Functional Demonstration of Reverse Transsulfuration in the Mycobacterium tuberculosis Complex Reveals That Methionine Is the Preferred Sulfur Source for Pathogenic Mycobacteria* □ S

Methionine can be used as the sole sulfur source by the Mycobacterium tuberculosis complex although it is not obvious from examination of the genome annotation how these bacteria utilize methionine. Given that genome annotation is a largely predictive process, key challenges are to validate these predictions and to fill in gaps for known functions for which genes have not been annotated. We have addressed these issues by functional analysis of methionine metabolism. Transport, followed by metabolism of 35 S methionine into the cysteine adduct mycothiol, demonstrated the conversion of exoge-nous methionine to cysteine. Mutational analysis and cloning of the Rv1079 gene showed it to encode the key enzyme required for this conversion, cystathionine (cid:1) -lyase (CGL). Rv1079, annotated metB , was predicted to encode cystathionine (cid:1) -synthase (CGS), but demonstration of a (cid:1) -elimination reaction with cystathionine as well as the (cid:1) -replacement reaction yielding cystathionine showed it encodes a bifunctional CGL/CGS enzyme. Consistent with this, a Rv1079 mutant could not incorporate sulfur from methionine

The ability of living organisms to produce or acquire sulfur compounds is central to biological processes. Key sulfur-containing metabolites derived from methionine include N-formyl methionine for the initiation of peptide chain biosynthesis and S-adenosylmethionine for many one-carbon transfer reactions.
Compounds, such as glutathione or its replacement in the actinomyctes, mycothiol, are synthesized from cysteine for redox maintenance. The mshC gene in the biochemical pathway for mycothiol biosynthesis is essential for growth of Mycobacterium tuberculosis (1). Coenzyme A, also derived metabolically from cysteine, is required for many reactions in intermediary and lipid metabolism. Since mycobacteria have both a high proportion of genes for lipid biosynthesis (2) and a notably high and diverse lipid content, the production of coenzyme A must be a core activity. Thus, for pathogenic mycobacteria, acquisition of sulfur from the host is a major requirement for growth in vivo.
Our interest in the pathways for the biosynthesis and utilization of cysteine and methionine arose from the observation that methionine, but never cysteine, auxotrophs of the M. tuberculosis complex could be isolated (3,4). These observations, together with the lack of further attenuation of a methionine auxotroph of Mycobacterium bovis BCG (4), suggested methionine might be an important source of sulfur in the host. However, this view conflicted with the lack of any obvious route for the conversion of methionine to cysteine. Cystathionine ␥-lyase (CGL) 1 is the key enzyme required for the operation of the reverse transsulfuration pathway (Fig. 1, unique reactions shown by blue arrows inside yellow box) from methionine to cysteine most likely to be needed for methionine auxotrophy. However, the four genes (Rv0391, Rv1079, Rv3340, and Rv3341) encoding enzymes of the appropriate protein family (5,6) in the genome of M. tuberculosis (2) were annotated without identifying a gene to encode CGL. One of these genes, Rv1079, annotated metB and predicted to encode cystathionine ␥-synthase (CGS), was also similar to CGL genes. During the course of our work, using a Bayesian method for "pathway hole" filling to identify enzymes missing in metabolic pathways constructed from annotated genomes, Rv1079 was identified as a CGL (7). Given that members of this protein family are enzymes with broad substrate specificity (5,6,8,9), we decided to investigate the activities Rv1079 encodes by both mutational analysis and cloning the gene.
Although reverse transsulfuration is a likely route for the conversion of methionine to cysteine, at least three biochemical pathways have been described for microbial conversion of methionine to cysteine (Fig. 1). It is important to establish which occur in the M. tuberculosis complex because if alternative pathways exist, their relative importance, for instance during growth within the host, can be ascertained. Therefore, we carried out a complete re-examination of sulfur amino acid metabolism in the M. tuberculosis complex. Briefly, our strategy was to trace the metabolic fate of 35 S-labeled amino acids, perform experiments with the ␥-lyase/␥-synthase inhibitor (see Fig. 1 to view its sites of action) propargylglycine (10), and to assay key enzymes in the pathways of sulfur amino acid metabolism to test and extend the genome annotation and to corroborate our own work, described herein, on Rv1079. A more general issue we are addressing is the continual need to refine and revise genome annotation. This is particularly important in research on the M. tuberculosis clade that includes the strains used in this study, M. bovis BCG and M. tuberculosis H37Rv, which have over 99% sequence identity. Hitherto almost intractable to investigation, expression profiling (11) and subtractive hybridization (12) of these important pathogens have allowed insights into the genes required for their virulence and those that are expressed once inside the host. Interpretation of this rapidly emerging data requires accurate annotation of the genes identified, and our work contributes to this process. Table I. The strains of mycobacteria used were M. tuberculosis H37Rv and its metB::hyg mutant (13) and M. bovis BCG Pasteur (Statens Serum Institute) and its cysA::Tn5367 mutant in which the Tn5367 transposon carries a kanamycin resistance gene (4) ( Table I). It should be noted that we shall use the original designation of the M. tuberculosis metB::hyg mutant throughout even though, as will become apparent, the gene name is somewhat misleading. These strains were routinely grown in Middlebrook 7H9 medium with 10% ADC (albumin-dextrose-catalase) supplement (Difco) plus 0.05% (w/v) Tween 80 at 37°C in Roller bottles rotated at 4 rpm or, where specifically stated, in square-bottomed bottles that were shaken daily and the OD 600 of the cultures determined. Hygromycin (100 g/ml) or kanamycin (25 g/ml) was included in cultures of the mutants with the corresponding antibiotic resistance. Modifications made to the media for experiments specified in the text were (i) making "sulfate-free 7H9" in which all the sulfate (4.16 mM), present as magnesium sulfate, was substituted with magnesium citrate keeping the concentrations of Mg 2ϩ and citrate ions the same as in the standard medium and (ii) the addition of methionine or cysteine, and (iii) the addition of propargylglycine (2-amino-4-pentynoic acid) at various concentrations as shown under "Results." Dry weights were determined by drying samples on cellulose acetate (0.22-m pore size) filters to constant weight for the BCG strains and correlated to OD 600 . Dry weights of virulent bacteria were estimated from the OD 600 of their cultures.

Strains of Tubercle Bacilli and Growth Conditions-Strains and plasmids used in this study are listed in
Generation of an Rv1079 Construct in Escherichia coli-E. coli DH5␣ was used as a general purpose cloning host, and E. coli⌬metB (ATCC 35636, F-metB rel-1 PO2B lambda), a methionine-requiring host strain (14) was used as the overexpression host for Rv1079. The E. coli strains were grown in Luria-Bertani (LB) medium with ampicillin (100 g/ml) as required.
The Rv1079 coding region from H37Rv chromosomal DNA was amplified using the primers metBF (5Ј-CGACCGTCATATGAGCGAAGAC-CGCACGGGAC-3Ј) and metBR (5Ј-CGTCTGCTGCGGCCGCTTAAC-CCAGGGCCTGTT-3Ј). PCR was carried out using Advantage-HF PCR Kit (Clontech) following the manufacturer's instructions. The amplified DNA was cloned into the pGEM-T Easy vector (Promega) to generate construct pLK99. After sequencing, the insert was excised from this vector by digestion with NdeI and NotI (restriction sites underlined) and cloned into NdeI/NotI-restricted pET21a (Novagen), generating the plasmid pLK101.
Complementation of E. coli ⌬metB-E. coli ⌬metB was lysogenized using a DE3 lysogenization kit (Novagen) to allow expression of target genes cloned in the pET vector under the control of the T7 promoter. pLK101 and pET21a were transformed into E. coli ⌬metB(DE3) and transformants selected on agar plates containing 100 g/ml ampicillin (Amp). Isolated colonies were selected from the plates, checked for the presence of the plasmid by PCR, and then grown in LBAmp to yield stocks of E. coli ⌬metB(DE3)/pET21a and E. coli ⌬metB(DE3)/pLK101. These strains were plated out on isopropyl-1-thio-␤-D-galactopyranoside (IPTG, 0.4 mM) and glucose (0.4%w/v) supplemented M9 minimal medium (15) in the presence and absence of methionine (0.15 mM) to demonstrate Rv1079 complementation of the metB mutation. For protein extraction E. coli ⌬metB(DE3)/pET21a and E. coli ⌬metB(DE3)/ pLK101 were grown in 0.4 mM IPTG supplemented LB broth with Amp.
Chemicals and Radiochemicals-All reagents were the purest grade available from Sigma or Merck (AnalaR or HPLC grade) except Oacetylhomoserine, which was prepared by acetylation of L-homoserine with acetic anhydride (16). The O-acetylhomoserine thus prepared was 98% pure with the 2% contaminant being unreacted homoserine (quantified by HPLC-MS, see below for methods). HPLC solvents were from Rathburn. Water was MilliQ grade. Radiochemicals were from Amersham Biosciences: to optimize stability we purchased methionine at low specific activity (Ͻ18.5 Bq/pmol) and added 0.5 mM dithiothreitol, and 40 M bathophenantholinedisulfonic acid (BDPS) to cysteine (ϳ37 kBq/ pmol) on receipt.
HPLC Instrumentation-All derivatives of metabolites were chromatographed on a Symmetry C8 (Waters) reverse phase HPLC column (3.9 ϫ 150 mm). UV analysis was accomplished with a Waters 625 LC system equipped with a 486 UV detector, whereas HPLC-RA or HPLC-MS employed a HP1050 HPLC system connected in series to either an A525 radioactivity-detector (Canberra Packard, Pangbourne, Berks) or an LCQ mass spectrometer (ThermoFinnigan, Hemel Hempstead, Berks), respectively.
Uptake of Methionine and Cysteine into Intact Bacteria-Initially, bacteria, at Ͼ1 mg (dry wt.)/ml, were suspended in 50 mM MOPS buffer at pH 7. The radioactivity taken up was determined by collecting bacteria on glass fiber filters followed by scintillation counting (4).
In incubations with methionine, (i) to obtain kinetic data, suspensions of bacteria were incubated for 8 or 12 min to obtain initial rates. Three experiments were performed and the range of K m and V max values obtained is reported. (ii) Subsequently, for metabolic-tracing experiments, bacteria were grown as 20-ml cultures in 7H9 with 60 M methionine at similar rates (mean generation time ϭ 1.0 -1.2 days) to OD 600 of 0.05, then [ 35 S]methionine (1.00 MBq; 0.054 mol) was added and the cultures harvested 4 -5 days later, still during log phase, when the OD 600 reached 0.4 -1.0. During the period of culture including [ 35 S]methionine, 1.8 -3.4% of the methionine was taken up per mg (dry wt.) bacteria. Cultures at the end of the period contained 3-6 mg of bacteria. Thus, typically, around 10% of the radioactivity in the culture flask was taken up.
In incubations with cysteine, uptake was rapid for 8 min but continued thereafter so that about 3ϫ more had been taken up by 1 h. Radioactivity was taken up at similar rates for 10 nM and 20 M cysteine, but at half the rate with 50 M cysteine. For metabolic tracing experiments, because cysteine is toxic in the Middlebrook media used in this study (4) (further results in this report), bacteria were grown to and OD 600 of 0.4 -0.8 (log phase), then 740 kBq of [ 35 S]cysteine and unlabeled cysteine to give a final concentration of 20 M was added, and the incubation was continued for 4 h.
Control incubations comprised sufficient heat-killed bacteria from a suspension of M. bovis BCG in 50 mM MOPS buffer at pH 7.5 with 0.4 mM MgCl 2 heated at 100°C for 15 min to give a pellet of 25 l. This was mixed with 150 kBq of 35 S-amino acid, then acid and solvent was added to the mixture, which was analyzed for thiols or amino acids by the methods given below starting from the first pellet stage.
Detection of Thiols Formed as Bimane Derivatives-Monobromobimane (mBBR) derivatives were prepared to give stable, UV-absorbing derivatives of thiols. The principle of this procedure is that the rapid reaction with mBBR minimizes losses of the unstable thiols (17). At the end of uptake experiments, the bacteria were collected by centrifugation for 5 min at 4000 ϫ g at 37°C. Typically the pellet was Ͻ50 l. Immediately (without washing) the pellet was acidified to pH 3-4 with 100 l of 10 mM methane sulfonic acid, then 100 l of acetonitrile was added, and the pellet was resuspended. The suspension was frozen and thawed twice then adjusted to pH 8 with 200 l of the following mixture: 0.5 M HEPPS-KOH, pH 8, water (MilliQ), 0.25 M KOH (16: 174:10, v/v/v). The alkaline suspension was mixed rapidly, 7 l of 150 mM mBBR in acetonitrile was added immediately, mixed, and allowed to react at 60°C for 20 min. The mixture was then acidified to pH 4 -5 with 1.25 M methane sulfonic acid, and clarified by centrifugation and, if necessary, filtration. This material was injected directly into HPLC in 20-l portions: in experiments with radiochemicals, this volume gave sufficient (Ͼ2 KBq) radioactivity for analysis. After injection, a gradient of 41 mM acetic acid adjusted to pH 3.5 with ammonium hydroxide (solvent A) and methanol (solvent B) was run: 0 min, 100% A; 25 min 76% A; 50 min 20% A; 52 min 20% A; 56 min 100% A. The flow rate was 1 ml/min. The major thiol detected was mycothiol, ranging from 19 nmol/mg (dry wt.) bacteria in M. bovis BCG grown with 270 M methionine to 4 nmol/mg (dry wt.) bacteria in M. bovis BCG ⌬cysA grown with 60 M methionine. These values are in agreement with the literature (18). No radioactive N-acetylcysteine was formed showing that the above protocol effectively preserved the bacteria after the end of the incubation and did not provoke a detoxifiction response (19). Mass spectra of metabolites were collected over the range m/z ϭ 150 -800 using the electrospray probe in the positive ion mode. Product ion mass spectra of protonated molecular ions were collected with a collision energy of 35% and an isolation width of 2 amu (atomic mass units).
Detection of Amino Acids-Typically, pellets of bacteria were treated as described above for thiol detection, except that (i) they were washed once in 1.5 ml of MilliQ water when the mass of methionine and cystathionine was to be quantified and (ii) ethanol was used instead of acetonitrile. After acid-ethanol extraction, the mixture was either derivatized as such or centrifuged to separate soluble, pooled amino acids and insoluble material including protein, which were derivatized separately. The material was dried under vacuum. The insoluble fraction was hydrolyzed for 18 h at 105°C in constant boiling 6 M HCl (Fisher Chemicals) purged with nitrogen, and treated with Na 2 SO 3 (20). In our hands, we recovered Ͼ80% of L-cysteine treated in this way.
PTC derivatives were prepared and analyzed by the method of Pramanik et al. (21). In our hands, we found that the ion pair reagent (triethylamine) completely suppressed the ionization and thus the identification of analytes. Removal of triethylamine required adjustment of the gradient by lessening the content of organic modifier. Thus, the optimal gradient used in this study was with 14 mM ammonium acetate adjusted to pH 6.4 with acetic acid as solvent A and acetonitrile/water (3:2 by vol) as solvent B: 0 min 100% A; 4 min 58% A; 16 min 58% A; 24 min 55% A; 25 min 10% A; 29 min 10% A; 30 min 100% A. The flow rate was 1 ml/min. Mass spectra of metabolites were collected as described above for bimane derivatives.
Identification of Metabolites-Radioactive metabolites were identified, and where indicated, quantified as Bq, by matching corresponding peaks from selected ion and HPLC-RA traces and comparison of product ion mass spectra with authentic reference chemicals where available. When metabolites were quantified as moles, keto acids and thiols were quantified by the peak area of their UV-absorbing derivatives and amino acid derivatives were quantified by the abundance of their mass ions.

Preparation of Extracts and Enzyme
Assays-Cell-free extracts were prepared by ultrasonication of washed bacteria and clarified as described previously for mycobacteria (22) and by the same method but with 40 s (2 ϫ 20 s) ultrasonication for E. coli, then desalted using a PD-10 column (Amersham Biosciences) according to the manufacturer's instructions. The buffer used throughout was designed to maintain enzyme activities (23). ␥and ␤-Lyases and cysteine desulfhydrase activities were determined by assaying for the keto acids produced (24). Concentrations of substrates are given in the results section. CGS (MetB) was assayed for both the elimination reaction by detecting 2-ketobutyrate (24) and the substitution reaction (25) by stopping reactions by adding 4 l of 1.25 M methane sulfonic acid, then detecting cystathionine formed by HPLC-MS of its PTC-derivative (see "Detection of Amino Acids" and Fig. 3). All enzyme assays were buffered with 50 mM potassium phosphate at pH 7.5. In all assays with cysteine, 1.25 mM dithioerythritol was included, or if the concentration of cysteine was above 1 mM, a 25% molar excess of dithioerythritol was included. The dithioerythritol was not transformed in any of the assays conducted (results not shown). Reactions were started by the addition of cell-free extract except when the effect of propargylglycine was tested (26,27). Then, extracts were preincubated with propargylglycine (range of concentrations: 4 -100 M) for 30 min at 37°C before starting the reaction by adding substrate. In all the protocols above, reactions were stopped by acidification to pH 3 or lower (24).

Growth Experiments with M. bovis BCG cysA::Tn5367 That
Can Be Used to Interpret Biochemical Pathways-If a methionine auxotroph is more susceptible to propargylglycine than its parent strain, this provides a line of evidence that it is dependent upon the propargylglycine-sensitive steps shown in Fig. 1, and that alternative pathways are not operating. First, we confirmed that M. bovis BCG cysA::Tn5367 used methionine as its sole sulfur source by showing its growth was limited by methionine. When Middlebrook 7H9 medium was supplemented with 30 M methionine, the mean generation time was 1.0 -1. All the pathways and genes or proteins considered in this report are shown in this figure. The pathways for which there are functional data are shown in thick lines; those which have been ruled out are shown in thin, dashed lines. A color code is used to denote the different pathways as follows: orange, biosynthetic and sulfur recycling, with inorganic sulfur as product or substrate; magenta, the unique reactions of the transsulfuration pathway from cysteine to methionine; blue, the unique reactions of the reverse transsulfuration pathway from methionine to cysteine; black, other relevant pathways and reactions shared by more than one pathway. Propargylglycine-sensitive steps are denoted by double green cross-lines. Genes and enzyme names are as follows: red, shown functionally; turquoise, present, discussed in this report; black with red strikethrough, shown to be absent. Cysteine desulfhydrase was not annotated in the M. tuberculosis genome.
unsupplemented Middlebrook 7H9, which supported growth of all the other strains (Table I) used in this study.
These data suggest that the methionine auxotroph relies on the activity of either CGL or methionine ␥-lyase (MdeA) (see Fig.  1) for its growth. These two activities were assayed for, revealing only CGL (Table II; discussed fully in the section "Enzyme Activities in Cell-free Extracts of Mycobacteria"). The results are discussed in full later, after all the work with intact bacteria is described. The alternative pathway via 3-methylthioproprionate (orange in Fig. 1) is not sensitive to propargylglycine (9, 28); thus, it cannot be operating in the M. tuberculosis.
Growth Experiments with M. tuberculosis H37Rv metB::hyg Used to Interpret Biochemical Pathways-None of the strains we used grew in sulfate-free Middlebrook 7H9. All the strains, except the metB::hyg mutant, grew with 270 M methionine added to the sulfate-free medium. None of the strains grew with cysteine (tried at 30, 60, and 300 M) as sole sulfur source. These results indicated that disruption of the metB gene, unexpectedly, disrupted methionine utilization. Therefore, further experiments were designed both to investigate sulfur amino acid metabolism in general, and to obtain more lines of evidence to deduce the role of the metB gene (Rv1079) in sulfur metabolism.
Uptake of Methionine-The next series of experiments, tracing labeled methionine, were designed to provide further evidence for the uptake and metabolism of methionine and to indicate the biochemical pathways that might be used so as to enable us to identify key enzymes to investigate further. First, by measuring the rate of uptake and affinity for methionine with intact bacteria, we would be able to discuss the relative rates that sulfate and methionine can be taken up by comparing with a previous report on sulfate transport using M. bovis BCG (4).
The apparent K m for methionine uptake into intact M. bovis BCG, was 80 M (range: 68 -93 M) and the V max for 1 mg (dry weight) bacteria was 440 pmol/min (range: 427 pmol/min to 453 pmol/min). For its cysA mutant, the apparent K m was not notably different at 86 M (range: 78 -95 M) but the V max for 1 mg (dry weight) bacteria was consistently higher at 563 pmol/ min (range: 537-588 pmol/min). As the uptake experiments involved taking rapid readings that were difficult to perform in Category III microbiological containment conditions, and we did not have a methionine auxotroph to compare with the H37Rv strain, we did not perform uptake experiments with M. tuberculosis H37Rv. However, we did show it metabolized methionine to mycothiol (below), an activity that requires the methionine to be taken up.
Incorporation of Sulfur from Methionine into Mycothiol, a Metabolic Product of Cysteine-We were unable to detect cysteine, either labeled or unlabeled, in the soluble metabolites of M. bovis BCG or M. tuberculosis H37Rv (see, for example, Fig.  2). Cysteine could be retrieved in the form of cystine from hydrolyzed protein, but proved to be too unreliable to achieve to use as a standard assay (see Supplemental Materials). There-   M. tuberculosis H37Rv incorporated sulfur from methionine into mycothiol more slowly, at 0.69 Ϯ 0.25 kBq/mg (dry wt.) bacteria (n ϭ 2), but this was clearly and significantly above the limit of detection, 0.015 kBq/mg (dry wt.) bacteria. No incorporation was detected in 4 determinations in the metB::hyg mutant of H37Rv, or heat-killed bacteria.
The evidence for incorporation of the labeled sulfur into mycothiol was obtained by identifying thiols in acid-acetonitrile soluble material as their mBBR derivatives (Fig. 2). The radiolabeled peak eluting at 19.90 min in HPLC-RA corresponded to a peak at 20.26 -20.29 min on HPLC-MS that yielded a protonated molecular ion at m/z ϭ 677 consistent with the mBBR derivative of mycothiol. The identity of this metabolite as mycothiol was further supported by ions at m/z ϭ 497 and 694 on full MS of ions collected between 19.99 and 20.49 min. These ions correspond with the desinositol fragment of mycothiol (18) and an ammonium adduct respectively. For other possible labeled products, protonated molecular ions of N-acetylcysteine (m/z ϭ 354) at 29.87 min, but not mass ions corresponding to cysteine (m/z ϭ 312) or homocysteine (m/z ϭ 356) derivatives were evident. However, only the mycothiol-mBBR corresponded to any of the radioactive peaks (Fig. 2). Note that whether UV absorbance or total ion current was used to detect peaks (Fig. 2), the peak that corresponds to both radioactivity and the mycothiol is the second peak of a distinct doublet. In this doublet, the first peak is never labeled and appears in DTNB (5,5Ј-dithiobis-[2-nitobenzoic acid])-pretreated samples (data not shown), showing that it is not a thiol, while the second is abrogated by DTNB pretreatment showing that it is a thiol. These metabolic tracing results provide further evidence that the sulfur from methionine is incorporated into the cysteine molecule, and the lack of incorporation in the metB::hyg mutant provides further evidence that this gene is involved in methionine utilization.
Accumulation of Cystathionine in M. tuberculosis H37Rv metB::hyg-Yet more evidence that the Rv1079 gene is involved in a biochemical pathway for converting methionine to cysteine would come from accumulation of one of the intermediates shown in Fig. 1 when the gene is disrupted. In the absence of propargylglycine, cystathionine accumulated in the pool amino acids of M. tuberculosis H37Rv metB::hyg but not the other strains used in this study. When duplicate cultures were grown to mid-log phase (OD 600 ϭ 0.6 -0.8) in the presence of 270 M methionine, 2.94 and 4.13 nmol of cystathionine per mg (dry wt.) was detected while in the parent strain (H37Rv), 0.07 and 0.14 nmol per mg (dry wt.) was detected. When 20 M propargylglycine, a CGL inhibitor, was included in the medium, accumulation to 9.61 nmol per mg (dry wt.) was detected in the parent strain. Cystathionine also accumulated in M. bovis BCG grown in the presence of methionine and propargylglycine (see Supplemental Materials). Cystathionine was always quantified from the abundance of mass ions of its PTCderivative (Fig. 3). In all these HPLC-MS runs, methionine could be detected but cysteine and cystine PTC-derivatives were absent, consistent with the results obtained in tracing experiments in which mBBR derivatization was used (Fig. 2). The m/z values for protonated molecular ions of PTC derivatives obtained in these experiments and in the following section are reported together in the following section: "Incorporation of Sulfur from Cysteine into Methionine." Incorporation of Sulfur from Cysteine into Methionine-Taken together, the failure of the metB::hyg mutant to incorporate sulfur from methionine into mycothiol, and the accumulation of cystathionine when grown with methionine, suggests that disruption of the Rv1079 gene removes CGL activity. This suggested that the annotation of Rv1079 as metB might be erroneous so we decided to investigate the conversion of cysteine to methionine, in which metB is central (see magenta arrows in yellow box, Fig. 1). In a representative experiment with intact bacteria, [ 35 S]cysteine (to 20 M: 740 kBq) was added to a growing culture of M. tuberculosis H37Rv metB::hyg. The bacteria took up 2.19% of the radioactivity per mg (dry wt.) bacteria in 4 h. After hydrolysis of acid-alcohol insoluble material and derivatization of the amino acids, 1.56% of the radioactivity per mg (dry wt.) bacteria was found in methionine. The radioactive peak at 8.20 min shown in Fig. 3 coincided exactly with the protonated molecular ion of PTC-methionine (m/z ϭ 285; Fig. 3). Mass ions corresponding to derivatives of the substrate, PTCcystine (m/z ϭ 511, and a fragmentation product m/z ϭ 255) had a retention time of 9.13 min (trace not shown) and a possible product, PTC-cystathionine (m/z ϭ 493) had a retention time of 7.89 min (Fig. 3) respectively. Further, in a control incubation when cysteine was incubated with killed bacteria and hydrolyzed, the radioactivity in the 8.2 min peak corresponded to only 0.17% of the radioactivity per mg (dry wt.) bacteria.
All the strains used in this study incorporated the sulfur from cysteine into methionine. Early experiments when we derivatized the amino acids with DNP gave results that concurred with those obtained when we used PITC-derivatization (see Supplemental Materials).
Enzyme Activities in Cell-free Extracts of Mycobacteria-So far, using intact bacteria, we have shown that the M. tuberculosis complex took up methionine and could incorporate the sulfur into cysteine. Disruption of the Rv1079 gene abrogated the reverse transsulfuration pathway needed for the utilization of methionine but the sulfur atom of cysteine could still be incorporated into methionine. Together with the accumulation of cystathionine in this mutant, our results so far suggested the gene encoded CGL and probably not CGS. Therefore we assayed extracts of M. tuberculosis and M. bovis BCG to determine these, and other key enzyme activities in the interconversion of methionine and cysteine.
The key enzyme needed for conversion of methionine to cysteine, CGL, was detected in all the strains tested except M. tuberculosis H37Rv metB::hyg. Further, this mutant also lacked CGS activity (Table II). Only the CGS activity was predicted to be encoded by the Rv1079 gene that was annotated metB (2). CGS (i.e. MetB) activity was present in all the other strains. Cystathionine ␤-lyase (MetC), the other enzyme assayed for in the cysteine to methionine pathway (Fig. 1), was present in all the strains used. The best substrate for CGS activity was O-succinylhomoserine. When O-acetylhomoserine was included in the assay in place of O-succinylhomoserine, the activity was 20 -30% of what it was with O-succinylhomoserine in either H37Rv or BCG (Table II), whereas homoserine was a poor substrate (Table II). As elsewhere in nature (26,27) this enzyme was highly sensitive to inactivation by propargylglycine. Thus our data showing the methionine auxotroph, M. bovis BCG cysA::Tn5367, to be 100-fold more sensitive than strains prototrophic for sulfur can be explained by inhibition of CGL by propargylglycine and indicates that the M. tuberculosis complex relies on this enzyme, and thus on reverse transsulfuration to convert methionine to cysteine.
The other propargylglycine-sensitive route (see Fig. 1), via methionine ␥-lyase (MdeA), is not possible, because neither was the enzyme activity detected, even when attempted to induced with methionine (Table II), nor was a gene predicted to encode this activity found in the genome (2). Further, the Rv1079 knock-out mutant would be able to compensate for lacking the reverse transsulfuration pathway if it possessed MdeA activity, although its failure to grow on methionine as sole sulfur source ruled out MdeA or any alternative pathway operating to rescue this mutant.
A further enzyme activity, not predicted in the genome analysis (2), was detected at very high specific activity. This was cysteine desulfhydrase, which catalyzes the ␤-elimination reaction shown in Reaction 1.
Usually, this reaction is catalyzed by the same protein that has CGL activity (8,9,29), but this cannot be the case here, because it was not inhibited by propargylglycine and was detected in extracts with no CGL activity (Table II). Given that cysteine is unstable, this activity may be an artifact of nonenzymatic catalysis. However, several lines of evidence indicate that the activity was indeed enzymatic. First, the assays were always performed in the presence of an excess of the thiol-protecting agent, dithioerythritol. Second, heated extracts or tuberculin, a denatured protein product from M. bovis, had no activity. Finally, the activity was saturable, with a K m for L-cysteine or D-cysteine of 1.4 mM. Desulfhydrase activity in the M. tuberculosis H37Rv metB::hyg was compared for three substrates at 1 mM: The specific activities (pmol of product formed/mg of protein/min) were: for L-cysteine, 4012 (Table II); for D-cysteine, 747; for L-cystine, 5111. In trace c, cystathionine (7.9 min), which was detected but not labeled, is shown. This peak accumulated in experiments when CGL was abolished (see text).
We could find no significant difference in specific activities of any of the enzymes (i) in the same strain grown with different additions to the culture media, or (ii) between parent strains and mutants except for the enzyme activities of CGL and CGS, which were completely lacking in M. tuberculosis H37Rv metB::hyg.
Enzyme Activity in Cell-free Extracts of E. coli ⌬metB(DE3)/ pLK101-The results obtained with the M. tuberculosis metB::hyg mutant introduced an ambiguity about the role of the gene annotated as metB, Rv1079 (Table II). To resolve this, we cloned Rv1079 into an E. coli ⌬metB mutant to determine whether it would confer both CGL and CGS activities, or whether the lack of either activity in the M. tuberculosis metB::hyg mutant was caused by a polar or a regulatory effect. The Rv1079 gene complemented the metB mutation in E. coli, allowing it to grow in minimal medium with sulfate as the sole sulfur source (see Supplemental Figure).
The strain of E. coli we chose gave very clean backgrounds in our assays as it is a metB mutant, and E. coli naturally lacks CGL activity. This was borne out in our assays, because we could not detect either enzyme activity in cell-free extracts of control clones in which an empty vector was introduced. Moreover, very little pyruvate, no more than 0.2 nmol/min/mg protein, was ever produced from cystathionine, indicating that the native E. coli ␤-lyase activity was low or absent in these extracts made from LB-grown bacteria in which the activity should be repressed. The Rv1079 gene clearly conferred both CGL and CGS (Table III) activities, but no cysteine desulfhydrase or cystathionine ␤-lyase activity. The properties of the enzyme activities conferred by the cloned gene were similar to those attributed to the native enzyme in the extracts of mycobacteria (Table II). These properties were the ratio of CGS activity with 10 mM O-succinylhomoserine to CGL activity with 2 mM L-cystathionine (Table III), the ratio of CGL activity with 2 mM L-cystathionine to 0.2 mM L-cystathionine with the higher concentration appearing almost saturating (Table III). Further, for the synthase activity, relative rates of the elimination reaction (i.e. to give keto acid (Fig. 1, red arrow)) with 10 mM O-succinylhomoserine and the replacement reaction (i.e. to give cystathionine) when 1 mM cysteine was included were similar. With either source of enzyme, only the CGS reaction occurred at 1 mM cysteine with CGL being completely inhibited at this concentration. Finally, the synthase activities with 10 mM Oacetylhomoserine was 16 -22%, and with 10 mM L-homoserine, 3.6% the activity with O-succinylhomoserine (Table IV).
Next, we wanted to address how the Rv1079 gene product is used in sulfur metabolism in the M. tuberculosis complex and how it could be regulated. Since the extracts of E. coli with the cloned Rv1079 gave its activity in isolation from the other enzymes involved in sulfur amino acid interconversion, we used this cell-free extract to investigate the effects of adding substrates and other metabolites at a range of concentrations. These can be related to the concentrations of the sulfur-containing metabolites inside tubercle bacilli that we determined during this study by estimating their concentrations (shown as a footnote to Table III) using an intracellular volume of 6 ϫ 10 Ϫ16 liter (30,31). We chose 1 mM as an arbitrary concentration of O-succinylhomoserine lower than that used in standard assays (25), because below this, enzyme activity was barely detectable (results using 0.25 mM O-succinylhomoserine not shown). Although CGL activity was inhibited by increasing concentrations of L-cysteine, even at 50 M cysteine (way in excess of any intracellular concentration) 60% of CGL activity remained (Table III). In contrast, L-methionine had no effect on CGL even at 0.8 mM (Table III). The elimination reaction with O-succinylhomoserine to give 2-ketobutyrate was affected similarly by L-cysteine, which in turn gave rise to the production of cystathionine through the replacement reaction catalyzed by CGS activity using cysteine as a substrate. The replacement reaction was not affected by the addition of either 0.5 mM L-cystathionine or 0.8 mM L-methionine (Table III). While the concentrations of cystathionine and O-succinylhomoserine appear to approach saturation in the reactions shown in Table III, that of L-cysteine does not. This may be a result of futile cycling of cysteine when it is at low concentrations and exerts little inhibitory effect on the CGL activity. A consequence of the operation of that futile cycle would be the net loss of some of the cystathionine formed, and thus an underestimate of its rate of  ( (2) a NA, not applicable as cystathionine could not be measured as product when it was the only substrate for the CGL/CGS enzyme. b Estimated intracellular concentrations of sulfur amino acids in tubercle bacilli: Cystathionine could reach 0.5 to 0.7 mM, and methionine 0.8 mM, in bacteria grown in 7H9 medium including 60 M methionine. In bacteria grown in 7H9 alone, both amino acids were at 20 -40 M. The cysteine concentration, deduced from data represented in Fig. 2, was always less than 20 M (from the limit of detection in the less sensitive, thiol determination). Using the data from Fig. 3, and a more sensitive detection method, the limit of detection would be even lower but we have presented the more conservative value as we could not show any major effects at 20 M in the experiments reported in this table.
c Total cystathionine was determined after stopping this assay, but controls allowed cystathionine added to be subtracted from the total giving the value reported here minus cystathionine formed. d ND, not determined because its appearance would be ambiguous, i.e. from both cystathionine or O-succinylhomoserine.
formation. Putting these results together, the Rv1079-encoded product must act solely as a CGL in the tubercle bacilli at the concentrations of sulfur-containing amino acids we observed in this study, with the issue of futile cycling of cysteine not arising at the very low concentrations that occurred. Consistent with this view was the lack of effect of possible inhibitors at the end-point of the reaction and the sulfhydration pathway, cystathionine and methionine, respectively.

DISCUSSION
Mutants of the M. tuberculosis complex unable to take up (3,4) or reduce (32) sulfate are auxotrophic for methionine. The ability of different auxotrophs to grow in the host has been used, indirectly, to deduce the nutrients available to them in the cellular compartments in which they reside, for example for Salmonella (33,34) and for Brucella suis (35). Given that methionine auxotrophs of M. bovis BCG, lacking functional sulfate transport, are not compromised for survival in experimentally infected mice (4), it appears that methionine may be their key source of sulfur in the host. To use methionine, the bacteria must be able to take it up and metabolize it to cysteine in order to fulfill all its sulfur-containing metabolites (Fig. 1). In the work described in this report, we have shown all the key steps in methionine utilization by the M. tuberculosis complex.
Methionine was transported with a V max 11ϫ higher that the V max for sulfate. In the host, sulfate is likely to be available at saturating concentration (4) but methionine is likely to be at concentrations around 20 -27 M (36), about one-third the apparent K m for its uptake. Nevertheless, even at this concentration methionine would be taken up about 3ϫ faster than sulfate. In the presence of methionine, weak (about 2-fold) repression of sulfate transport occurred (4) so tubercle bacteria would take up methionine 5-6ϫ faster than sulfate at the estimated physiological concentration. Although the genes that encode the methionine transporter were not identified (2), there are candidates among the many predicted ABC transporters for amino acid, including methionine, transport (37)(38)(39).
Next, we showed that methionine was converted to cysteine in intact bacteria and demonstrated, in cell extracts, the key enzyme for this process in the reverse transsulfuration pathway (Fig. 1). Because intact M. tuberculosis bacteria did not accumulate cysteine in their metabolic pools, the conversion to cysteine had to be shown by tracing radioactive sulfur from methionine into mycothiol, its most abundant thiol (18). As the sulfur of mycothiol is obtained only from cysteine in a pathway starting with ligation of cysteine to a pseudodisaccharide (glucosaminylinositol) (40), our demonstration of labeling of mycothiol is unambiguous evidence that M. tuberculosis is able to convert methionine to cysteine while conserving the sulfur atom.
The mycothiol labeling experiments alone did not provide sufficient evidence for reverse transsulfuration, as the sulfur atom is known to be recycled in some organisms (Fig. 1). However, we showed the sole key enzyme to explain the labeling experiments and methionine auxotrophy to be CGL as growth experiments, accumulation of the intermediate cystathionine, and enzyme assays were all concordant. We ruled out the alternatives (Fig. 1) of a methionine ␥-lyase (MdeA) and the pathway via 3-methylthioproprionate by experiments using the ␥-lyase inhibitor, propargylglycine and performing enzyme assays initially. Then, when we discovered the M. tuberculosis H37Rv metB::hyg mutant lacked CGL, we found that the inability of this mutant to grow on methionine as sole sulfur source provided further support for our conclusion. A gene in the 3-methylthioproprionate pathway, mtn (Fig. 1), has been annotated (genolist.pasteur.fr/TubercuList/) but it is in a branch that can be used in anabolism, and its likely product is the methylthiopentose that is part of lipoarabinomannan (41).
Both mutagenesis and cloning show the Rv1079 gene encodes a bifunctional CGL/CGS. It also catalyzed an elimination reaction with homoserine and its derivatives, but, unusually, had no elimination activity with cysteine. This is the first demonstration of bifunctional CGL/CGS activity in a single protein in an organism possessing both transsulfuration and reverse transsulfuration (reactions unique to these pathways entirely within the yellow box in Fig. 1) pathways. It is likely that Saccharomyces cerevisiae has a bifunctional CGL/CGS. Though its CGS activity was only assayed using the elimination reaction with O-acetylhomoserine (so cystathionine formation was never directly established) (42), the structures of the binding sites of this yeast enzyme and E. coli CGS are superimposable, suggesting it also may possess CGS activity (6). Streptomyces phaeochromogenes has a single protein shown both to produce cystathionine by a replacement reaction with homoserine and cysteine (43) and to have CGL activity (8). Yet neither organism has a cystathionine ␤-lyase so their bifunctional CGL/CGS can only be used in the methionine to cysteine direction. Where both pathways co-exist in the same organism and the enzymology was investigated, CGL and CGS activities were conferred by separate proteins, for example in Klebsiella aerogenes in which a metBϪ/metCϪ double mutant retains CGL activity (28).
In general, enzymes of the Cys-Met metabolism pyridoxal-5Ј-phosphate-dependent protein family possess a range of ␤-elimination, ␥-elimination, and ␥-replacement reactions (6) exemplified by cystathionine ␤-lyase, CGL, and CGS (cystathionine-forming), respectively. During evolution, these activities have been separated into proteins each catalyzing a subset of these reactions (5,44), presumably under selective pressure that has driven precise control of enzyme functions and avoiding wasteful futile cycling of intermediates. In the M. tuberculosis complex, cystathionine ␤-lyase and CGL/CGS are encoded by two different genes, metC and Rv1079, but given the complete pathways in which CGL and CGS participate how can these latter two activities be controlled? We postulate that it is by controlling the intracellular pool of cysteine. With its reactive thiol group, cysteine can be highly toxic, so its intracellular concentration is tightly regulated in nature (45,46). We suggest that another enzyme that we discovered in the tubercle bacilli during this study, a cysteine desulfhydrase, has this role. Although CGLs often have this activity of releasing pyruvate from cysteine (8,9,29), in this case its persistence in the M. tuberculosis metB::hyg mutant and insensitivity to propargylglycine show it is encoded by a different, but as yet unknown, gene to Rv1079. Its high activity (Table II) and preference for L-cysteine over D-cysteine are consistent with a role in detoxification, and help explain the absence of any detectable pool cysteine. Thus, in this essentially cysteine-free environment, the bifunctional CGL/CGS enzyme only ever acts as a CGL.
Even in the presence of cysteine, CGS activity is not needed to incorporate the sulfur atom of cysteine into methionine as the metB::hyg mutant, lacking CGS, still carries out this activity. This can be rationalized by the action of the cysteine sulfhydrase, which releases sulfide. The sulfide can then be recycled by O-succinylhomoserine sulfhydrylase (MetZ) to give the 4-carbon sulfur amino acid homocysteine directly, thus by-passing the effect of disruption to metB (Fig. 1). Indeed, MetY, an O-acetylhomoserine sulfhydrylase that carries out the same function as MetZ in the biosynthesis of homocysteine, has been shown functionally to enable a metB-deficient mutant of Corynebacterium glutamicum to grow on sulfate as the sole sulfur source (47).
Although we show a bifunctional CGL/CGS enzyme for the first time in an organism possessing both transsulfuration and reverse transsulfuration pathways, it seems likely that such an enzyme occurs in other actinomycetes. The orthologues of the gene that encodes this enzyme are extremely closely related in C. glutamicum (5), Streptomyces coelicolor (63% identical over the whole protein) and Mycobacterium leprae (84% identical over the whole protein), though the full range of activities that they encode has not been ascertained.
Overall, this work shows that the existence of the reverse transsulfuration allows the M. tuberculosis complex to utilize methionine as a sole sulfur source. The Rv1079 gene, formerly annotated metB, encodes the key enzyme and should be reannotated to reflect the discovery that it encodes a bifunctional CGL/CGS. Cysteine is an important intermediate in the production of many sulfur compounds but it is strictly an intermediate that does not accumulate, and we suggest a mechanism, a powerful desulfhydrase activity, for decomposition of any excess. In the host, methionine may be a more important sulfur source than sulfate for growth of the M. tuberculosis complex, since a cysA mutant unable to transport sulfate survived as well as its parent strain (4) while a Rv1079 mutant that could not convert methionine to cysteine was somewhat attenuated (48). This preference for methionine over sulfate in the host also explains how M. leprae remains a pathogen, despite being a natural methionine auxotroph, because of its loss (see Fig. 1) of cysTWA for sulfate transport and metC for cystathionine ␤-lyase (49). 2