The Mycobacterium tuberculosis IdeR is a dual functional regulator that controls transcription of genes involved in iron acquisition, iron storage and survival in macrophages

Identification of IdeR boxes

IdeR was predicted to bind a similar DNA sequence to the C. diphtheriae DtxR, given the following observations: the DNA-binding domain of IdeR is 92% identical and 100% similar on the amino acid level with that of DtxR (Doukhan et al., 1995); these proteins have almost identical crystal structures (Pohl et al., 1999); IdeR can complement a dtxR mutant for iron-dependent regulation of DtxR-regulated promoters (Schmitt et al., 1995); and finally, IdeR binds to the C. diphtheriae tox promoter and other DtxR target gene promoters (Schmitt et al., 1995). Therefore, we hypothesized that many of the IdeR boxes in the M. tuberculosis genome could be identified by conducting a computer search using a DtxR binding site sequence as a surrogate. A preliminary search for IdeR boxes using the blast program against every published sequence (not specific for M. tuberculosis) resulted in the identification of the M. tuberculosis hisE–irg2 promoter region (Rodriguez et al., 1999). We have now conducted a broader search of the M. tuberculosis published genome (Cole et al., 1998) for IdeR boxes to identify genes regulated by IdeR using the available computer program tuberculist (http://genolist.pasteur.fr/TubercuList/Moszer et al., 1995) (Table 1). The criteria chosen as qualities for a DNA region regulated by IdeR were limited to those within +250 to +100 bp of the predicted start codon of an open reading frame (ORF) and containing five or fewer mismatches from our query sequence of TWAGGTWAGSCTWACCTWA (where W = A or T, and S = G or C). There are two major reasons why the putative IdeR boxes in Table 1 are limited to five mismatches. First, there was a dramatic increase in potential IdeR binding sites with each subsequent mismatch allowed (estimated as more than 100 targets for six mismatches and more than 700 targets for seven mismatches). Secondly, if IdeR boxes were randomly distributed in the M. tuberculosis chromosome, there should be approximately four times more putative IdeR boxes inside an average length ORF (≈ 1000 bp) compared with a 250 bp promoter region. This was not the case, as there was only a small number of putative IdeR binding sites with ≤ four mismatches (two in ORFs versus 17 located 250 bp upstream of the putative translational start site), whereas five mismatches resulted in 16 in ORFs versus 19 located 250 bp upstream of the putative translational start site. Although this indicated that the computer search for IdeR boxes was probably more accurate with ≤ four mismatches, putative binding sites with five mismatches were left in Table 1, as some of these were located upstream of genes whose transcription was predicted to be regulated by IdeR (annotated to be involved in iron storage and acquisition).

Over 40 genes preceded by putative IdeR boxes were identified and fell into several categories (Table 1). One group encodes proteins that are annotated to be directly involved in either iron acquisition (mbtA–mbtB, mbtI, rv1348, rv1347c) or iron storage (bfrA, bfrB). The second group of genes, encoding annotated proteins involved in amino acid biosynthesis (hisB, hisB-like, pheA, lysyl tRNA) may be involved in siderophore biosynthesis. The last group contains many genes that either have no obvious relationship to iron (murB, acp) or have no significant homologies to any known proteins in databases.

To test the functionality of the IdeR boxes identified by the computer search, we analysed four that were between or in front of the following genes: mbtA–B, mbtI, rv3402c and bfd–bfrA. mbtA, mbtB and mbtI encode enzymes involved in the biosynthesis of mycobactin, the major M. tuberculosis siderophore (Quadri et al., 1998; De Voss et al., 2000). rv3402c encodes a protein with > 40% amino acid similarity to three types of enzymes: an aminotransferase, a dehydratase and an enzyme involved in perosamine/O-antigen biosynthesis. bfrA encodes a bacterioferritin subunit that is involved in iron storage. bfd is not currently annotated in the M. tuberculosis genome, but we found that this ORF has homology (30–40% amino acid identity, > 50% conserved amino acids) to Bfd from Escherichia coli, a protein that associates with BfrA and may have a role in donating iron to the bacterioferritin complex (Garg et al., 1996; Quail et al., 1996). In M. tuberculosis, bfd and bfrA are adjacent and divergently transcribed (Fig. 5C) and, in E. coli, bfd and bfrA are also neighbouring genes but are transcribed in the same direction (Andrews et al., 1993).

IdeR binds to DNA fragments containing putative IdeR boxes

To demonstrate that IdeR could bind to the putative IdeR boxes identified from our computer search, DNA fragments containing these sequences were PCR amplified and labelled with ³²P for gel retardation analysis. The labelled probes were mixed with various concentrations of purified IdeR (0.06–1 µM) and then separated on 8% polyacrylamide gels (Fig. 1). Nickel was used as the divalent metal in the binding reactions on account of its redox stability compared with ferrous iron. The specificity of purified IdeR to its target sequences has been demonstrated in a previous work (Rodriguez et al., 1999). In addition, we have shown previously IdeR's broad in vitro divalent metal specificity and IdeR's metal dependence for function (Schmitt et al., 1995; Dussurget et al., 1996).

The migration of all four DNA fragments tested was retarded by IdeR in a concentration-dependent manner (Fig. 1). The labelled probes from mbtA–B, mbtI and rv3402c, with one IdeR box, had one major shifted product (B2), whereas bfd–bfrA, containing two IdeR binding sites, had two major shifted products (B2 and B3). All four fragments exhibited one minor intermediate band (B1) that appeared with the lower concentrations of IdeR, which may represent an intermediate with one IdeR dimer bound to the probe (White et al., 1998). These results indicate that IdeR can bind to these DNA fragments that contain putative IdeR-binding sequences.

IdeR binds to IdeR box sequences

We performed DNase protection assays with IdeR bound to the four DNA fragments containing putative IdeR boxes, mbtA–B, mbtI, rv3402c and bfd–bfrA, in order to identify the DNA sequences bound by the protein. The results of the DNase footprinting experiments are shown in Fig. 2, and the sequences protected by IdeR are summarized in Fig. 3. IdeR protected the predicted IdeR box sequence for mbtA–mbtB, mbtI, rv3402c and both IdeR boxes for bfd–bfrA. The protection pattern covers the 19 bp IdeR box sequence and also includes up to a 10 bp overhanging region of protection 3′ to the end of the IdeR box on both the top and the bottom strand. The sequences protected by IdeR are similar to those found in our previous studies with the M. smegmatis fxbA and the M. tuberculosis hisE–irg2 (Dussurget et al., 1999; Rodriguez et al., 1999), and these are included here for comparison (Fig. 3).

IdeR functions as a repressor by blocking the −10 region of the promoter

We mapped the transcriptional start points (TSP) of mbtB, mbtI, rv3402c, bfd and bfrA by primer extension analysis using RNA from M. tuberculosis grown in either low- or high-iron-containing medium 5, 6; data not shown), and the results are summarized in Fig. 3. The mRNA levels of mbtA were too low for this analysis. The TSP of mbtB, mbtI, rv3402c, bfrA and bfd all mapped within the IdeR box or 5 bp downstream of the box. We identified putative −10 sequences with similarity to the sequence TAgRcT (Gomez and Smith, 2000) that were the correct distance from the TSPs (Fig. 3). The mapping of the TSPs indicates that these IdeR boxes are located in close proximity to the −10 region, which fits a model that, in the presence of iron, IdeR functions as a repressor by occluding the site of RNA polymerase binding (Fig. 8), as is the case for DtxR binding to its target genes (Schmitt et al., 1992).

Genes preceded by an IdeR box are regulated by iron

The expression of genes with IdeR boxes was analysed in M. tuberculosis cultures grown with low and high levels of iron by reverse transcriptase–polymerase chain reaction (RT–PCR) coupled with molecular beacons (mbRT–PCR) (Manganelli et al., 1999). We used the sigA mRNA for normalization, as its levels remain constant in low- and high-iron concentrations (data not shown; Dussurget et al., 1999; Rodriguez et al., 1999). The average ratio of mRNA copies in low/high iron were 14 (mbtA), 49 (mbtB), 17 (mtbI), 17 (rv3402c) and 40 (bfd) (Fig. 4). The effect of iron on bfrA expression is described in the next section. As was predicted from the presence of an IdeR box in their promoter region, the transcription of these genes was indeed repressed by high iron levels.

IdeR acts positively on bfrA expression

We expected that the expression of bfrA, encoding an iron storage protein, bacterioferritin, would be upregulated in high-iron conditions as has been shown in other bacteria (Andrews, 1998; Miller et al., 2000). However, as all the other IdeR- and DtxR-regulated genes have been shown to be repressed by high iron levels (Schmitt et al., 1992; Schmitt and Holmes, 1994; Lee et al., 1997; Dussurget et al., 1999; Rodriguez et al., 1999), it was intriguing to study how bfrA would be regulated by IdeR. In addition, the bfrA promoter is unlike the other IdeR-regulated genes because it has two tandem IdeR boxes, beginning at 207 and 231 bp upstream of the annotated ORF start site (Fig. 5).

As we speculated that bfrA might be positively regulated, we wanted to determine whether the expression of this gene required IdeR and the two IdeR binding sites. To address these questions, a bfrA–lacZ fusion construct was made that contained 378 bp upstream from the bfrA initiating codon and was integrated into the chromosome of wild-type M. smegmatis MC²155 and its isogenic ideR mutant, SM3 (Dussurget et al., 1996). In addition, a bfrA–lacZ fusion was made with the tandem IdeR boxes deleted (bfrAΔIB–lacZ), and this construct was also incorporated into MC²155 and SM3.

Initial experiments using the intact bfrA–lacZ fusion integrated into M. smegmatis MC²155 demonstrated that levels of bfrA decreased (≈ twofold) during iron limitation. When the bfrA–lacZ and bfrAΔIB–lacZ constructs were integrated into M. smegmatis SM3, there was little or no lacZ expression (data not shown). The analysis of the bfrA promoter region was extended to locate the TSP, using RNA isolated from the M. smegmatis strains containing the M. tuberculosis bfrA–lacZ and bfrAΔIB–lacZ fusions. The primers used were specific for the M. tuberculosis bfrA and not endogenous M. smegmatis bfrA, as the primer extension products were not seen using RNA isolated from M. smegmatis not containing the M. tuberculosis promoter fusions (data not shown). The TSPs of bfrA, from both M. tuberculosis total RNA and the M. smegmatis strain carrying the M. tuberculosis bfrA–lacZ fusion were identical when grown in both high iron (Fig. 5A) and low iron (Fig. 5B), and the location of these TSPs is shown in Fig. 5C. The stronger band intensity of the M. tuberculosis P_low1 and P_low2 in Fig. 5B results from longer iron starvation. These results indicated that the regulation of the M. tuberculosis bfrA could be studied using M. smegmatis as a surrogate host.

A TSP was detected in high iron and was located 110 bp upstream of the ATG (P_high) 5, 6. P_high activity is decreased during iron starvation (Fig. 6B, lanes 1 and 2) and is dependent on IdeR, as the levels of P_high transcript are very low in the ideR mutant strain (Fig. 6B, lanes 3 and 4). If the IdeR boxes are deleted (bfrAΔIB–lacZ), the expression of P_high in high iron is reduced (Fig. 6C, lanes 1 and 2) compared with when the IdeR boxes are present (Fig. 6B, lanes 1 and 2). A combination of deleting the IdeR boxes and lacking IdeR completely abrogated bfrA P_high activity (Fig. 6C, lanes 3 and 4). The mRNA produced by P_high is not a processed mRNA produced by P_low, as it is made, albeit at lower levels, even when P_low is deleted (Fig. 6C, lanes 1 and 2). Two primer extension products were present only during iron starvation, which mapped to the distal IdeR box (P_low1 and P_low2) 5, 6. P_low1 and P_low2 were always present in equal amounts and will henceforth be named P_low to refer to both promoters. The repression of P_low in high iron is dependent on IdeR, as the transcription of P_low remains high in the ideR mutant (Fig. 6D). The activity of P_high was estimated to be 5–10 times stronger than P_low by quantitative RT–PCR and band intensity from primer extension (data not shown).

The contributions of P_low and P_high were measured more precisely using mbRT–PCR of M. tuberculosis RNA derived from cultures grown in low- and high-iron medium. The total amount of bfrA mRNA (P_low + P_high) decreased in a time-dependent manner to about 30% during iron starvation (Fig. 6E). Using RT–PCR with a beacon specific for the P_low mRNA (which recognizes a sequence 5′ to P_high), P_low was demonstrated to be induced about 11-fold during iron starvation (Fig. 6E). The decrease in bfrA mRNA (P_low + P_high) during iron starvation can be attributed to a combination of mRNA differentially expressed from both P_low and P_high(Fig. 8). P_high is expressed during growth in conditions of iron excess and requires IdeR for expression. During conditions of iron starvation, P_high is shut off, and IdeR releases its repression of P_low, allowing for low-level expression from P_low.

IdeR-controlled genes are derepressed during M. tuberculosis infection of human macrophage-like THP-1 cells

The levels of mbtB, mbtI, Rv3402c and bfrA were compared in broth cultures of M. tuberculosis and M. tuberculosis infection of the human monocytic cell line THP-1 that had been differentiated into macrophages (Fig. 7). THP-1 macrophages were infected with M. tuberculosis and, at various time points, macrophages containing internalized M. tuberculosis were harvested, and RNA was isolated for subsequent analysis by mbRT–PCR. sigA mRNA was used to normalize all ratios, as it had been determined previously that the M. tuberculosis sigA mRNA levels remain constant during M. tuberculosis infections of THP-1 cells (Manganelli et al., 2001; E. Dubnau et al., submitted). The expression of genes previously reported to be expressed during intracellular growth in macrophages, icl, encoding isocitrate lyase (McKinney et al., 2000), and hspX, encoding α-crystallin (Yuan et al., 1998), was analysed, and both were found to be strongly induced (data not shown; E. Dubnau et al., submitted). Two genes from the mycobactin biosynthesis operon were each induced to essentially the same levels at both 24 and 72 h: mbtB (≈ 38-fold) and mbtI (≈ fourfold) (Fig. 7). rv3402c was induced 11-fold by 24 h, and 25-fold by 72 h (Fig. 7). The levels of bfrA (P_low + P_high) did not change significantly from broth-grown cultures at either time point. To control for induction of gene expression during the 2 h period while the macrophages were ingesting M. tuberculosis, the levels of mRNA for all the genes were compared with cultures incubated in RPMI or 7H9 for 2 h. Only mbtB was responsive to a 2 h exposure to RPMI and was induced threefold, and the other genes were unaffected. Accordingly, if mbtB is normalized against RNA from M. tuberculosis exposed to RPMI for 2 h, its induction is 11-fold (versus 38 if normalized to 7H9). The induction of the mycobactin biosynthesis genes, mbtB and mbtI, suggests that the mycobacterial phagosome may be limiting for iron, consistent with previous findings (De Voss et al., 2000).

Identification of IdeR iron boxes

An initial low-stringency search of the M. tuberculosis genome using the blast program, querying for the DtxR/IdeR binding sequence, identified one promoter region of the divergently transcribed genes, hisE–irg2, which was in fact regulated by IdeR (Rodriguez et al., 1999). In this work, various combinations of the consensus DtxR iron box were used to probe the M. tuberculosis H37Rv published sequence (Cole et al., 1998) with the program tuberculist (http://genolist.pasteur.fr/TubercuList/) (Moszer et al., 1995). The query sequence used was TWAGGTWAGSCTWACCTWA (where W = A or T and S = G or C), and we allowed for up to five mismatches and limited the search to within −250 to +100 bp of each ORF's predicted start codon.

Bacterial strains, media and growth conditions

Escherichia coli strains JM109 and XL1-Blue (Stratagene) were used for cloning and grown in Luria–Bertani (LB) broth. Wild-type M. smegmatis mc²155 and the M. smegmatis ideR strain SM3 were grown in Middlebrook 7H9 liquid medium supplemented with 0.2% glycerol and 0.05% Tween 80, and M. tuberculosis strain H37Rv was grown in the same medium with the addition of 0.5% bovine serum albumin (BSA) fraction V, 0.2% dextrose and 0.085% NaCl (ADC supplement). When required, media were supplemented with 20 µg ml⁻¹ streptomycin, 10 µg ml⁻¹ kanamycin and 100 µg ml⁻¹ ampicillin. To assay gene expression by mbRT–PCR or primer extension analysis in low and high iron, M. tuberculosis was grown in MMT-ADN (minimal medium with 0.05% Tween 80, treated overnight with Chelex 100; Bio-Rad) supplemented with either 0 or 50 µM FeCl₃. Final concentrations of iron in low-iron MMT-ADN have been shown previously by atomic absorption spectroscopy to be in the range of < 1 µM iron (Rodriguez et al., 1999). In order to achieve conditions of iron starvation, M. tuberculosis H37Rv was passaged for two generations in low-iron medium and then diluted to an optical density of 0.05 in the same medium lacking iron or in a high-iron medium.

β-Galactosidase assays

β-Galactosidase assays were performed as described previously (Miller, 1972; Dussurget et al., 1999). Briefly, M. smegmatis was grown in LB medium supplemented with 0.1% Tween to mid-late logarithmic phase (OD₆₀₀ = 0.4–0.6), which was made iron limiting by the addition of 2,2′-dipyridyl to 200 µM and grown for an additional 3 h.

bfrA–lacZ constructs

A 430 bp construct encompassing −378 to +52 of bfrA was amplified by PCR, cloned into the vector psM128 and integrated in single copy into the att site of M. smegmatis wild type (MC²155) and ideR mutant (SM3) (Dussurget et al., 1999). A construct was made by deleting the two iron boxes (−255 to −199), resulting in bfrAΔIB–lacZ, which contains an additional XhoI site. Constructs were verified by sequencing.

IdeR purification

IdeR was overexpressed in M. smegmatis, and the protein was purified by techniques similar to those described previously but with some modifications (Schmitt et al., 1995; Pohl et al., 1999). The wild-type copy of the M. tuberculosis ideR was cloned downstream of the L5 mycobacteriophage promoter, P_left, and then placed into pSM232 (pMV261 + streptomycin/spectinomycin resistance cassette), resulting in pSM273. Strain SM84 was created by transforming the ideR mutant, SM3, with pSM273. Functional IdeR activity was visualized by complementation of SM3's derepression of siderophore synthesis on CAS high-iron plates (Dussurget et al., 1996). A culture of SM84 was grown to stationary phase in 7H9 medium, and cells were harvested and resuspended in buffer A (50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 2 mM β-mercaptoethanol) containing a protease inhibitor cocktail (Complete; Boehringer Mannheim). Cells were disrupted by passaging three times by French press at 1200 psi. Cell debris and lipids were removed by 30 min ultracentrifugation at 20 000 r.p.m. Crude extracts were loaded onto a nickel–NTA column (Qiagen) and separated by fast protein liquid chromatography (FPLC) using an imidazole gradient of 0–100 mM in buffer A. IdeR was eluted between 20 and 30 mM imidazole. Fractions containing IdeR were pooled and dialysed against 50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM dithiothreitol (DTT) and 1 mM EDTA to remove residual nickel. A second FPLC purification was performed using anion exchange on a Source 15Q column (Amersham Pharmacia Biotech) with 50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM DTT and an NaCl gradient of 50–500 mM. The major IdeR peak eluted between 65 and 85 mM NaCl. IdeR was > 99% pure as verified by Coomassie staining and Western blotting, with a yield of about 150 µg of purified IdeR l⁻¹ culture.

Gel retardation assay

The gel retardation assay was carried out essentially as described previously (Hamoen et al., 1998; Rodriguez et al., 1999) with minor modifications. PCR fragments containing the putative IdeR consensus binding sites were end labelled with T4 polynucleotide kinase and [γ-³²P]-ATP. Binding reactions [20 mM Tris-HCl at pH 8, 1 mM DTT, 50 mM KCl, 5 mM MgCl₂, 0.05 mg ml⁻¹ poly-(dI–dC), 0.05 mg ml⁻¹ BSA and 10% glycerol] containing ≈ 0.1 ng of probe (20 000 c.p.m.) were incubated with purified IdeR (0.06–1 µM) in 20 µl volumes for 30 min at room temperature. NiSO₄ (200 µM) was added to binding reactions and, when required, to the acrylamide gels and running buffers. Reactions (15 µl) were loaded without dye to an 8% polyacrylamide gel containing 40 mM Tris acetate at pH 8. Gels were run at 110 V at room temperature, dried, and bands were visualized by autoradiography.

DNase footprint analysis

Primers corresponding to either the top or bottom strands of the following promoter regions, mbtA–mbtB, mbtI, rv3402c, bfrA–bfd, were labelled with T4 polynucleotide kinase and [γ-³²P]-ATP and then used to PCR amplify the corresponding fragments. PCR products were ethanol precipitated and then purified on a 2% agarose gel. Binding reactions with IdeR were performed as described for the gel retardation assay in 20 µl with ≈ 100 000 c.p.m. probe and 1 µM IdeR. Reaction volumes were adjusted to 100 µl with a Ca²⁺/Mg²⁺ solution (final concentrations are 2.5 mM CaCl₂ and 5 mM MgCl₂). Then, 0.15 U of DNase (Promega) was added, and mixtures were incubated for 1 min at room temperature. Reactions were terminated by the addition of 90 µl of stop solution (200 mM NaCl, 20 mM EDTA, 1% SDS and 100 µg ml⁻¹ yeast RNA). Samples were subsequently extracted with phenol–chloroform, ethanol precipitated and resuspended in a formamide loading dye. Samples were electrophoresed on a 6% TBE polyacrylamide–urea sequencing gel, dried and exposed by autoradiography. Maxam–Gilbert A+G sequencing reactions were performed to identify protected regions.

Primer extension analyses

RNA was extracted essentially according to the protocol of Manganelli et al. (1999). M. tuberculosis H37Rv was grown as described above in low- and high-iron MM-ADN to late log phase. Bacterial pellets were resuspended in LETS buffer (100 mM LiCl, 10 mM EDTA, 10 mM Tris, pH 7.8, 1% SDS) and then mixed with an equal volume of phenol–chloroform–isoamyl alcohol (P:C:IAA; 25:24:1) and broken with a bead beater. Membranes were removed by centrifugation at 14 000 r.p.m. for 15 min, and the resulting supernatant was extracted once in P:C:IAA and twice with TRI reagent (Molecular Research Center), according to the manufacturer's protocol. RNA integrity was verified by electrophoretic analysis on a 2% TBE agarose gel. For primer extension, 20–60 µg of RNA was hybridized to 1.5 pmol of γ-³²P-labelled reverse primer in 1×Carboxydothermus hydrogenoformans reverse transcription buffer (C. therm. kit; Roche) in a final volume of 10 µl. The hybridization programme was 94°C for 1.5 min, 64°C for 3 min, 57°C for 5 min, and then the tubes were allowed to cool rapidly to 50°C, when they were placed directly on ice. A reverse transcription mixture using the C. therm reverse transcriptase (Roche; final of 1 mM dNTPs, 2.5 mM MgCl₂, 5% DMSO and 5 mM DTT) was added to the annealed primer–RNA mixture at a final volume of 20 µl. The cDNA synthesis reaction was at 60°C for 1 h followed by 2 min at 94°C, and then the cDNA–RNA mixture was ethanol precipitated and washed with 75% EtOH. Samples were resuspended in 6 µl of water and 4 µl of formamide loading buffer and separated on a 6% TBE gel containing 8 M urea at 1800 V. In some cases, a 10 bp ladder (Gibco BRL) was used as a molecular weight standard to determine the length of the primer extension product and, in other cases, a sequencing reaction was performed using the Sequenase 7-deaza-dGTP kit (USB). All transcriptional start points were verified using two primers whose 3′ ends were 10–20 bp apart.

Molecular beacon RT–PCR (mbRT–PCR)

To obtain RNA, frozen bacterial pellets were resuspended in 1 ml of TRI reagent and 500 µl of silica beads and then broken using a bead beater at maximum speed for 2 × 1 min, with chilling on ice between breaking periods. RNA extraction was performed according to the manufacturer's instructions (Molecular Research Center). The mbRT–PCR was performed essentially as described by Manganelli et al. (1999; 2001), and cDNA was synthesized as described above for primer extension assays. A cocktail of up to five reverse primers was added per reaction to have all reverse transcriptions (always including sigA) performed in the same tube. RNA (50–200 ng) was added to each reverse transcription and, in all cases, a mock reaction lacking reverse transcriptase was performed to obtain copy numbers of DNA contaminating the RNA sample. The PCR reactions typically contained 100–800 nM of a gene-specific molecular beacon (coupled to tetramethylrhodamine or fluorescein) and were performed with the following PCR programme: 95°C for 10 min, 10 cycles of 95°C for 20 s, 65°C for 30 s and 72°C for 30 s, and then 35 cycles of 95°C for 20 s, 60°C for 30 s and 72°C for 30 s. Molecular beacons were designed with the aid of the mfold server (http://bioinfo.math.rpi.edu/~mfold/dna/form1.cgi) and synthesized as described previously (Tyagi and Kramer, 1996). All molecular beacon and primer sequences are available upon request. Ratios of induction at each time point were solved using the following formula, where X and A are the absolute number of copies of mRNA from geneX or sigA respectively: (X_{low Fe}/A_{low Fe})/(X_{high Fe}/A_{high Fe}).

Isolation of RNA from THP-1 macrophages infected with M. tuberculosis

Mycobacterium tuberculosis infections of THP-1-derived macrophages were performed as described by Manganelli et al. (2000). Briefly, THP-1 cells were grown in RPMI-1640 (supplemented with 0.45% glucose, 0.15% sodium pyruvate and 4 mM l-glutamine), differentiated into macrophages by 24 h treatment with 50 nM 12-O-tetradecanoylphorbol-13-acetate (PMA) (Tsuchiya et al., 1982) and plated at a final concentration of 7.5 × 10⁵ cells ml⁻¹. To improve long-term macrophage adherence in large volumes, flasks were pretreated with 0.2% gelatin overnight at 4°C. M. tuberculosis H37Rv were grown in 7H9 to an OD₅₄₀ of 0.2 and infected into the THP-1 cells at a multiplicity of infection of 0.5–1. Extracellular bacteria were removed after 2 h by washing twice with phosphate-buffered saline (PBS), and then fresh RPMI containing 50 µg ml⁻¹ gentamicin and 20% fetal calf serum (FCS) was added back to the macrophages. At 24 and 72 h, RPMI was removed, macrophages were washed once with RPMI, and then the remaining cells were resuspended in 10 ml of TRI reagent containing polyacryl carrier, frozen immediately on dry ice and then stored at −80°C. At each time point, macrophages were checked for viability using Trypan blue exclusion (Sigma), and viable counts of M. tuberculosis growing intra- and extracellularly were performed in a parallel infection. Expression levels of all genes were normalized against M. tuberculosis that had been exposed to RPMI for 2 h or to an M. tuberculosis rolling culture in 7H9. To make RNA, 500 µl of silica beads was added to the frozen mix of M. tuberculosis and macrophages, and cells were broken using a bead beater at maximum speed for 2 × 1 min, with chilling on ice between breaking periods. RNA extraction was performed with TRI reagent according to the manufacturer's instructions (Molecular Research Center). RNA was subjected to DNase treatment (DNA-free kit; Ambion) and then precipitated and resuspended in water treated with diethylpyrocarbonate (DEPC). Ratios of induction at each time point were calculated using the following formula, where X and A are the absolute number of copies of mRNA from geneX or sigA respectively: (X_{intracellular}/A_{intracellular})/(X_{extracellular}/A_{extracellular}).