The AraC-type Regulator RipA Represses Aconitase and Other Iron Proteins from Corynebacterium under Iron Limitation and Is Itself Repressed by DtxR*

The mRNA level of the aconitase gene acn of Corynebacterium glutamicum is reduced under iron limitation. Here we show that an AraC-type regulator, termed RipA for “regulator of iron proteins A,” is involved in this type of regulation. A C. glutamicum ΔripA mutant has a 2-fold higher aconitase activity than the wild type under iron limitation, but not under iron excess. Comparison of the mRNA profiles of the ΔripA mutant and the wild type revealed that the acn mRNA level was increased in the ΔripA mutant under iron limitation, but not under iron excess, indicating a repressor function of RipA. Besides acn, some other genes showed increased mRNA levels in the ΔripA mutant under iron starvation (i.e. those encoding succinate dehydrogenase (sdhCAB), nitrate/nitrite transporter and nitrate reductase (narKGHJI), isopropylmalate dehydratase (leuCD), catechol 1,2-dioxygenase (catA), and phosphotransacetylase (pta)). Most of these proteins contain iron. Purified RipA binds to the upstream regions of all operons mentioned above and in addition to that of the catalase gene (katA). From 13 identified binding sites, the RipA consensus binding motif RRGCGN4RYGAC was deduced. Expression of ripA itself is repressed under iron excess by DtxR, since purified DtxR binds to a well conserved binding site upstream of ripA. Thus, repression of acn and the other target genes indicated above under iron limitation involves a regulatory cascade of two repressors, DtxR and its target RipA. The modulation of the intracellular iron usage by RipA supplements mechanisms for iron acquisition that are directly regulated by DtxR.

activity of aconitase (EC 4.2.1.3), which catalyzes the stereospecific and reversible isomerization of citrate to isocitrate via cis-aconitate, varies depending on the carbon source and that this is caused by transcriptional regulation (2). A repressor of the TetR family, called AcnR, was identified, which represses aconitase by binding to an imperfect inverted repeat within the acn promoter region and interfering with the binding of RNA polymerase (2). The factors that control binding of AcnR to its operator are not yet known. DNA microarray experiments revealed that acn expression is not only influenced by the carbon source but also by the iron concentration of the medium (2). Under iron limitation, the acn mRNA level in the wild type was 3-fold lower than under iron excess. In the ⌬acnR mutant, this decrease was even larger (4.8fold), presumably because the increased expression of aconitase, which contains a 4Fe-4S cluster, leads to an enhanced iron starvation.
We now have identified a new transcriptional regulator, designated RipA, which is responsible for iron-dependent regulation of aconitase and several other iron-containing proteins. Evidence is provided that RipA represses acn and six other target operons under iron limitation and is itself repressed under iron excess by the global iron repressor DtxR.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Media, and Growth Conditions-All strains and plasmids used in this work are listed in supplemental Table S1. The C. glutamicum type strain ATCC13032 (3) was used as wild type. Strain ⌬ripA is a derivative containing an in-frame deletion of the ripA gene. For growth experiments, 5 ml of brain heart infusion medium (Difco) was inoculated with colonies from a fresh LB agar plate (4) and incubated for 6 h at 30°C. After washing, the cells of this first preculture were used to inoculate a 500-ml shake flask containing 50 ml of CGXII minimal medium (5) with 4% (w/v) glucose and either 1 M FeSO 4 (iron starvation) or 100 M FeSO 4 (iron excess). This second preculture was cultivated overnight at 30°C and then used to inoculate the main culture to an A 600 ϳ1. The main culture contained the same iron concentration as the second preculture. The trace element solution with iron salts omitted and the FeSO 4 solution were always added after autoclaving. For growth of C. glutamicum strains carrying plasmid pJC1 or pJC1-ripA, the medium was supplemented with 25 g/ml kanamycin. For all cloning purposes, Escherichia coli DH5 (Invitrogen) was used as host, for overproduction of RipA and DtxR E. coli BL21(DE3) (6). The E. coli strains were cultivated aerobically in LB medium at 37°C (DH5) or at 30°C (BL21(DE3)). When appropriate, kanamycin was added to a concentration of 50 g/ml.
Recombinant DNA Work-The enzymes for recombinant DNA work were obtained from Roche Applied Science or New England Biolabs (Frankfurt, Germany). The oligonucleotides used in this study were obtained from Operon (Cologne, Germany) and are listed in supplemental Table S2. Routine methods like PCR, restriction, or ligation were carried out according to standard protocols (4). Chromosomal DNA from C. glutamicum was prepared as described (7). Plasmids from E. coli were isolated with the QIAprep spin miniprep kit (Qiagen, Hilden, Germany). E. coli was transformed by the RbCl method (8), and C. glutamicum was transformed by electroporation (9). DNA sequencing was performed with a Genetic Analyzer 3100-Avant (Applied Biosystems, Darmstadt, Germany). Sequencing reactions were carried out with the Thermo Sequenase primer cycle sequencing kit (Amersham Biosciences).
An in-frame ripA deletion mutant of C. glutamicum was constructed via a two-step homologous recombination procedure as described previously (10). The ripA up-and downstream regions (ϳ500 bp each) were amplified using the oligonucleotide pairs orf1558-A-for/orf1558-B-rev and orf1558-C-for/orf1558-D-rev, respectively, and the products served as template for cross-over PCR with oligonucleotides orf1558-A-for and orf1558-D-rev. The resulting PCR product of ϳ1 kb was digested with EcoRI and HindIII and cloned into pK19mobsacB (11). DNA sequence analysis confirmed that the cloned PCR product did not contain spurious mutations. Transfer of the resulting plasmid pK19mobsacB-⌬ripA into C. glutamicum and screening for the first and second recombination event were performed as described previously (10). Kanamycin-sensitive and saccharose-resistant clones were tested by PCR analysis of chromosomal DNA with the primer pair orf1558amp-for/orf1558-amp-rev (supplemental Table S2). Of 10 clones tested, five showed the wild-type situation (2.0-kb fragment) and five had the desired in-frame deletion of the ripA gene (1.1-kb fragment), in which all nucleotides except for the first six codons and the last 12 codons were replaced by a 21-bp tag.
In order to complement the ⌬ripA mutant, the ripA coding region and 250-bp upstream DNA containing the promoter region were amplified using oligonucleotides (ripAϩ250-for (2) and ripAϩ250-rev) introducing a SalI and a PstI restriction site, respectively. The resulting 1245-bp PCR product was cloned into the vector pJC1 (12). The resulting plasmid pJC1-ripA and pJC1 were used to transform C. glutamicum wild type and the ⌬ripA strain.
For overproduction and purification of RipA with an N-terminal StrepTag-II (13), the ripA coding region was amplified using oligonucleotides that introduce an NdeI restriction site, including the start codon (ripA-2-for) and an XhoI restriction site after the stop codon (ripA-2-rev). The purified PCR product was cloned in the modified expression vector pET28b-Streptag (14), resulting in plasmid pET28b-Streptag-ripA. The RipA protein encoded by this plasmid contains 14 additional amino acids (MASWSHPQFEKGAH) at the amino terminus. For overproduction and purification of DtxR, the dtxR coding region (equivalent to NCgl1845) was amplified using oligonucleotides that introduced an NdeI restriction site at the translation initiation codon (dtxR-for-1) and four histidine codons plus an XhoI restriction site before the stop codon (dtxR-rev-1). The PCR product was cloned into the pET24b vector, resulting in plasmid pET24b-dtxR-C. The DtxR protein encoded by this plasmid contains 12 additional amino acids at the carboxyl terminus (HHHHLEHHHHHH). The PCR-derived portions of pET28b-Streptag-ripA and pET24b-dtxR-C were analyzed by DNA sequence analysis and found to contain no spurious mutations. For overproduction of RipA and DtxR, the two plasmids were transferred into E. coli BL21(DE3).
Preparation of Total RNA-Cultures of the wild type and the ⌬ripA mutant were grown in CGXII minimal medium containing 4% (w/v) glucose under iron limitation (1 M FeSO 4 ) or iron excess (100 M FeSO 4 ). In the exponential growth phase at an A 600 of 4-6, 25 ml of the cultures were used for the preparation of total RNA as described previ-ously (15). Isolated RNA samples were analyzed for quantity and quality by UV spectrophotometry and denaturing formaldehyde-agarose gel electrophoresis (4), respectively, and stored at Ϫ70°C until use.
DNA Microarray Analyses-The generation of whole-genome DNA microarrays (16), synthesis of fluorescently labeled cDNA from total RNA, microarray hybridization, washing, and data analysis were performed as described previously (2,(17)(18)(19). Genes that exhibited significantly changed mRNA levels (p Ͻ 0.05 in a Student's t test) by at least a factor of 1.7 were determined in two series of DNA microarray experiments: (i) five comparisons of the wild type and the ⌬ripA mutant cultivated in CGXII minimal medium with 4% (w/v) glucose under iron limitation (1 mM FeSO 4 ); (ii) two comparisons of the wild type and the ⌬ripA mutant cultivated in CGXII-glucose medium under iron excess (100 M FeSO 4 ).
Aconitase Assay-Aconitase activity was determined as the rate of cis-aconitate formation from isocitrate (20), as described previously (2), except that the assay was performed at 30°C. Cells of the 20-ml main culture were harvested by centrifugation at 5,000 ϫ g for 10 min and 4°C. The cell pellet was resuspended in 90 mM Tris/HCl, pH 8.0, and used for the preparation of cell extract. The assay mixture contained 950 -995 l of 90 mM Tris/HCl, pH 8.0, and 20 mM DL-trisodium isocitrate. The reaction was started with the addition of 5-50 l of cell extract, and the formation of cis-aconitate was followed by measuring the absorbance increase at 240 nm using a Jasco V560 spectrophotometer. An extinction coefficient for cis-aconitate of 3.6 mM Ϫ1 cm Ϫ1 at 240 nm was used. One unit of activity corresponds to 1 mol of isocitrate converted to cis-aconitate per min.
Overproduction and Purification of RipA-E. coli BL21(DE3) carrying the plasmid pET28b-strep-ripA was grown at 30°C in 200 ml of LB medium with 50 g/ml kanamycin to an A 600 of ϳ1.2 before adding 1 mM isopropyl ␤-D-thiogalactoside. After cultivation for another 4 h, cells were harvested by centrifugation, washed once, and stored at Ϫ20°C. For cell extract preparation, thawed cells were resuspended in 10 ml of buffer W (100 mM Tris/HCl, pH 8.0, 150 mM NaCl). After the addition of 1 mM diisopropylfluorophosphate and 1 mM phenylmethylsulfonyl fluoride, the cell suspension was passed three times through a French pressure cell (SLM Aminco, Spectronic Instruments, Rochester, NY) at 207 megapascals. Intact cells and cell debris were removed by centrifugation (15 min, 5,000 ϫ g, 4°C), and the cell-free extract was subjected to ultracentrifugation (1 h, 150,000 ϫ g, 4°C). The supernatant obtained after ultracentrifugation was applied to a StrepTactin-Sepharose column with a bed volume of 1 ml (IBA, Göttingen, Germany). The column was washed with 6 ml of buffer W, and RipA tagged with StrepTag-II was eluted with 8 ϫ 0.5 ml of buffer W containing 7.5 mM desthiobiotin (Sigma). Fractions containing RipA were pooled, and the buffer was exchanged against TG buffer (30 mM Tris/HCl, pH 7.5, 10% (v/v) glycerin) using Vivaspin concentrators with a cut-off of 10 kDa. Protein concentrations were determined with the BCA protein assay kit (Pierce) using bovine serum albumin as a standard. The purity of the protein preparations was assessed by SDS-PAGE and subsequent protein detection with Gel Code blue stain reagent (Pierce). Using this protocol, ϳ0.2 mg of RipA protein was purified to apparent homogeneity from 200 ml of culture.
Overproduction and Purification of DtxR-E. coli BL21(DE3) carrying the plasmid pET24b-dtxR was grown at 30°C in 100 ml of LB with 50 g/ml kanamycin. Expression was induced at an A 600 of ϳ0.3 with 1 mM isopropyl ␤-D-thiogalactoside. Four hours after induction, cells were harvested by centrifugation and stored at Ϫ20°C. For cell extract preparation, thawed cells were washed once and resuspended in 10 ml of TNGI5 buffer (20 mM Tris/HCl, pH 7.9, 300 mM NaCl, 5% (v/v) glycerol, 5 mM imidazol) containing 1 mM diisopropylfluorophosphate and 1 mM phenylmethylsulfonyl fluoride. Disruption of the cells and fractionation by centrifugation was performed as described above for RipA purification. DtxR present in the supernatant of the ultracentrifugation step was purified by nickel affinity chromatography using nickel-activated nitrilotriacetic acid-agarose (Novagen). After washing the column with TNGI50 buffer (which contains 50 mM imidazol), DtxR protein was eluted with TNGI100 buffer (which contains 100 mM imidazol). Fractions containing DtxR were pooled, and the elution buffer was exchanged against TG buffer (30 mM Tris/HCl, pH 7.5, 10% (v/v) glycerin). From 100 ml of culture, ϳ3 mg of DtxR was purified to apparent homogeneity.
Gel Shift Assays-For band shift assays of RipA with putative target promoters, purified RipA protein was mixed with DNA fragments (200 -630 bp, final concentration 8 -13 nM) in a total volume of 20 l. The binding buffer contained 20 mM Tris/HCl, pH 7.5, 0.5 mM EDTA, 5% (v/v) glycerol, 1 mM dithiothreitol, 0.005% (v/v) Triton X-100, 50 mM NaCl, 5 mM MgCl 2 , and 2.5 mM CaCl 2 . Approximately 20 nM of different nontarget promoter fragments (clpC, clpP, ripA, and porB) were added as a negative control. After incubation for 30 min at room temperature, the samples were separated on a 10% native polyacrylamide gel at room temperature and 170 V using 1ϫ TBE (89 mM Tris base, 89 mM boric acid, 2 mM EDTA) as electrophoresis buffer. The gels were subsequently stained with Sybr Green I according to the instructions of the supplier (Sigma) and photographed.
Binding of DtxR to the ripA promoter was carried out in a 20-l reaction mixture containing 100 mM Tris/HCl (pH 7.5), 5 mM MgCl 2 , 40 mM KCl, 10% (v/v) glycerol, 1 mM dithiothreitol, 150 M MnCl 2 , an 18 nM concentration of a 300-bp ripA promoter DNA fragment, and DtxR in concentrations ranging from 0 to 3.6 M. The ripA fragment covered the region from position Ϫ230 to ϩ70 relative to the translation start and was obtained by PCR with primers ripA-Prom-for and ripA-Promrev. As a negative control, a 23 nM concentration of a 200-bp acn promoter fragment extending from position ϩ190 to Ϫ50 relative to the acn transcription start site (2) was added. This fragment was amplified with primers acn-Prom4-for and acn-Prom4-rev. The reaction mixture was incubated at room temperature for 30 min and then loaded onto a 10% native polyacrylamide gel containing 1 mM dithiothreitol and 150 M MnCl 2 . Electrophoresis was performed at room temperature and 170 V using 1ϫ TB (89 mM Tris base, 89 mM boric acid) supplemented with 1 mM dithiothreitol and 150 M MnCl 2 as electrophoresis buffer. All PCR products used in the gel shift assays were purified with the PCR purification Kit (Qiagen, Hilden, Germany) and eluted in EB buffer (10 mM Tris/HCl, pH 8.5).

RESULTS
Identification of RipA as a Potential Iron-dependent Regulator of the Aconitase Gene-In a previous study, we showed that expression of the aconitase gene acn of C. glutamicum is influenced by the iron availabil-ity, being reduced under iron limitation (2). This regulation also occurred in a mutant lacking AcnR, a repressor of the acn gene, and thus must be mediated by a different regulator or regulatory mechanism. A candidate gene that might be responsible for iron-dependent regulation of acn was identified in the DNA microarray experiments used to compare the gene expression profile of C. glutamicum under iron excess and iron limitation. Expression of the gene NCgl0943 was strongly influenced by the iron availability (2). Its mRNA level was always found to be increased under iron-limiting conditions, and it thus behaved like typical iron starvation genes. The protein derived from NCgl0943 is composed of 331 amino acid residues (36.044 kDa) and contains a DNA binding domain of the AraC family (PF00165 in the PFAM data base (21), PS01124 in the PROSITE data base (22)) with two helix-turn-helix motifs extending from position 113 to 159 and from position 165 to 208. It is flanked by amino-and carboxyl-terminal domains of 112 and 123 residues, respectively, which show no significant sequence similarity to other proteins. Based on the results described below, the NCgl0943 gene was designated ripA (repressor of iron proteins A).
In order to test an involvement of the RipA protein in acn regulation, a ripA deletion mutant of C. glutamicum was constructed. In a first set of experiments, the growth behavior of the ⌬ripA mutant was tested. As shown in Fig. 1A, no differences were observed between wild type and mutant cultivated in glucose minimal medium containing excess iron  (100 M). However, under iron-limiting conditions (1 M), the ripA mutant grew initially like the wild type, but after an A 600 of about 5, the growth rate of the mutant decreased more strongly than that of the wild type. The final cell density of the mutant (A 600 of 20) was only half that of the wild type (A 600 ϭ 40). Thus, the ⌬ripA mutant has a growth defect under iron limitation but not under iron excess. As shown in Fig. 1C, this growth defect could be reversed by transformation with a plasmid carrying the ripA gene with its native promoter region (pJC1-ripA), but not with pJC1 alone (Fig. 1B).
In a second set of experiments, aconitase activity was determined in wild-type and ⌬ripA cells from cultures grown under iron excess and iron limitation. As shown in Fig. 2, the aconitase activity of the two strains was nearly identical under iron excess, whereas under iron limitation, the ⌬ripA mutant had a 1.5-2-fold higher activity than the wild type at four different time points. Thus, the absence of ripA might result in an increased expression of the acn gene under iron limitation, but not under iron excess.
Comparison of the Expression Profiles of ⌬ripA Mutant and Wild Type with DNA Chips-In order to determine the effects of RipA on acn expression as well as on global gene expression, whole genome DNA microarrays of C. glutamicum (16) were used to compare the mRNA ratios of the ⌬ripA mutant and the wild type under iron limitation and iron excess. Under iron starvation (1 M iron), nine genes showed a Ͼ1.7-fold higher mRNA level in the ⌬ripA mutant (TABLE ONE). This group included the aconitase gene acn, supporting the above made assumption that increased acn expression is responsible for the elevated aconitase activity in the ⌬ripA mutant under iron limitation. Besides acn, catA (catechol 1,2-dioxygenase), leuCD (isopropylmalate dehydratase), narKGHJI (nitrate/nitrite transporter and nitrate reductase), sdhCAB (succinate dehydrogenase), and pta (phosphotransacetylase) showed higher mRNA levels in the ⌬ripA mutant compared with the wild type. The mRNA level of the ackA gene for acetate kinase, which is co-transcribed with pta (23), was slightly increased in the ⌬ripA mutant but below the cut-off used. Except for the transporter NarK, phosphotransacetylase, and acetate kinase, the enzymes encoded by these genes are known to contain iron, mostly in the form of iron-sulfur clusters (aconitase, isopropylmalate dehydratase, nitrate reductase, succinate dehydrogenase) and/or heme (nitrate reductase, succinate dehydrogenase) (24). Remarkably, the mRNA level of the genes mentioned above was changed only under iron limitation but not under iron excess (TABLE ONE).
Besides ripA, seven other genes showed a Ͼ1.7-fold decreased mRNA level in the ⌬ripA mutant under iron limitation (TABLE ONE). There is no obvious common property of these genes; however, dps (starvationinduced DNA protection protein) and ftn (ferritin) are critically involved in iron homoeostasis (25,26). In contrast to the genes with an increased mRNA level in the ⌬ripA mutant, dps and ftn had decreased mRNA levels not only under iron limitation but also under iron excess.
Binding of Purified RipA Protein to the acn Promoter-In order to test whether the influence of RipA on acn expression is direct, binding of RipA to the acn promoter was analyzed. For that purpose, the RipA protein containing an amino-terminal StrepTag-II was overproduced in E. coli and purified to apparent homogeneity by affinity chromatography (Fig. 3). Gel shift assays showed that the RipA protein bound with high affinity to fragment 1 covering the entire acn promoter region, whereas a control fragment covering the promoter of the porB gene encoding an anion channel (27) was not shifted (Fig. 4). A RipA-fragment 1 complex was already observed at a 5-fold molar excess of RipA. At a 10-fold excess, two RipA-fragment 1 complexes were observed, and at a 30-fold excess, only the second RipA-DNA complex was observed, suggesting the presence of two binding sites. Gel shift assays with 10 different subfragments (Fig. 4) clearly confirmed the presence of two  FeSO 4 ). The genes leuC, sdhB, narK, and narH show an average mRNA ratio below 1.7 but were included, since they are organized in operons with genes (leuD, sdhCA, or narKGJI) having an mRNA ratio above 1.7. b This column provides the mRNA ratio (⌬ripA mutant/wild type) of the genes under iron excess conditions. It represents the average of two DNA microarray experiments performed with RNA isolated from two independent cultivations in CGXII minimal medium under iron excess (100 M FeSO 4 ).
distinct binding sites, extending from position Ϫ212 to Ϫ194 (binding site A) and from Ϫ155 to Ϫ137 (binding site B) relative to the transcription start site of acn determined previously (2). Fragments lacking these regions (e.g. fragment 2) were not shifted, fragments containing one of the two regions formed a single RipA-DNA complex (e.g. fragment 7), and fragments containing both regions (e.g. fragment 8) formed two RipA-DNA complexes (Fig. 4). Inspection of the two regions revealed that they contained a similar sequence motif but in opposite orientation (Fig. 5). The relevance of this motif was tested by mutational analysis, in which three or four nucleotides were exchanged simultaneously. As shown in Fig. 5, all mutations within the proposed motif prevented RipA binding, whereas the mutations outside did not inhibit binding. These results confirmed the importance of the sequence G(A/T)GCGN 6 GAC for RipA binding.
Binding of Purified RipA Protein to Additional Target Promoters-As a result of the DNA microarray experiments, the operons catA, leuCD, narKGHJI, sdhCAB, and pta-ack were identified as further putative target genes of RipA, since their mRNA level was also increased in the ⌬ripA mutant. We therefore tested the binding of RipA to the corresponding promoter regions. As shown in Fig. 6, all five promoter fragments were shifted by RipA at a molar excess (protein/DNA) of 5-10, and in all cases, two RipA-DNA complexes were formed. This indicates that there are two RipA binding sites in the corresponding promoter regions, as shown above for the acn promoter. Since expression of the katA gene encoding the hemoprotein catalase was also altered in some of the DNA microarray experiments, the katA promoter region was also tested for RipA binding and shown to contain two RipA binding sites with affinities comparable with those described above. In addition, a third binding site of lower affinity was detected (Fig. 6). Binding of RipA was also tested with the promoter regions of ripA, dps, and ftn. In the case of ripA and dps, no shift was observed, suggesting that there is no autoregulation of ripA and no direct control of dps expression by RipA. In the case of ftn, a weak binding was observed, with about 30% of the ftn fragment shifted at a 100-fold molar excess of RipA (data not shown). For the other RipA targets, a complete shift was observed at a 10 -30fold molar RipA excess. Thus, the affinity of RipA to the ftn promoter appears to be much lower. If RipA directly influences ftn expression, it should act as an activator, since the ftn mRNA level was decreased in the ⌬ripA mutant. Considering that induction of an iron storage protein under iron limitation appears counterproductive, the role of RipA in ftn expression is not yet clear.
As described above for acn, the RipA binding sites within the sdhCAB promoter were narrowed down with six different subfragments (data not shown). In this way, binding site A was shown to be located in the   Table S2 (acn-for-3-acn-Prom10-for). At the right, it is indicated whether the fragment was shifted once (ϩ), twice (ϩϩ), or not at all (Ϫ). The boxes labeled A and B indicate the regions that were identified to contain RipA binding sites. B, gel showing binding of purified RipA (5-50-fold molar excess) to fragment 1 (8.5 nM 632-bp fragment). A 485-bp porB promoter fragment (11 nM) served as negative control. The DNA-protein mixture was incubated for 30 min at room temperature before separation by native polyacrylamide gel electrophoresis (10%) and staining with SybrGreen I. C, gel showing binding of RipA (5-and 10-fold molar excess) to fragment 2 (no binding site), fragment 7 (one binding site), and fragment 8 (two binding sites). region between Ϫ180 and Ϫ90 relative to the sdhC start codon and binding site B between Ϫ90 and ϩ12. Inspection of these regions revealed sequence motifs similar to the ones identified in the acn promoter. The relevance of these sites for RipA binding was again confirmed by mutational analysis (supplemental Fig. S1). Based on the four RipA binding sites identified upstream of acn and sdhC, the other RipA target promoters were searched for motifs similar to G(A/ T)GCGN5(T/C)GAC, and the relevance of putative motifs was subsequently tested by changing three adjacent nucleotides within the motif. In this way, two binding sites were identified upstream of narK and pta, three upstream of katA, and one upstream of leuC and catA. Fig. 7 gives an overview of all identified RipA binding sites, their position relative to the respective start codon, and their orientation. From the alignment of the 13 binding sites, the RipA consensus motif RRGCGN 4 RYGAC was derived. From the 13 motifs, two were present in inverted orientation (acn-B and pta-A). The distance between neighboring RipA binding sites varied between 57 and 339 bp.
Regulation of ripA Expression by DtxR-As shown previously (2), expression of ripA followed the same pattern as that of typical iron acquisition genes (i.e. its mRNA level was always increased under ironlimiting conditions). In Corynebacterium diphtheriae, DtxR in complex with iron represses expression of the iron starvation proteins under iron excess but is inactivated under iron limitation (28). C. glutamicum contains a protein with 72% sequence identity to C. diphtheriae DtxR (encoded by NCgl1845), and the C. glutamicum homolog was previously shown to repress the tox promoter from C. diphtheriae in an iron-dependent manner (29). It was therefore tempting to speculate that expression of ripA is repressed under iron excess by DtxR and derepressed under iron starvation. A 19-bp consensus operator of DtxR from C. diphtheriae has been defined as TWAGGTTAGSCTAAC-CTWA (30). Inspection of the C. glutamicum ripA promoter region revealed a sequence motif (i.e. TGAGGTTAGCGTAACCTAC) that differs in only three positions from the consensus binding site and ends 32 bp upstream of the ripA start codon. In order to test whether this motif is a DtxR binding site, the DtxR protein from C. glutamicum was overproduced in E. coli and isolated by means of a carboxyl-terminal histidine tag (Fig. 2). Gel shift analysis revealed that the purified DtxR protein bound to the ripA promoter region (Fig. 8). A partial shift was observed at a 20-fold molar excess (protein/DNA), whereas a 100-fold molar excess was required for a complete shift. Binding of DtxR to the ripA promoter was strictly dependent on the presence of divalent cations (e.g. Mn 2ϩ ) (data not shown). As a negative control, the promoter region of acn was used, which was not shifted. These results clearly support a regulation of ripA expression by DtxR.  (1, 2, and 3) and outside (4 and 5) the proposed RipA binding sites A and B are listed below the wild-type sequence. Fragments containing these mutations were obtained with the primer pairs acn-A.1/acn-for-3 to acn-A.5/acn-for-3 and acn-B.1/acn-Prom4-rev to acn-B.5/acn-Prom4-rev (see supplemental Table S2). C, gel showing binding of RipA to the mutated DNA fragments. Approximately 30 nM fragments A1-A5 and B1-B5 were incubated for 30 min at room temperature either without RipA (lanes labeled with a minus sign) or with 1.2 M of purified RipA protein (lanes labeled with a plus sign). Subsequently, the samples were separated on a 10% nondenaturating polyacrylamide gel, and the gels were stained with SybrGreen I.  DECEMBER 9, 2005 • VOLUME 280 • NUMBER 49

DISCUSSION
Iron is a critical element for bacteria, being essential as a co-factor in a multitude of enzymes, poorly soluble and dangerous, by catalyzing the formation of reactive oxygen species (25). Therefore, most cells have sophisticated regulatory systems to ensure a sufficient supply of iron but to avoid high levels of free Fe 2ϩ , the form responsible for hydroxyl radical production via the Fenton reaction (31). In many Gram-negative and low GC Gram-positive bacteria, the Fur protein is the central regulator of iron regulation (32,33), whereas in many high GC Grampositive genera (e.g. Corynebacterium, Mycobacterium, Rhodococcus, or Streptomyces), DtxR and homologous proteins are the key regulators in iron metabolism (28,34). Under iron excess, DtxR in complex with its co-repressor Fe 2ϩ represses its target genes, in particular uptake systems for iron siderophores, heme, or other iron sources. When iron becomes limiting, Fe 2ϩ dissociates from DtxR, and apo-DtxR dissociates from its target promoters. The DtxR protein was first identified in C. diphtheriae, where it regulates the expression of the diphtheria toxin gene carried by corynebacteriophage ␤. In this work, we have unraveled a completely new aspect of DtxR (i.e. its influence on the expression of several prominent iron-containing proteins via the AraC-type regulator RipA).
The involvement of RipA in iron-dependent regulation was suggested by recent microarray experiments in which the ripA mRNA level was always found to be increased under iron limitation, similar to a multitude of iron acquisition genes (2). In our present study, transcriptome comparisons of a ⌬ripA mutant and the wild type revealed seven operons whose mRNA level was increased in the ⌬ripA mutant under iron limitation, but not under iron excess (i.e. those encoding aconitase (acn), isopropylmalate dehydratase (leuCD), succinate dehydrogenase (sdhCAB), nitrate/nitrite transporter and nitrate reductase (narKGHJI), catechol 1,2-dioxygenase (catA), phosphotransacetylase (pta), and catalase (katA)). The hypothesis that RipA functions as a repressor of these operons was supported by gel shift assays showing that purified RipA binds to the seven corresponding promoter upstream regions. In all cases, at least two RipA-DNA complexes of distinct mobility were identified in the gel shift experiments, suggesting the presence of at least two RipA binding sites. Using subfragments and mutational analysis, the binding sites upstream of acn and sdhCAB were identified and used to search for similar sequences in the other target promoters. Subsequently, mutational analysis led to the identification of three binding sites upstream of katA and of two binding sites upstream of narKGHJI and pta, whereas in the case of catA and leuCD only one of the binding sites could be identified up to now. Alignment of the corresponding sequences revealed a minimal consensus sequence of the type RRGCGN 4 RYGAC. AraC-type regulators (35,36) contain two adjacent helix-turn-helix (HTH) 3 motifs, which in the case of MarA insert in two adjacent segments of the major groove of the mar promoter (37). Thus, one might speculate that one HTH motif of RipA interacts with the conserved RRGCG motif and the adjacent HTH with the RYGAC motif.
Whereas the vast majority of AraC-type regulators investigated to date function as transcriptional activators (35,36), the results presented here indicate that RipA predominantly acts as a transcriptional repressor. Repression is usually accomplished by binding of the regulator between the Ϫ35 and Ϫ10 regions of the promoter and blocking access of RNA polymerase. From the RipA target operons identified in this work, transcriptional start sites have been determined for acn (located 113 bp (TS2) and 110 bp (TS1) upstream of the start codon (2)) and for pta (located 158 bp (TS2) and 46 bp (TS1) upstream of the initiation codon (38)). In the case of acn, the two identified RipA binding sites are centered at Ϫ203.5 and Ϫ146.5 with respect to TS2. Since these sites are far upstream of the RNA polymerase binding site, the question arises of how RipA represses acn expression. One possibility is the presence of one or more additional binding sites that we have not yet identified. A weak third RipA-acn complex that was observed at high RipA concentrations (Fig. 4B) supports this suggestion. A promising RipA binding motif is located immediately downstream of the acn start codon (GAGCTCACTGTGAC). However, fragment 4 in Fig. 4A, which contains this motif, was not shifted by RipA, at least under the conditions used in the experiment. Possibly, this site can only be occupied after previous binding to one of the other sites or under different conditions. Another possibility could be that an additional protein is involved whose binding is influenced by the presence of RipA. In the case of pta, the identified RipA binding sites are centered at Ϫ111.5 and ϩ156.5 with respect to TS2, with the second binding site overlapping the pta 3 The abbreviation used is: HTH, helix-turn-helix.  plus and minus signs. The designations A, B, and C of the binding sites were assigned according to the distance to the translation start site, with the A sites located most distantly. In the derived consensus sequence, single residues are indicated when they occur in at least 10 binding sites. The first two and the last two bases shown are probably not essential for binding, since mutation of these sites did not inhibit RipA binding in the case of acn-A, acn-B, sdhC-A and sdh-B. The bases interfering with RipA binding in the case of the acn and sdhC regions are shown in Figs. 5 and S1, respectively. The relevance of the other binding sites was confirmed by showing that mutation of three consecutive bases inhibited binding of RipA to the fragment containing the proposed binding site. The bases mutated were GAC for narK-A, GGC for narK-B, GCG for pta-A, GTC for pta-B, GCG for katA-A, GAG for katA-B, GCG for katA-C, GCG for leuC-A, and GCG for catA-A. start codon. In this case, direct inhibition of transcription by RipA can be envisaged. For narKGHJI, the RipA binding sites are centered at Ϫ149.5 and Ϫ2.5 with respect to the narK start codon. As in the case of pta, the second site very likely interferes with transcription of the nar operon. In the case of sdhCAB, katA, catA, and leuCD, no predictions can be made on the mechanism of repression yet. The presence of at least two binding sites in each RipA target promoter and the large and varying distances between these binding sites might suggest that DNA looping is involved in the mechanism of action of RipA, as reported, for example, for the AraC-type regulators AraC (39) and MelR (40).
Except for the nitrate/nitrite transporter NarK and presumably phosphotransacetylase, all of the enzymes repressed by RipA contain iron; aconitase and isopropylmalate dehydratase possess one iron-sulfur cluster, succinate dehydrogenase probably harbors three iron-sulfur clusters and two hemes (24), nitrate reductase presumably contains four iron sulfur-clusters and two hemes (24), catalase contains heme, 4 and catechol 1,2-dioxygenase contains one non-heme iron (41). Therefore, it is obvious to assume that a major function of RipA is to reduce the synthesis of iron proteins under iron-limiting conditions, thus reducing the cell's iron demand and preventing the formation of inactive apoenzymes lacking iron. In agreement with such a function, the mRNA levels of acn, leuCD, sdhCAB, narKGHJI, catA, pta, and katA were decreased under iron limitation compared with iron excess, both in the wild type and in the ⌬acnR mutant (2). Whereas repression of the non-iron protein NarK can be explained by co-transcription with the nitrate reductase structural genes, the inclusion of phosphotransacetylase in the RipA regulon must have other reasons. A dependence on iron has only been described for the enzyme of Clostridium acidiurici (42), whereas phosphotransacetylase from other species apparently does not require iron. Phosphotransacetylase catalyzes the reversible conversion of acetylphosphate and acetyl-CoA and, in concert with acetate kinase, is involved in the catabolism of acetate (38) as well as in the formation of acetate from acetyl-CoA. C. glutamicum, in contrast to E. coli, usually does not form acetate as a product of aerobic overflow metabolism, and therefore the primary function of phosphotransacetylase in this species appears to be in acetate utilization. Since acetate catabolism involves a 2-3-fold higher carbon flux through the citric acid cycle compared with growth on glucose (43), repression of pta by RipA may serve to reduce acetate utilization under iron limitation and thus to prevent an increased citric acid cycle flux, which cannot be maintained if aconitase and succinate dehydrogenase are repressed at the same time.
A regulatory cascade with an analogous function to that of DtxR and RipA in Corynebacterium is found in E. coli. Here the role of DtxR is fulfilled by Fur, whereas the small RNA RyhB plays a function similar to that of RipA (44). Expression of ryhB is repressed by Fur under iron excess and increases under iron limitation. RyhB acts as an antisense RNA and inhibits translation of the mRNAs encoding succinate dehydrogenase (sdhCDAB), aconitase A (acnA), fumarase A (fumA), ferritin (ftnA), bacterioferritin (bfr), and superoxide dismutase B (sodB). Although the spectrum of target genes regulated by RipA and RyhB only partially overlaps, it is remarkable that both involve the iron-containing proteins of the citric acid cycle (the only fumarase of C. glutamicum belongs to the type II fumarases and does not contain iron).
A search for the distribution of RipA revealed that homologous proteins are only present in Corynebacterium efficiens (CE1047; 70.1% sequence identity) and C. diphtheriae (DIP0922; 51.5% sequence identity), but not in Corynebacterium jeikeium (45) and other high GC gram positives (e.g. the genera Mycobacterium or Streptomyces). The C. efficiens ripA gene, as annotated in the genome sequence (46), encodes a protein of 400 amino acids. We prefer an ATG start codon that is located 68 codons downstream of the annotated GTG start codon, because the derived protein has a length comparable with the RipA proteins from C. glutamicum and C. diphtheriae (supplemental Fig. S2) and because a well conserved DtxR binding site (TGAGGTTAGCGTA-ACCTAC) deviating in only two positions from the consensus sequence (30) ends 40 bp upstream of the ripA ATG start codon proposed here (164 bp downstream of the annotated GTG start codon). In C. diphtheriae, the annotated genome sequence from strain NCTC13129 predicts that the ripA homologous gene DIP0922 encodes a protein of 335 amino acid residues (47). Inspection of the corresponding upstream sequence revealed a putative DtxR binding site (CGAGCAAGGAGTAAC-CTTA) ending 87 bp upstream of the proposed start codon, which, however, differed in eight positions from the consensus sequence and thus is quite speculative. Interestingly, Lee et al. (48) previously identified a DtxR-regulated gene region designated IRP3 from C. diphtheriae strain C7, which is equivalent to the one described above. The DtxR binding site they identified experimentally by DNase I footprinting  Under iron limitation, DtxR repression is relieved, and RipA protein is synthesized and partially represses expression of its target genes, which encode iron-containing proteins, except for narK, pta, and ackA. In this way, intracellular iron usage is modulated and supplements mechanisms for iron uptake that are directly regulated by DtxR.
DIP0922. Further studies are required to determine the relevance of the strain differences and the functionality of the putative DtxR binding site upstream of DIP0922.
The discovery of RipA as a repressor of iron proteins and its own repression by DtxR have unraveled a new aspect of the regulatory network controlling iron metabolism in Corynebacterium (Fig. 9). Aspects that have to be addressed in future work are the mechanism(s) of repression by RipA and the mechanism of RipA inactivation after a shift from iron limitation to iron excess. This will probably require an understanding of the function of the N-and C-terminal domains that show no homology to other proteins.