase free controls. The template RNA samples were prepared separately from those used for macroarray analysis.
Macroarray Hybridization and Scanning-Labeled cDNA was hybridized independently to two Panorama TM E. coli gene arrays (Sigma-Genosys) essentially as described by the manufacturer. Hybridized and washed arrays were then exposed to low intensity phosphorimaging cassettes (Molecular Dynamics) for 45 h and the cassettes were then scanned with a Phosphoimager SI system (Molecular Dynamics) at 200-m resolution. Filters were stripped prior to re-use as described by the manufacturer.
Data Analysis-The phosphoimager image files were analyzed using ImageQuant software (Molecular Dynamics). Pixel intensities for each spot were manipulated using Microsoft Excel. The 632 blank spots were used to calculate the background, which was then subtracted from the 8580 gene-specific spots (each gene present in duplicate). Spots with intensities less than two standard deviations above mean background values were considered to display no significant expression. The intensity of each of the gene-specific spots within an individual array was normalized by expressing values as percentages of total gene-specific spot intensity. This allowed comparisons between array experiments. Array experiments were performed in duplicate, and average values calculated along with confidence levels (Student's t test values).
Whole Cell Spectroscopy-E. coli wild type (MC4100) and fur mutant (MC4100 fur) were grown aerobically in L broth, with or without 200 M dip, to mid-log (OD 650 , 0.5), late-log (OD 650 , 1.0), and stationary phase (OD 650 , 3.0). Cells were harvested by centrifugation and 0.5 g of wet cells were resuspended in 10 ml of 0.1 M phosphate buffer (pH 7). Samples of 4 ml were oxidized with 0.005% (w/v) potassium ferricyanide and 0.005% (w/v) ammonium persulfate or were reduced with 0.005% (w/v) sodium dithionite. Spectra were then recorded at room temperature using a dual beam spectrophotometer over a wavelength range of 400 -700 nm, with a reference wavelength of 500 nm (13). Cytochrome d-and b-type cytochrome levels were quantified as previously described (14,15). Prior to whole cell EPR spectroscopy, cells were washed in 0.1 mM phosphate buffer (pH 7) containing 50 mM EDTA in order to remove Mn from the cell surface. Half of each sample was reduced by the addition of 0.005% (w/v) sodium dithionite followed by repetitive freeze-thawing. The EPR spectra were measured at 10 K on a Bruker EMX EPR spectrometer equipped with an Oxford Instruments liquid helium system. A spherical high quality Bruker resonator SP9703 was used. The EPR samples were frozen in Wilmad SQ EPR tubes. Measurements were as follows: microwave power, 3.18 milliwatts; microwave frequency, 9.46 GHz; modulation frequency, 100 kHz; modulation amplitude, 3 G; sweep rate, 3.58 G/s; time constant, 0.082 s. Protein concentrations were determined as previously described (16).
Native Gel Electrophoresis of 55 Fe-labeled Soluble E. coli Proteins-E. coli cultures were grown in 5 ml of L broth containing 55 FeC1 3 (ϳ0.5 MBq/ml). Cells (5 OD 650 nm units) were harvested by centrifugation in the postexponential phase and washed in saline at 4°C. Soluble cell extracts were then prepared by the spheroplast osmotic lysis method (17) with the following modifications. Tricine was used in place of Tris, dithiothreitol was omitted, 0.1% Triton X-100 was included at the lysis stage and glycerol was added postlysis to a final concentration of 5%. Soluble cell extracts were electrophoresed in native-acrylamide gels containing 0.1% Triton X-100 and Tricine in place of Tris. Gels were then dried under vacuum and autoradiographed with Kodak Biomax MS film.

RESULTS AND DISCUSSION
Identification of Fe 2ϩ -Fur-regulated Genes by Transcriptional Profiling-Genomic transcriptional profiling was used to identify genes regulated by both Fe and Fur (see "Materials and Methods"). The transcription profile of MC4100 (wild-type) grown in rich broth was compared with those of both the wild-type grown with an Fe 2ϩ chelator (dip) and the fur mutant (MC4100 fur) grown without chelator. Samples were harvested at an OD 650 nm of 1.0, corresponding to early postexponential phase. No major growth differences were observed for the three experimental conditions. 33 P-labeled cDNA was prepared from the RNA samples using reverse transcriptase and random hexaoligonucleotides, and was hybridized to E. coli Panorama Macroarrays. Each array experiment was performed in duplicate using pooled RNA samples prepared from three identical cultures. Duplicate experiments gave good reproducibility with correlation coefficients of 0.97. Comparison of the L broth versus L-broth plus dip, or wild-type versus fur mutant, gave lower correlation coefficients (0.95 and 0.90). Since members of the Fur modulon are regulated by iron and Fur in conjunction, only those genes that were Ն2-fold regulated by both dip and the fur mutation are considered further here. Accordingly, 101 genes were found to be regulated by the Fe 2ϩ -Fur complex of which 53 were repressed and 48 induced. These genes fall into three major categories: iron metabolism (Table I), energy production (Table II), and miscellaneous/unknown (Table IV).
Iron Metabolism: Potential Novel Iron Transporters-Reassuringly, most of the known iron-acquisition genes were induced by both the chelator and the fur mutation (Table I) validating the experimental procedure. The enterobactin biosynthesis (entA-F) genes were among the most highly de-repressed (average ϳ21-fold) genes (Table I), presumably to ensure that energy is not needlessly squandered on enterobactin production during conditions of iron sufficiency. In contrast, the ferric-enterobactin uptake genes (fepABCDEG) were weakly de-repressed (average ϳ2-fold) indicating that enterobactin production systems are more strongly controlled by iron than the ferri-enterobactin acquisition apparatus. As previously observed, the fecIRABCDE, feoAB, fhuE, fhuF, and cirA genes involved in ferric-dicitrate, ferrous iron, ferri-coprogen/ rhodotorulic acid, ferrioxamine B, and ferric-dihydroxybenzoate utilization were all repressed by the Fe 2ϩ -Fur complex, as were the tonB and exbBD genes required for energy-dependent ferrisiderophore transfer across the outer membrane. However, the fhuACDB operon, specifying the ferric hydroxamate uptake apparatus, was not significantly affected by the chelator or fur mutation although it is known that fhuA-lacZ transcriptional fusions are Fe 2ϩ -Fur regulated (19). The reason for this discrepancy is unclear, but could be related to growth-phase ef-  1 Potential novel iron transporters are boxed, gene orientation is indicated by left (clockwise) and right (counter clockwise) justification, and transcriptional organization is indicated by the vertical arrows. 2 Relative expression levels for the wild type in L broth as parts per million (ppm). 3 Fold increases (ϩ) and decreases (Ϫ) in expression levels (with respect to wild type in L broth) caused by dip or the fur mutation are indicated by black and white backgrounds, respectively. 4 Expression changes of less than 2-fold are shown with gray backgrounds and were included either because the corresponding genes are associated with adjacent Fe 2ϩ -Fur-regulated genes or because the expression changes are considered significant (Ͻ0.05 probability in Student's t-test). 5 The data for the genes of the fhuACDB operon are combined and are included for completion sake although they do not show any major change in expression. 6 The most strongly predicted Fur-binding site is indicated with the probability score (and number of base-pairs matching the 19-bp consensus indicated in parentheses) determined as described by Robison et al (18) using 19 known Fur sites with an average probability score of 19.6 (high scoring sites are more strongly predicted than low scoring sites). NI indicates that no Fur-binding site was predicted using the above approach. 7 The distance of the Fur site upstream (negative) or downstream (positive) of the translational start site of the corresponding gene is indicated. 8 Descriptions are mostly derived from ExPASy.
fects. The suf operon, which probably functions in iron-sulfur cluster assembly during iron starvation and redox stress (20), was also Fe 2ϩ -Fur repressed (average of 5-fold), as previously indicated (21). However, the iscSUA-hscBA-fdx cluster, which encodes genes with a housekeeping role in iron-sulfur cluster assembly, was not Fe 2ϩ -Fur controlled (not shown). Appropriately, the bfd gene encoding a ferredoxin thought to be involved in iron release from bacterioferritin, was de-repressed (22) whereas the ftnA gene specifying the iron-storage protein, ferritin A, was repressed by the chelator and fur mutation, as previously observed (5,11). A significant absentee from the list of genes in Table I is bfr, coding for bacterioferritin. Although this gene is known to be Fe 2ϩ -Fur induced, its expression is RpoS dependent and so is restricted to the stationary phase. 2 Thus, the array data are generally consistent with numerous previous expression studies on the Fur-regulated components of iron metabolism. The array analysis enabled the identification of several unknown genes with potential functions in iron acquisition (Table  I). These were initially recognized either because of their chelator-and fur-dependent expression or by their chromosomal co-location with such genes. They are organized into 6 clusters (boxed in Table I), of which the largest (ybiM-ybiLXI-ybiJ) consists of 5 co-polar genes encoding: YbiM and YbiJ, two TABLE II Iron-and Fur-regulated genes involved in energy metabolism The number of iron atoms per encoded protein is indicated. All other details are as for Table I. 1 The data for the genes of the cyo, nuo and tdc operons are combined to reduce the Table size. 2 Abbreviations: s-u, subunit; and red-ase reductase.
putative-periplasmic/exported proteins that are related to each other, but not to any other E. coli protein; YbiL (or Fiu), a probable TonB-dependent outer membrane receptor previously shown to be involved in Fe 3ϩ -dihydroxybenzoylserine and -dihydroxybenzoate utilization (23); YbiX, a homologue of a protein encoded by an Fe-repressed Pseudomomas aeruginosa gene (piuC) thought to be involved in iron uptake (24); and YbiI, a probable C4-Zn finger protein of unknown function. The above suggests that the Fiu-mediated Fe-uptake system is more complex than hitherto believed.
The second cluster (ycdN-ycdOB) consists of three copolar genes encoding: YcdN, a homologue (24% amino acid sequence identity) of the high-affinity ferrous iron transporter (Ftr1p) of yeast; YcdO, a potential exported lipoprotein of unknown function; and YcdB, another potentially exported protein of unknown function. A twin arginine transporter (tat) motif was previously identified in the N terminus of YcbB (25), and a similar motif is present in the N terminus of YcdO. This suggests, as for many other tat-exported proteins, that these proteins could possess prosthetic groups inserted prior to export to the periplasm. Homologues of the ycdNOB genes are found co-located in the chromosomes of at least 7 other bacteria indicating that these 3 genes form a functional unit. YcdB bears homology to STY2683 of Salmonella typhimurium, a putative iron-dependent peroxidase. By analogy with the yeast Ftr1p system (26), we speculate that YcdN acts as a ferrous iron transporter, and that YcdO and YcdB act together as a novel periplasmic iron oxidase or reductase (this possibility is currently being tested). The third cluster consists of three genes, yddAB-pqqL, that appear to form an operon encoding: YddA, which is homologous to ABC transporters; YddB, homologous to TonB-dependent outer membrane receptors; and PqqL, a potential Zn peptidase (pqqL is also induced by iron restriction in Pasteurella multocia; 45). These genes are likely to specify a new 3-component iron-uptake system. The fourth cluster is also likely to represent a newly identified iron-uptake pathway. It consists of two divergently arranged genes encoding: YcnD, a probable TonB-dependent outer membrane receptor; and YncE, which is predicted to be a pyrolo-quinoline quinone containing periplasmic oxidase. Note that all four of the above Fe 2ϩ -Fur repressed gene clusters are associated with predicted Fur boxes.
The fifth locus consists of a single gene (ydiE) that is related gene regulation Fold changes in expression with respect to the wild type in L broth were determined using macroarrays and quantitative RT-PCR. Other details are as for Table I. 3 Included as positive controls. 4 Included as a negative control. (36% identity at the amino acid sequence level) to hemP, a gene that forms part of the heme utilization operon (hemPRST) of Yersinia enterocolitica (27). The specific function of hemP is uncertain. The ydiE gene is associated with a well predicted Fur box and its homologue (hemP) in S. typhimurium is Fur repressed (28), supporting the Fe 2ϩ -Fur dependent regulation observed here. The final locus also comprises a single gene, yqjH, homologous (28% identity at the amino acid sequence level) with the siderophore-utilization gene (viuB) of Vibrio cholerae. yqjH has an appropriately positioned potential Fur box (29) consistent with its Fe 2ϩ -Fur dependence.
Energy Metabolism: Control of Iron-rich Proteins-Unex-pectedly, a large number of genes encoding proteins involved in energy metabolism were found to be Fe 2ϩ -Fur-regulated (Table  II). Of those genes and operons listed in Table II, only cyoA and gpmA have previously been reported to be Fe 2ϩ -Fur-controlled (8,6). Most (36) were induced by the Fe 2ϩ -Fur complex of which 32 encode iron-containing respiratory complexes associated with a total of 148 iron atoms (per subunit). We speculate that this Fur-dependent control of iron-containing respiratory proteins represents a newly recognized iron homeostatic mechanism whereby the production of a subset of iron proteins is regulated according to iron availability. Such a mechanism would allow the cellular demand for iron to be reduced under iron-restricted growth conditions, enabling available iron to be utilized more economically and helping to ensure that production of Fe-requiring proteins does not exceed iron availability.
Partly consistent with the data in Table II, previous work has shown that the anaerobic expression of cydA, cyoA, narG, and frdA is 2-14-fold reduced by dip (note that for cyoA and cydA this effect required an fnr background) (30). However, in contrast to the findings presented here, this effect appeared to be Fur-independent (the iron-dependent regulator was not identified; Ref. 30). Nearly all of the Fe 2ϩ -Fur induced genes in Table II require the anaerobic regulator, Fnr, and anaerobiosis for full induction (31). Although the growth conditions used here were aerobic, O 2 tensions are very low during late-log growth in rich broth (32) and thus could favor Fnr-dependent expression. It is also possible that anaerobic conditions were introduced at the sample harvesting stage. A potential complication is the possible inactivation of Fnr by the iron chelator and consequent down-regulation of the Fnr-dependent genes in Table II (30). However, such an effect would not be anticipated for the fur mutant and so would not explain both the dip and the fur expression effects shown in Table II. The Fe 2ϩ -Fur induced genes listed in Table II are not essential and the observed reductions in their expression during aerobic growth would not be expected to lead to a major growth defect. Indeed, although fnr mutants are unable to express anaerobic respiratory complexes, they retain the ability to grow both anaerobically (via fermentation) and aerobically.
The induction of expression by the Fe 2ϩ -Fur complex is likely to be indirect since for most of the Fe 2ϩ -Fur induced genes (Tables I, II, and IV), there appear to be no associated Fur boxes. It is probable that the recently identified Fur-dependent regulatory RNA molecule, RyhB, acts as the direct regulator of these genes (11). It is noteworthy that not all Fe-protein-encoding genes appear to be induced by Fe 2ϩ -Fur. This may reflect the high importance of some Fe-proteins, such as the aerobic ribonucleotide reductase (NrdAB) required for DNA biosysnthesis. It is also interesting to note that the expression of gpmA (encoding phosphoglycerate mutase in the glycolytic pathway) and gltA (encoding the TCA cycle enzyme, citrate synthase) is repressed by the Fe 2ϩ -Fur complex (Table II), whereas acnA (encoding the TCA cycle enzyme aconitase A that converts citrate into isocitrate) is known to be induced by Fe 2ϩ -Fur, suggesting that E. coli may respond to iron restriction by producing citrate to mediate iron uptake. Expression of the citrate synthase gene (prpC) is also Fur repressed in Shewanella oneidensis (44).
Quantitative RT-PCR was used to confirm the Fe 2ϩ -Fur dependent expression of 12 iron-regulated genes (Table III). Although the directions of regulation were identical for the RT-PCR and array data, the degree of regulation varied considerably and was generally greater for the RT-PCR analysis, presumably reflecting the higher quantitative precision obtained with RT-PCR. Control experiments with polA (encoding FIG. 3. Effect of the fur mutation on the Fe-S cluster composition of E. coli. Fe-S cluster composition was determined by EPR spectroscopy in the as prepared (oxidized) and dithionite-reduced whole cells harvested in the mid-exponential phase (OD 650 nm 0.5) following aerobic growth in L broth. Spectra for MC4100 (wt) and H1941 (fur) are shown. A, the position of the components of signals from the 3Fe-4S clusters (g ϭ 2.019, paramagnetic in the oxidized state) and signals from the centers spectrally identical to those in complex I (N1b, g 1 ϭ 2.022, g 2 ϭ 1.935, g 3 ϭ 1.924; N2, g 1 ϭ 2.052, g 2 ϭ g 3 ϭ 1.936; N3, g 1 ϭ 2.03, g 2 ϭ 1.93, g 3 ϭ 1.87; N5, g 1 ϭ 2.078, g 2 ϭ 1.93, g 3 ϭ 1.90; all paramagnetic in the reduced state (35,38)) are indicated. The prominent feature at g ϭ ϳ1.935 (arrowed) results from a superimposition of signals from 4Fe-4S clusters. Note: the wild type-reduced spectrum was recorded at half-amplification with respect to the other spectra. B, the same spectrum recorded at a lower field range; the signal from the mononuclear non-heme iron centers (g ϭ 4.3) is indicated. Broadly similar results were obtained for cells grown to OD 650 nm 1.0 and 3.0 (not shown). Spectra with dip are not shown due to interference from a large Mn 2ϩ signal, presumably arising from an Mn 2ϩ -dip complex.
DNA polymerase I) showed no significant Fe 2ϩ -Fur effect and thus matched the DNA array data.
Cellular Iron-Protein Composition Is Fur-dependent-The observed Fe 2ϩ -Fur dependent expression of Fe-proteins is consistent with the 70% reduction in iron levels caused by the fur mutation (5). Part (50%) of this reduction in iron content is due to low levels of FtnA iron stores, and it was suggested that the residual (20%) reduction in iron content is due to decreased expression of Fe-proteins (5). This suggestion is supported by the array data presented in Table II. In order to determine whether the fur mutation does indeed lead to significantly decreased levels of cellular Fe-proteins, a comparison of the soluble Fe-protein composition of wild-type and fur mutants grown in L broth with 55 Fe was performed (Fig.  1). This analysis confirmed that the fur mutant does indeed have a much lower level (ϳ2-fold, as estimated by densitometric analysis of the autoradiograph presented in Fig. 1) of Fe-proteins than the wild-type supporting the notion that Fur controls cellular levels of Fe-proteins as suggested here and elsewhere (5,11,30). Whole Cell Spectroscopy Reveals That Cytochrome b and d and Fe-S Protein Levels Are Fur and/or Iron-regulated-To provide further evidence for the Fe 2ϩ -Fur control of Fe-protein levels, whole cell spectroscopy was used to measure the effect of dip and/or the fur mutation on the relative abundances of cytochromes and Fe-S proteins (Figs. 2 and 3). Visible difference spectroscopy measurements revealed that the cytochrome d-dependent signal at 630 nm is decreased by 3-fold (from 0.06 to 0.02 nmol/mg of protein) by dip or the fur mutation (Fig. 2), which correlates well with the 2.7-fold decrease in expression observed for the corresponding genes (cydAB; Table II). The signal at 560 nm (Fig. 2) is due to overlapping ␣-bands from b-type cytochromes (33), primarily cytochrome b 562 of the cytochrome bo 3 quinol oxidase, the low-spin cytochrome b 558 of the cytochrome bd quinol oxidase, and heme b 556 of succinate dehydrogenase (34). The 560-nm band was 2.9-fold reduced (from 0.2 to 0.07 nmol/mg protein) by dip or the fur mutation, which is consistent with the expression data for both the cydAB and cyo genes (Table II).
Low temperature electron paramagnetic resonance (EPR) spectroscopy of whole cells was used to compare the levels of fully assembled Fe-S-containing proteins. When the cells are frozen as prepared, the Fe-S proteins are found mainly in the oxidized state (Fig. 3A). In this state the major EPR-detectable species is the 3Fe-4S cluster that can be found in, for example, succinate dehydrogenase and fumarate reductase. The content of the 3Fe-4S clusters in the fur mutant is ϳ4-fold lower than in the wild type sample (likely to be derived from fumarate reductase, rather than succinate dehydrogenase; Table II). This correlates well with the 4.2-fold decrease in expression of the fumarate reductase operon (frdABCD) in the fur mutant. The macroarray data showed no significant effect of dip or the fur mutation on sdhCDAB expression and this lack of Fe 2ϩ -Fur response is supported by RT-PCR on sdhB and sdhC (not shown). However, preliminary array studies (not shown) using minimal (as opposed to rich) medium showed a clear iron-dependent expression of the sdh operon, as reported previously (11,36). A signal at g ϭ 4.3, arising from iron in a "rhombic" conformation, is also present in the oxidized samples (Fig. 3B). The most likely species responsible for this signal is a mononuclear non-heme iron center (35). This feature was 7-fold higher in the fur mutant. The identity of this species is unknown but it could correspond to a labile free-iron species, as previously observed in fur mutants (37).
Other Fe 2ϩ -Fur-regulated Genes-Forty-six genes, from various other functional categories, were also Fe 2ϩ -Fur controlled (Table IV). Of these, only nrdH, sodB, and yhhY have previously been reported to be Fe 2ϩ -Fur regulated (8,39). Of particular note is the apparent Fe 2ϩ -Fur induction (average 2.1fold) of the Ni 2ϩ -transport genes (nikA-R; 40). Since all known nickel-containing proteins in E. coli are hydrogenases, which also require iron, it is not surprising that Ni 2ϩ transport is induced by the Fe 2ϩ -Fur complex. Of the four hydrogenase operons, the array data show only that of hydrogenase 2 (hybOA-G) is iron and Fur-regulated (1.4-and 1.7-fold-induced, respectively; data not shown). However, this effect is relatively weak and how it is mediated is unclear. Interestingly, a heatshock gene cluster (hslTS-yidE) was Fe 2ϩ -Fur repressed (average 3.7-fold), while a group of cold-shock genes (cspI, cspB, and cspF) within a phage gene cluster was induced (average 2.7fold) by the Fe 2ϩ -Fur complex. This suggests that temperature shock-dependent gene induction may be partly due to transient changes in intracellular iron availability or Fe 2ϩ -Fur stability. The nrdHIEF operon encoding a non-essential iron-containing ribonucleotide reductase (41) appears to be Fe 2ϩ -Fur repressed (average 2.6-fold) which is supported by previous work showing Fe 2ϩ -Fur repression of nrdH (8). This indicates that this isoenzyme (there are three alternative ribonucleotide reductases in E. coli) may have a role under iron-restricted conditions. The Fe 2ϩ -Fur induced ompW gene encoding an outer membrane of uncertain function is also induced by iron in Pasteurella multicoda (45), although it is repressed by iron in S. oneidensis (44).
In addition to bfr, fhuABCD, and sdhCDAB, there are several other known Fe 2ϩ -Fur regulated genes (acnA, flbB, fumA, fumB, fumC, purR, nohB, sodA, ygaC, and yhhX) that were not affected by dip and Fur in this study. The reasons for this are probably related to the growth conditions employed here, which are unlikely to favor expression of all Fur-controlled genes. For instance, acnA expression is s and SoxSR dependent and so requires stationary phase and redox-stress for maximum induction (42), and our preliminary data (not shown) suggest that the PurR-controlled purine regulon is iron-dependent in minimal (rather than rich) medium. Note that the Fe 2ϩ -Fur regulated promoter originally identified upstream of yhhX (8) appears to be specific for ryhB (11), which would explain its lack of Fe 2ϩ -Fur control in this work.
General Relevance-This study has revealed that, despite much previous work, there is still a great deal to be discovered concerning iron metabolism in E. coli K-12. A major new role for Fur in iron homeostasis has been demonstrated in which Fur mediates the control of cellular iron-protein levels in response to iron availability. This new mechanism is relatively simplistic in that it only involves iron-responsive gene regulation. It is currently unclear how widespread this mechanism is, but given its simplicity, it is possible that it could be common, especially among organisms with the capacity to globally regulate gene expression in response to iron. Recent global transcriptional profiling studies on the effects of iron restriction and/or fur inactivation provide some evidence for iron regulation of Fe-protein genes in other bacteria. In Bacillus subtilis (43) several cytochrome systems (e.g. cydABCD) and aconitase (citB) were reported to repressed by iron limitation. In S. oneidensis, a fur mutation resulted in reduced expression of genes encoding proteins (cytochrome c oxidase, cytochrome c maturation protein B, cytochrome b 561 ) involved in electron transport (44), and in P. multocida, genes encoding proteins (e.g. fumarate reductase, dimethylsulfoxide reductase, putative iron-sulfur protein, NapF, formate dehydrogenase) involved in energy metabolism and electron transport were reduced during iron restriction by 2-6-fold (45). In P. aeruginosa, 87 genes were repressed by iron restriction, although their identities have not yet been reported (46).