Rv0132c of Mycobacterium tuberculosis Encodes a Coenzyme F420-Dependent Hydroxymycolic Acid Dehydrogenase

The ability of Mycobacterium tuberculosis to manipulate and evade human immune system is in part due to its extraordinarily complex cell wall. One of the key components of this cell wall is a family of lipids called mycolic acids. Oxygenation of mycolic acids generating methoxy- and ketomycolic acids enhances the pathogenic attributes of M. tuberculosis. Thus, the respective enzymes are of interest in the research on mycobacteria. The generation of methoxy- and ketomycolic acids proceeds through intermediary formation of hydroxymycolic acids. While the methyl transferase that generates methoxymycolic acids from hydroxymycolic acids is known, hydroxymycolic acids dehydrogenase that oxidizes hydroxymycolic acids to ketomycolic acids has been elusive. We found that hydroxymycolic acid dehydrogenase is encoded by the rv0132c gene and the enzyme utilizes F420, a deazaflavin coenzyme, as electron carrier, and accordingly we called it F420-dependent hydroxymycolic acid dehydrogenase. This is the first report on the involvement of F420 in the synthesis of a mycobacterial cell envelope. Also, F420-dependent hydroxymycolic acid dehydrogenase was inhibited by PA-824, and therefore, it is a previously unknown target for this new tuberculosis drug.


Introduction
The cell wall of Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis [1,2], has an extraordinarily complex and very hydrophobic structure. Consequently it offers an exceptionally low permeability and makes the Mtb cells poorly accessible to drugs and less vulnerable to attack by the host immune system [3]. For this reason, cell wall synthesis enzymes of Mtb have been targeted for TB drug development [4]. Mycolic acids (MAs) are some of the key lipid components of the mycobacterial call wall. These ''high-molecular weight beta-hydroxy fatty acids with a long alpha-alkyl side chain'' [5] (Fig. S1) are constituents of mycolylarabinogalactan-peptidoglycan complex and trehalose mono-/dimycolates (TMM and TDM) [6][7][8]. By helping to build a strong cell wall and being immunogenic [7,9,10], these complexes contribute to the development of TB [3,[10][11][12][13][14][15]. Mtb generates three structural types of MAs which are called a-, methoxy-and keto-mycolic acids (a-, M-and K-MAs) and under in vitro growth conditions it does not contain epoxymycolic acids (E-MAs) that are found in Mycobacterium smegmatis [16]; the respective chemical structures are shown in the Supporting Material (Fig. S1). The keto-and methoxy-derivatives enhance the pathogenic nature of Mtb [17,18], and the bacterium uses these compounds to modulate the host immune response [9,[19][20][21]. A recent report shows that K-MAs allow Mtb to form pellicle structures, which in turn make this pathogen drug-resistant [22]. Thus, the enzymes that introduce keto-and methoxy-groups in mycolic acids are of research interest [3,17,[23][24][25][26]. These oxygenated lipids are generated through common immediate precursors, hydroxymycolic acids (H-MAs) ( Fig. 1) [3,24,27]. Whereas it is known that in Mtb the conversion of H-MAs to M-MAs is catalyzed by an adenosylmethionine-dependent methyltransferase (Mma3 or CmaB) encoded by the ORF Rv0643c [7,24,26] (Fig. 1), the enzyme that oxidizes H-MAs to K-MAs remains unknown. We call this unknown enzyme hydroxymycolic acid dehydrogenase (HMAD). In this report we describe the gene that encodes HMAD in Mtb and demonstrate that the enzyme utilizes coenzyme F 420 , a deazaflavin derivative, as electron carrier (Fig. 1). Thus, we named the enzyme fHMAD for F 420 -dependent Hydroxy Mycolic Acid Dehydrogenase. Also, we show that fHMAD is inhibited by PA-824, a nitroimidazopyran and a new TB drug that is currently on clinical trial [28].

Results and Discussion
Identification of Rv0132c as Coenzyme F 420 -dependent Hydroxymycolic Acids Dehydrogenase (fHMAD) in M. tuberculosis This work began with an analysis of the available data, and the resulting hypothesis was tested via genetic analysis of an Mtb gene in Mycobacterium smegmatis. The rationale for the selection of M. smegmatis as the experimental host has been elaborated below.
Selection of Mycobacterium smegmatis as a facile screening host in a search for the HMAD encoding gene of Mycobacterium tuberculosis. As mentioned above, Mtb produces a-, K-and M-MAs, and it does not contain epoxymycolic acids (E-MAs) under in vitro growth conditions [16]. In this regard Mycobacterium bovis strain BCG (BCG) is similar to Mtb except some of the strains of the former do not produce M-MAs as the cmaB or mma3 gene of the organism is non-functional due to a point mutation [16,25,26]. M. smegmatis produces a-, a9-, and E-MAs but is devoid of K-and M-MAs [16,29]. The structures of these species are shown in Fig. S1; in M. smegmatis five variations of the a group, a1-, a2-, a3-, a4and a5, are found [29]. The investigation described in this report concerns only the longer aliphatic chains (the R groups) of the MAs (Fig. S1 and Fig. 1). Fig. 1 shows the proposed pathways for synthesis of H-, K-, Mand E-MAs in wild-type and recombiant Mtb, BCG and M. smegmatis [17,23,24,27]; a common precursor for the aliphatic chains of various MAs serves as the starting point in this scheme. The deletion of the hma gene (also called mma4 and cmaA) in Mtb abolishes the production of K-and M-MAs and causes the production of E-MAs and an intermediate that is similar to a-MAs of M. smegmatis [17,23,27]. Heterologous expression of the Mtb or BCG hma gene (orf rv0642 or mb0661, respectively) in M. smegmatis allows the synthesis of H-MAs and reduces the production of aand E-MAs in the recombinant strain [23,24,27]. Therefore, in M. tuberculosis the hma gene encodes the enzyme that generates H-MAs as precursors for both keto and methoxy forms, and this process competes well with the E-MA formation. The accumulation of H-MAs in a M. smegmatis strain carrying heterologous hma shows that the organism lacks both Mma3 (or CmaB) and HMAD and therefore cannot transform this intermediate into M-MAs and K-MAs [23,24,27]. Accordingly, a recombinant M. smegmatis strain carrying Mtb hma could be used to screen candidate Mtb genes for HMAD activity via complementation. This is advantageous, as unlike Mtb, M. smegmatis is not pathogenic and it grows much faster than Mtb or BCG [30].
Identification of rv0132c as a candidate gene encoding HMAD. We searched for this gene in the Mtb H37Rv genome [31] by using the following criteria. It must be present in both Mtb and BCG while absent in M. smegmatis. It should encode a dehydrogenase capable of catalyzing a two-electron transfer process, as the conversion of H-MAs to K-MAs involves the oxidation of a secondary alcohol group to a keto group. This dehydrogenase must also possess the structural elements for interaction with a hydrophobic substrate such as a mycolic acid. One of the Mtb ORFs that matched these characteristics was Rv0132c and it has been known as Fgd2 [31,32]. It is a structural Figure 1. Proposed pathways for the synthesis of hydroxy-, keto-, methoxy-and epoxymycolic acids in mycobacteria [7,24]. A common intermediate for various R groups is used as the starting point. Where MmaA2 and CmaA2 are involved in the formation of cis cyclopropane group, CmaA2 and an yet to identified enzyme (indicated by?) catalyze trans-cyclopropanation [48,67]. The details of the individual R groups are shown in Fig. S1. * indicates that it is not known whether the cyclopropanation step follows or precedes oxygenation. All protons (except for the isolated groups) that have been target for NMR data analysis have been shown in red. The OH group shown in italics and underlined in the box at the left corner of the figure was converted to a methoxy group during saponification of mycolic acids; the process generated mycolic acids methyl esters (MAMEs). doi:10.1371/journal.pone.0081985.g001 F420-Dependent Hydroxymycolic Acid Dehydrogenase PLOS ONE | www.plosone.org homolog of coenzyme F 420 -dependent glucose-6-phosphate (G6P) dehydrogenase (Fgd or Fgd1) that catalyzes two-electron oxidation of G6P [glucose-6-phosphate +F 420 R 6-phosphogluconolactone + reduced F 420 (F 420 H 2 )] [33,34]. Coenzyme F 420 is a deazaflavin derivative that is found in all mycobacteria [35,36]. At the ground state it functions similar to nicotinamide coenzymes or NAD(P), mediating hydride transfer reactions [37]. Fgd2 does not oxidize G6P and its substrate remains unknown [32,38]. M. smegmatis expresses Fgd1 and it lacks Fgd2, whereas both Mtb and BCG carry Fgd1 and Fgd2 ( [31,32]; NCBI Accession Number: NC_008596). Both Fgd1 and Fgd2 are also homologs of F 420dependent methylenetetrahydromethanopterin reductases (Mer) that are found in methanogenic archaea [34]. To obtain some clues to the nature of the substrate that Fgd2 or Rv0132c acts on, we analyzed the primary structure of this protein based on X-ray crystallographic structures of three well characterized Mer homologs: Fgd1 of Mtb (PDB ID: 3B4Y) [32], F 420 -dependent methylenetetrahydromethanopterin reductase from Methanopyrus kandleri (MkMer; PDB ID: 1EZW) [39], and an F 420 -dependent secondary alcohol dehydrogenase (Adf; PDB ID: 1RHC) from Methanoculleus thermophilicus [40]; M. kandleri and M. thermophilicus are methanogenic archaea. In Fgd1, His 40 , Ser 73 , Val 74 and Glu 109 help to bind F 420 and these residues, except Ser, are functionally conserved in MkMer and Adf [39,40] as well as in Rv0132c (Fig. 2). Ser 73 of Fgd 1 interacts with F 420 via the oxygen atom of the backbone carbonyl [32], and in Rv0132c and MkMer this residue has been substituted with Gly and in Adf the equivalent position is occupied by a Cys (Fig. 2). Ser, Cys and Gly are highly compatible in terms of their hydrophobicities and sizes [41][42][43]. Hydrophobe compatibility indices for Ser-Cys and Ser-Gly pairs in a scale 1-20 (1 and 20 being least and fully compatible, respectively) are 17.7 and 16.8, respectively [41]. The isoelectric points of Ser, Cys and Gly are 5.7, 6.0 and 5.1, respectively [41]. The volumes of Ser, Cys and Gly are 73, 86 and 48 cubic Angstroms, respectively, which are considered similar [44]; the amino acid volumes range from 48 cubic Angstroms for Gly to 163 cubic Angstroms for Trp. Consequently, the above-mentioned substitutions at Ser 73 position will not appreciably change the ability of an Fgd1 homolog to bind F 420 . Therefore, in our investigation, we had considered Rv0132c as a potentially F 420dependent enzyme. A recent report shows that Rv0132c indeed binds F 420 [38].
In our next analysis we tried to understand whether Rv0132c has the potential to transform hydrophobic substrates such as mycolic acids. Adf and MkMer interact with hydrophobic substrates whereas Fgd1 accommodates charged glucose-6-phosphate. In Adf the Val 193 and Leu 227 , which are hydrophobic, not only interact with the hydroxybenzyl unit of F 420 but also help to position the hydrocarbon chain of the substrate [40]. Similarly, Ala 197 and Tyr 229 in MkMer interact with both the F 420 and the hydrophobic pterin ring of tetrahydromethanopterin [40]. Rv0132c shows partial conservation of these characteristics, as homologous residues in this protein are Ala and Glu, respectively (Fig. 2). In contrast, the equivalent positions in Fgd1 are occupied by Ser and Glu [32], which are less hydrophobic and polar, respectively. In Fgd1, Lys 232 , which has a charged side chain, helps to bind a citrate ion, which is a competitive inhibitor of the enzyme [32]. In Rv0132c, Adf and MkMer, this residue has been replaced with Phe, Trp and Cys, respectively (Fig. 2). Additionally, Fgd1 utilizes a positively charged pocket formed by Lys 198 , Lys 259 , and Arg 283 to hold the phosphate group of glucose-6-phosphate [32] and these residues are not conserved in Rv0132c (Fig. 2). Hence, Rv0132c has the potential of interacting with a hydrophobic substrate.
We had observed that two tandem Arg residues in the NH 2terminus (amino acid residues 1-27, MTGISRRTFGLAAGF-GAIGAGGLGGGC; bold and underlined, characteristic residues) form a signature for translocation into the periplasmic space via a Tat-dependent protein export pathway which exists in the mycobacteria [45] and the features shown underlined (see above) represent a putative prokaryotic membrane lipoprotein lipid attachment site (PS00013) where Cys 27 could carry lipophilic substrates [46]. Indeed, as our work was complete, Rv0132c was found to be exported to the cell envelope of Mtb [38]. Thus, it is reasonable to assume that Rv0132c could interact with the hydrocarbon chains of the mycolic acids (R group, Fig. 1). The relevance of the demonstrated cellular location of Rv0132c to our findings has been discussed below.
The initial identification of the individual mycolic acid bands on the TLC plates was performed via comparison with previously reported patterns [18,23,24]. Then we carried out mass spectrometric and NMR spectroscopic analysis with materials recovered from the relevant TLC bands. For the H-MAs and K-MAs bands, MALDI-TOF mass spectrometry yielded spectra that were characteristics of respective myoclic acids with 77-82 carbon atoms (Fig. 4) [23]. The mass for every characteristic H-MA ion (Fig. 4A) was 2 units higher than that for a K-MA (Fig. 4B) and this is consistent with the respective structures shown in Fig. 1. 1 H NMR data provided more detailed characterization of relevant mycolic acid species and we discuss the findings below with a focus on the H-atoms marked in red in Fig. 1. This analysis is based on previously reported NMR data on mycolic acids [23,27,29,48,49]. The resonances at 2.7 ppm observed with the E-MA preparation obtained from M. smegmatis mc 2 155 cells (Fig. 5A) were characteristics of the methine protons associated with a trans-epoxide group [23]. In the spectrum obtained with the H-MAs preparation from M. smegmatis (pEP-hma) strain (Fig. 5B) the resonances for the above-mentioned epoxy group were not seen and instead it exhibited a resonance at 3.5 ppm representing the methine proton on the carbon that carried the characteristic hydroxyl group of H-MA. Similarly, the 1 H resonances of the methylenic and methine groups that flank the carbonyl group in K-MAs were found at 2.31-2.39 ppm in the spectra for the K-MA preparation from M. smegmatis (pEP-rv0132c/hma) strain (Fig. 5C). The spectra for the E-MA, H-MA and K-MA preparations exhibited the following common resonances and this observation is consistent with previous reports [23,27,29,48,49] (Fig. 5A-C): 1.29 ppm -broad, isolated methylene proton; 0.85 ppm -triplet, terminal methyl groups; 3.71 ppm -singlet, methyl ester; 2.50 ppm -multiplet, methine at postion C-2 with respect to the terminal carboxyl group (see within the box at the left corner of Fig. 1). None of the above spectra showed the resonances of the protons that are associated with the cyclopropane groups of mycolic acids produced by Mtb; these resonances appear at 20.40, 0.50, and 0.58 ppm for ciscyclopropanation and 0.01-0.16 ppm for trans-cyclopropanation [49]. Major mycolic acids produced by M. smegmatis lack cyclopropanation under normal growth conditions [23,50]. This modification occurs during growth at 25uC [51] and the growth temperature in our study was 37uC.
The above-described analysis showed that the heterologous expression of Mtb hma in M. smegmatis caused the suppression of the synthesis of E-MAs and the production of H-MAs, and Rv0132c protein converted the H-MAs to K-MAs. Thus, in Mtb Rv0132c encoded a hydroxymycolic acid dehydrogenase (HMAD).
The next step was to determine if HMAD was coenzyme F 420dependent. The fbiC is a key gene for the production of F 420 chromophore in mycobacteria [52] and mycobacterial strains lacking a functional fbiC gene are devoid of this coenzyme [52,53]. We found that a M. smegmatis DfbiC strain [53] generated H-MAs but not K-MAs when complemented with pEP-rv0132c/hma (Fig. 3B, right most lane or the lane rv0132c/hma for DfbiC). Complementation with pEP-hma also provided H-MAs in M. smegmatis DfbiC (Fig. 3B, lane hma) and pEP-rv0132c did not provide either K-MAs or H-MAs (Fig. 3B, lanes rv0132c); the left most lane (lane rv0132c/hma for wild-type (wt) strain) served as a positive control, where production of K-MAs was observed. Hence, HMAD required F 420 for activity and we call it hereafter fHMAD.
In this context we address two sets of contradicting reports in the literature that concern the biosynthesis of H-MAs and K-MAs in M. smegmatis strains carrying clones for the Mtb hma gene. In one case the hma gene caused the synthesis of both H-MAs and K-MAs [23,24] and in the latter only H-MAs were found in the recombinant [18]. Our result is consistent with the latter [18], as the conversion of a hydroxyl group to a keto group would be catalyzed by an electron transfer enzyme or dehydrogenase such as fHMAD, and not by a methylase/hydrase activity such as seen in Hma.   1 shows two mycobacterial MA oxygenation pathways, one of which leads to H-MAs, K-MAs and M-MAs, and the other is for the production of E-MAs. It has been shown that when the former operates, the latter is suppressed [17,23,24,27]. We observed a more stringent form of this regulation in our studies. The data in Fig. 3A show that when M. smegmatis was made capable of producing K-MAs, it did not produce E-MAs; a comparison of rv0132c/hma lane with any other lane in Fig. 3A leads to this conclusion. This effect was not due to the Rv0132c protein or the DNA elements cloned into pEP-rv0132c/hma, as their presence did not suppress E-MA production when the host lacked fbiC (Fig. 3B, lanes hma, rv0132c and rv0132c/hma). Hence it could be hypothesized that K-MAs either inhibit one or more E-MA synthesis enzymes and/or suppresses the expression  (Fig. 3A). Only a part of each spectrum is shown and the annotations for the ion masses are based on reference [23]: labels H & K, ions from hydroxy-and ketomycolic acids preparations; numbers 77-82: total number of carbon atoms in free acids. The unlabeled peaks belong to unidentified species that were present in both preparations. doi:10.1371/journal.pone.0081985.g004 F420-Dependent Hydroxymycolic Acid Dehydrogenase PLOS ONE | www.plosone.org of respective genes. Other possibilities are the interference with the translocation of the precursor of E-MAs to the modification site such as periplasm or a flux-based competition between the two pathways. We also observed that the cellular level of H-MAs increased when K-MAs were produced (Fig. 3A, lane pEP-rv0132c/hma). It is possible that K-MAs enhanced the activities of one or more enzymes that generate H-MAs from a-mycolic acids (Fig. 1) and/or increased the expressions of their genes, and the prevailing fHMAD activity was not at par with the rate of H-MA production. The other explanation is that in M. smegmatis (pEP-rv0132c/hma) the cellular level of Hma activity was much higher than that of fHMAD. In this context we note that the overproduction of M-MAs through over-expression of Mma3 or CmaB suppresses K-MAs in Mtb [18,22]. A more detailed study is needed to elucidate the mechanisms underlying these competitions between mycolic acids oxygenation pathways.

Inhibition of fHMAD by PA-824
PA-824, a new TB drug, inhibits the formation of K-MAs and causes an accumulation of H-MAs in Mtb [54]. We tested whether this effect is specifically due to the inhibition of fHMAD. As shown in Fig. 3C, in the presence of PA-824, M. smegmatis (pEP-rv0132c/ hma) accumulated a high level of H-MAs and contained a reduced level of K-MAs (Fig. 3C). Hence, PA-824 inhibited the heterologously expressed fHMAD. To establish further that fHAMD was inhibited by PA-824, we determined the relative levels of K-MAs in M. smegmatis (pEP-rv0132c/hma) cultivated in the presence of this drug at various concentrations. The results showed that the inhibition began at a PA-824 concentration between 10-25 microgram per ml culture and increased further as the drug concentration was raised (Fig. S2). Such a dose-dependent increase in the inhibition of K-MA synthesis by PA-824 has been reported also for M. tuberculosis [54]. However, the K-MA synthesis process in M. smegmatis (pEP-rv0132c/hma) was much less sensitive to PA-824 than that observed in wild-type M. tuberculosis; in M. tuberculosis this inhibition begins at a PA-824 concentration between 30-60 nanogram per ml culture [54]. It is possible that the higher minimum inhibitory concentration of PA-824 observed with M. smegmatis (pEP-rv0132c/hma) was due to the presence of a higher level of Rv0132c protein in this recombinant strain; rv0132c was expressed from a multi-copy plasmid [47] and was driven by both the native promoter as well as the strong and constitutive mycobacterial hsp60 promoter [47,55]. Another explanation is that compared to M. tuberculosis, M. smegmatis takes up PA-824 poorly and as a result for achieving an inhibitory concentration of the drug inside the cell, it had to be supplied in the culture medium at a higher concentration; wild-type M. smegmatis is naturally resistant to PA-824 [56]. Nevertheless, the results presented in Figs. 3C and S2 show that the phenomenon of inhibition of K-MAs synthesis by PA-824 that was observed in wild-type M. tuberculosis could be reproduced in a M. smegmatis strain carrying cloned hma and rv0132c genes from the former.
PA-824 kills Mtb under both aerobic and anaerobic conditions [54,57,58]. The anaerobic killing occurs through the reduction of PA-824 by an F 420 H 2 -dependent nitroreductase called Ddn, which is followed by the production of toxic NO [57]. Since F 420 H 2 is produced by Fgd1, Mtb and BCG strains lacking Fgd1 activity are resistant to PA-824 [54]. The aerobic killing of Mtb by PA-824 has been thought to occur due to the elimination of K-MAs via unknown mechanisms [54,58]. Our data has now linked this concept to a gene, rv0132c. Curiously, F 420 is an integral part of both mycobacterial systems, Ddn and fHAMD, that interact with PA-824.
As mentioned above, fHMAD found to be exported to the cell envelope of Mtb [38]. Also, it is thought that complete mycolic acids are transported to the plasma membrane as trehalose monomycolates or TMM [59]. In combination these observations suggest that in Mtb at least one additional modification, formation

Conclusion
The hydroxymycolic acid dehydrogenase of Mtb was shown to be an F 420 -dependent enzyme encoded by the ORF Rv0132c and it is inhibited by PA-824, a new TB drug.
Our data suggest that there is only one bona fide Fgd in the mycobacteria. Citing the lack of glucose-6-phosphate dehydrogenase activity in Rv0132c, it has been recently suggested that this protein should no longer be called Fgd2, and Fgd1 should be called simply Fgd [38]. Our data supports this proposal and provides a functional name for Rv0132c, F 420 -dependent hydroxymycolic acid dehydrogenase (fHMAD). Coenzyme F 420 is universally present and essential in the strictly anaerobic methanogenic archaea [37]. In the bacterial domain, a similarly wide distribution of this deazaflavin derivative is seen in the Actinobacteria phylum, which includes the mycobacteria [35,36]. Every mycobacteria examined thus far contains F 420 [35,36]. As mentioned above, in the hydride transfer function F 420 mimics NAD(P). The mid-point electrode potential of the F 420 /F 420 H 2 couple 2360 mV, which is 40 mV lower than that of the nicotinamides [37]. Perhaps in the mycobacteria F 420 participates in a set of hydride transfer reactions that cannot be accomplished at all or efficiently by the nicotinamides due to thermodynamic reasons, such as a need to operate at a lower redox potential. Such a specialized role has now been seen in the neutralization of nitrosative stress [53] (via a chemical reaction with Fgd-derived F 420 H 2 ) and in the introduction of a key functionality to the complex mycobacterial cell envelope (the fHMAD reaction as demonstrated here). Both of these actions bring resilience to the mycobacteria against environmental stresses such as those imposed by the human immune system. It is noteworthy, that the current report presents the first example for the involvement of F 420 in the biosynthesis of mycobacterial cell wall. The nitroreductase (Ddn) that helps to activate PA-824 with F 420 H 2 and the F 420 -dependent enzymes that allow the mycobacteria to decolorize triphenylmethane dyes or to degrade aflatoxins [57,[60][61][62] could also fulfill yet to be described key and normal physiologically relevant cellular functions in these organisms.

Oligonucleotides, Plasmid, DNA, Bacteria and Growth Conditions
Oligonucleotides, plasmids and bacteria used in this study have been described in the Supporting Material (Table S1). M. tuberculosis H37Rv chromosomal DNA was obtained from the National Institutes of Health's TB Vaccine Testing and Research Materials Contract (TBVTRMC) at the Colorado State University. E. coli was grown in Luria-Bertani broth or solid media. Mycobacterium smegmatis strains were grown in Middlebrook 7H9 broth or on agar solidified medium with 0.2% glycerol as the carbon and energy source [63]. For liquid cultures Tween 80 at the concentration of 0.05% was also added. When required, M. smegmatis strains bearing antibiotic resistance genes were selected on or grown with kanamycin and hygromycin at the concentration of 20 and 150 microgram/ml, respectively, and for similar work with E. coli strains ampicillin, kanamycin, and hygromycin concentrations were 100, 20, and 150 microgram/ml, respectively. To study the effect of PA-824 on the mycolic acids content of M. smegmatis (pEP-rv0132c/hma), a freshly inoculated culture was grown overnight to an optical density of 0.3 at 600 nm (as measured by use of a DU800 UV/Vis Spectrophotometer, Beckman Coulter, Brea, CA). It was then supplemented with PA-824 to a desired final concentration from a stock solution (80 mg/ml) in DMSO and grown for additional 36 h. The control culture received DMSO at a concentration of 1.25 ml/liter. PA-824 was a gift from Global Alliance for TB Drug Development (New York, NY) through the Global Health program of the RTI International (Research Triangle Park, NC).

Molecular Biology Techniques
M. smegmatis chromosomal DNA was isolated as described previously [63]. Transformation of M. smegmatis with plasmids was performed via electroporation [64] at 2.5 KV using an Electroporator 2510 (Eppendorf North America, Hauppauge, NY) and a cuvette with a 0.2 cm electrode-gap. For PCR amplification, Phusion polymerase with the GC buffer (Finnzymes Inc., Woburn, MA) was used. Plasmid purification and DNA recovery from agarose gels were done using Qiaprep and Qiaquick columns (Qiagen Inc., Valencia CA), respectively. Manipulations of DNA were performed using standard methods [65].

Construction of Protein Expression Plasmids and Bacterial Strains
The protein expression plasmids were based on pSMT3, a mycobacteria-Escherichia coli shuttle vector that allows selection for hygromycin resistance and gene expression under the control of the strong and constitutive hsp60 promoter [47,55]. To generate the plasmids pEP-hma and pEP-rv0132c for the expression of hma (rv0642c) and rv0132c of M. tuberculosis, respectively, the corresponding coding sequences along with the respective upstream regions bearing the promoters and ribosome-binding sites (253 bp for hma and 316 bp for rv0132c) and a bit of the downstream sequences (4 bp for hma and 20 bp for rv0132c) were PCRamplified from M. tuberculosis H37Rv chromosomal DNA and cloned into pSMT3; the cloning sites were EcoRV and ClaI for hma and BamHI and EcoRV for rv0132c. The primers used for this work have been described in the Supporting Material (Table S1). The cloned genes in pEP-hma and pEP-rv0132c were expressed in M. smegmatis from their native promoters and perhaps also from the plasmid resident mycobacterial hsp60 promoter. The plasmid pEP-rv0132c/hma that allowed simultaneous expression of hma and rv0132c was constructed by cloning the hma coding sequence along with the respective upstream and downstream sequences as mentioned above at the EcoRV and ClaI sites (or at the 39end of the rv0132c segment) of pEP-rv0132c. The construction of M. smegmatis DfbiC::aph strain has been described previously [53].

Preparation and Analysis of Mycolic Acid Methyl Esters
Mycolic acid methyl esters (MAMEs) were prepared as described previously [66]. Briefly, pelleted mycobacterial cells were saponified via incubation in 15% tetrabutyl ammonium hydroxide at 110uC overnight, followed by the addition of water, diazomethane, and dichloromethane and shaking at room temperature. From this mixture the MAMEs were recovered in the dichloromethane fraction and washed sequentially with equal volumes of water, 0.1 N HCl and water. The dichloromethane solution of MAMEs was dried under a stream of nitrogen, dissolved in a toluene-acetonitrile mixture (2:1), and then precipitated at room temperature with an addition of acetonitrile (final toluene:acetonitrile, 2:3). The pellet of MAMEs was dissolved in dichloromethane. Analysis of MAMEs was carried out by thin layer chromatography (TLC) on an aluminum-backed silica gel plate (10610 cm, Merck 5735-silica gel 60F254) by multiple developments using a solvent comprised of petroleum F420-Dependent Hydroxymycolic Acid Dehydrogenase PLOS ONE | www.plosone.org ether and diethyl ether (9:1). Mycolic acid spots were revealed by charring at 110uC for 15 min after spraying with 5% ethanolic molybdophosphoric acid [66].

Mass Spectrometric and NMR Analysis of Mycolic Acids
This work concerned the methyl esters of hydroxy-and ketomycolic acids. After performing multi-lane TLC separation for a sample, a terminal lane was cut off and processed for color development as described above. Then using a relevant band in this lane as a guide, the desired mycolic acid spots (silica layer) were scrapped off from the rest of the lanes. From the recovered silica particles, mycolic acids were extracted with dichloromethane and analyzed via MALDI-TOF mass spectrometry at School of Chemical Sciences Mass Spectrometry Laboratory at the University of Illinois at Urbana-Champaign. The Bruker peptide calibration mixture II (Angiotensin II, Angiotensin I, Substance P, Bombesin, ACTH clip 1-17, ACTH clip 18-39, Somatostatin 28, Bradykinin Fragment 1-7, and Renin Substrate Tetradecapeptide porcine Covered mass range: ,700 Da -3200 Da) was used for calibration and the matrix was 2,5-dihydroxybenzoic. A Bruker UltrafleXtreme mass spectrometer (Fahrenheitstr. 4,D-28359 Bremen, Germany) equipped with a smart beam II laser was used in the positive mode to acquire MALDI-TOF mass spectra. Samples were analyzed in the Reflectron mode.
A Bruker Avance III 600 MHz available at the NMR Laboratory, Department of Chemistry, Virginia Tech, was used to obtain 1 H NMR spectra of the purified mycolic acid methyl esters preparations described above. The solvent was CDCl 3 (100% D) and the reported chemical shifts were relative to the methyl resonances of tetramethylsilane (0 ppm). Figure S1 Structures of mycolic acids in and Mycobacterium tuberculosis complex and Mycobacterium smegmatis. The detailed structures of R groups in various mycolic acids are shown [4]. The reference cited here is listed in File S1. (TIF) Figure S2 Dose-dependent inhibition of K-MA production in M. smegmatis (pEP-rv0132c/hma) by PA-824. Wild-type M. smegmatis was used as control; the data in Fig. 3 show that neither the expression constructs pEP-hma and pEP-rv0132c nor the vector pSMT3 allow the production of K-MAs in M. smegmatis. The other details of the study have been presented in the MATERIALS AND METHODS. Mycolic acid types: a, a9, epoxy (E), hydroxy (H), and keto (K) [Fig. 1S shows the respective chemical structures.]. (TIF)

Supporting Information
Table S1 Oligonucleotides, plasmids, and strains. The references cited in this table are listed in File S1. (DOC) File S1 References for Table S1 and Figure S1. (DOC)