The Mycobacterium tuberculosis cmaA2 Gene Encodes a Mycolic Acid trans- Cyclopropane Synthetase*

Infection with Mycobacterium tuberculosis remains a major global health emergency. Although detailed understanding of the molecular events of M. tuberculosis pathogenesis is still limited, recent genetic analyses have implicated specific lipids of the cell envelope as important effectors in M. tuberculosis pathogenesis. We have shown that pcaA , a novel member of a family of M. tuberculosis S -adenosyl methionine (SAM)-dependent methyl transferases, is required for a -mycolic acid cyclopropanation and lethal chronic persistent M. tuberculosis infection. To examine the apparent redundancy between pcaA and cmaA2 , another cyclopropane synthetase of M. tuberculosis thought to be involved in a -my-colate synthesis , we have disrupted the cmaA2 gene in virulent M. tuberculosis by specialized transduction. Inactivation of cmaA2 causes accumulation of unsaturated derivatives of both the methoxy- and ketomycolates. Analysis by proton NMR indicates that the mycolic acids of the cmaA2 mutant lack trans- cyclopropane rings but are otherwise intact with respect to cyclopropane and methyl branch content. Thus, cmaA2 is required for the synthesis of the trans cyclopropane rings of both the methoxymycolates and ketomycolates. These results define cmaA2 as a trans -

Mycobacterium tuberculosis infection continues to overwhelm the populations of the developing world. It has been estimated that in 1997 there were 8 million new cases of active tuberculosis that were added to the already existing 16 million cases (1). In the same year, 2 million people died of tuberculosis as a result of an astonishing case fatality rate of 23-50% (1). This high death rate for a disease treatable with available antibiotics reflects the geographic superimposition of HIV 1 and M. tuberculosis infection, and the logistical and economic burden of at least 6 months of multidrug therapy required to treat the disease. New drugs to shorten therapy and vaccine candidates to prevent M. tuberculosis infection are badly needed but will only come with a more thorough understanding of the mechanisms of M. tuberculosis pathogenesis.
The cell envelope of M. tuberculosis is a highly complex array of distinctive lipids and glycolipids that has been intensely scrutinized as a potential effector in the interaction of M. tuberculosis with the human host (2)(3)(4). Investigation into the role of the cell envelope in virulence has been hampered by a lack of defined mutants of M. tuberculosis that fail to synthesize specific components of the cell surface. Recently, advances in the genetic manipulation of M. tuberculosis have allowed isolation of several mutants with defined cell envelope deficiencies and altered virulence (5)(6)(7). M. tuberculosis synthesizes three classes of mycolic acids, very long chain ␣-alkyl, ␤-hydroxyl fatty acids ( Fig. 1) in its cell envelope. These three classes of mycolic acids, ␣-, methoxy-, and ketomycolates, are modified with cyclopropane rings and methyl branches through the combined action of a large family of S-adenosyl methionine (SAM)-dependent methyl transferases that modify double bonds in the meromycolate chain. The oxygenated mycolic acids contain either cis-or trans-cyclopropane rings at their proximal position. Whereas the putative cis-cyclopropane synthetase of the methoxymycolates has been identified (8,9), the trans-cyclopropane synthetase is unknown. pcaA, one of the members of this distinctive gene family, has been established as essential for M. tuberculosis pathogenesis because a mutant of pcaA cannot establish a chronic persistent M. tuberculosis infection in mice (6). Biochemically, pcaA is required for the synthesis of the proximal cyclopropane ring of the ␣-mycolate molecule (Fig. 1). The finding that pcaA was required for proximal cyclopropanation of the ␣-mycolate molecule was surprising because this function had been previously attributed to cmaA2, another cyclopropane synthetase of M. tuberculosis (3,10,11). When introduced into Mycobacterium smegmatis on a multicopy plasmid, cmaA2 introduces cis-cyclopropane rings at the proximal position of the ␣-mycolate and the epoxymycolate, a position occupied by a double bond in the wild-type mycolates of this strain (11). Despite this lack of substrate specificity in M. smegmatis, the function of cmaA2 in M. tuberculosis was thought to be proximal cyclopropanation of the ␣-mycolate molecule. Thus, the functions of pcaA and cmaA2 appeared to overlap. To define the function of cmaA2 and to more completely explore the substrate specificity of the SAM-dependent methyl transferases of M. tuberculosis, we have inactivated cmaA2 in M. tuberculosis and shown here that cmaA2 is the trans-cyclopropane synthetase for both the methoxy-and ketomycolates.

Bacterial Strains and Growth Conditions-Wild-type M. tuberculosis
Erdman is a stock of an animal-passaged strain that has been passaged once in vitro. M. tuberculosis strain Erdman was grown at 37°C in 7H9 (broth) or 7H10 (agar) (Difco) media with OADC enrichment (Becton Dickinson), 0.5% glycerol, 0.05% Tween 80 (broth), and where appropriate, hygromycin (Roche Molecular Biochemicals) at 50 g/ml or kanamycin (Sigma) 20 g/ml. The M. tuberculosis Erdman strain with the ⌬cmaA2::hyg allele is designated mc 2 3120. For mycolic acid analysis, the wild-type strain was wild-type Erdman transformed with pmv306 hygro, an integrating vector that supplies a single copy hygromycin resistance gene.
Disruption of cmaA2 and Complementation-A ⌬cmaA2::hyg allele was constructed by amplifying the flanking regions of the cmaA2 gene and inserting these fragments on either side of the hygromycin resistance gene. Specifically, a 619-bp flanking region of cmaA2 5Ј to the start codon was amplified by PCR using primers omsg33 and omsg34, which contain XbaI and Asp7181 sites at their respective 5Ј-termini. A 646-bp flanking region 3Ј to the stop codon was amplified using primers omsg35 and omsg36, which introduce HindIII and SpeI sites, respectively. The PCR products were cloned, sequenced, and inserted flanking the hygromycin cassette in pMSG284, a cloning vector containing a bacteriophage lambda cos site, a PacI site, and the hygromycin resistance gene flanked by resolvase sites. The final knockout construct (pMSG104) was packaged into phAE87 as previously described (6) to create phMSG104. PhMSG104 was used to transduce wild-type M. tuberculosis to hygromycin resistance as previously described (6).
For complementation, an M. tuberculosis Erdman cosmid library was screened for cmaA2-containing clones by PCR. Cosmid 3E4 was digested with XbaI/NcoI and a 2093-bp fragment containing Rv504c, cmaA2, and part of Rv502 was cloned into pmv206 hyg to create pMSG129. To create an inframe deletion of Rv504c, the 1522-bp BstEII/ HindIII fragment, the 2279-bp BstEII/MluI fragment, and the 2203-bp MluI/HindIII fragment from pMSG129 were isolated after creating blunt ended BstEII ends with the Klenow fragment and were ligated in a three piece ligation to create pMSG133. Using this strategy, Rv504c was reduced to a truncated fusion protein of 81 amino acids. The reading frame of the fusion joint was confirmed by DNA sequencing. To create single copy complementation constructs, the inserts of pMSG129 and pMSG133 were subcloned into pmv306kan, a site-specific integrating mycobacterial vector (12) to create pMSG134 and pMSG136, respectively.
Expression of cmaA2 in M. smegmatis-M. smegmatis strain mc 2 155 was transformed with pMSG129 or vector control and total mycolic acids that were prepared as described below. Total mycolic acids were examined by proton NMR for the presence of cis-or trans-cyclopropane residues. For coexpression of cmaA2 with mmaA1, the mmaA1 openreading frame with its putative promoter (13) was cloned as a 1056-bp NgoMIV/AvrII fragment into pMSG137 digested with NgoMIV/NheI to create pMSG148.
Preparation and Analysis of Mycolic Acids-For radiolabeled mycolic acids, 50 ml of mid-log phase liquid cultures were incubated with 50 Ci of [ 14 C]acetate (PerkinElmer Life Sciences) for 12-18 h. Total mycolic acid methyl esters were prepared as described previously (6) and precipitated with toluene/acetonitrile. Analytical and preparative TLC was performed as previously described (6), and radio TLCs were analyzed on a phosphorimager cassette (Molecular Dynamics).
NMR Spectroscopy-One-dimensional 1 H NMR spectra were acquired at 27°C on either a Bruker DRX300 or DRX600 spectrometer in deuterochloroform (Cambridge Isotope Labs) and were referenced to the chloroform peak. Two-dimensional DQF-COSY and TOCSY NMR experiments were performed at 27°C on a Bruker DRX600 spectrometer equipped with a 5 mm TXI probe. Typically, 256 T 1 increments, each with 64 scans and 4000 data points over a spectral width of 5 kHz, were collected for each spectrum. The two-dimensional TOCSY experiment employed a 100 ms MLEV17 mixing sequence with a 9kHz spinlock field. Data processing and analysis was performed using Bruker XWIN-NMR software.
Sequence Analysis-Sequence alignment and phylogenetic tree construction was performed as described (14) on the Multalin server.

Inactivation of cmaA2 in M. tuberculosis by Allelic Exchange and Complementation with
Wild-type cmaA2-To define the function of cmaA2 in M. tuberculosis, we sought to delete cmaA2 from the chromosome of the Erdman strain of M. tuberculosis by allelic exchange. We constructed a substrate for allelic exchange at cmaA2 by replacing the coding region with a hygromycin resistance gene as described under "Experimental Procedures." We packaged this knockout construct into a specialized transducing mycobacteriophage and infected wildtype M. tuberculosis as previously described. (6,16). 2 Antibiotic-resistant M. tuberculosis clones were screened for allelic exchange at cmaA2 by Southern blotting. Three hygromycinresistant clones contained the cmaA2 disruption (Fig. 2B), and one was designated mc 2 3120 and used for further studies.
To show that any phenotype observed for the cmaA2 mutant was attributable to the cmaA2 mutation, we complemented the cmaA2 mutant with cmaA2 in single copy under its own promoter. Inspection of the genomic sequence surrounding cmaA2 suggests that this gene is transcribed as the second gene in a two gene operon with Rv504c, a gene of unknown function (see Fig. 2A for diagram). To complement the cmaA2 mutant with only cmaA2 under its native promoter, we reconstructed the cmaA2 operon with an inframe deletion in Rv504c and complemented the cmaA2 mutant in single copy with this inframe deletion construct (pMSG136). The strains mc 2 3120 and mc 2 3120 (pmsg136) were analyzed in the subsequent experiments.
Inactivation of cmaA2 Alters the Oxygenated Mycolic Acids of M. tuberculosis-As shown previously for pcaA (6) in the absence of a cyclopropane synthetase, the mycolic acids of M. tuberculosis would likely acquire an unsaturation. Therefore, we examined [ 14 C]acetate-labeled mycolic acids of the cmaA2 cmaA2 Encodes a trans-Cyclopropane Synthetase mutant by two-dimensional argentation TLC. This TLC system has been described previously for the analysis of M. tuberculosis mycolic acids (6,11). Briefly, the TLC plate is impregnated with silver nitrate leaving an unimpregnated strip along the left edge. The sample is developed first along the unimpregnated strip to separate the mycolates by polarity (Fig. 3A,  arrow 1). The plate is then developed in the silver dimension (Fig. 3A, arrow 2), separating the mycolates by degree of unsaturation. Silver nitrate retards the migration of unsaturated lipids relative to saturated or cyclopropanated lipids. Therefore, in the absence of a cyclopropane synthetase, an unsaturated mycolic acid retarded in the second dimension might appear.
The TLC pattern of wild-type M. tuberculosis mycolates is shown in Fig. 3A. The ␣-, methoxy-, and ketomycolates are labeled and correspond to the structures given in Fig. 1. cisand trans-Cyclopropanated oxygenated mycolates are not distinguished in this TLC system. Inactivation of cmaA2 alters the oxygenated mycolic acids. Specifically, two new mycolic acid species are visible in the cmaA2 mutant with the polarity of methoxy-and ketomycolates but which are retarded by silver impregnation (Fig. 3B). The ␣-mycolate of the cmaA2 mutant is identical to that from wild-type in its mobility. To demonstrate that this phenotype is due specifically to the loss of cmaA2, we examined the mycolic acids from the complemented mutant. Wild-type mycolic acid patterns were restored in the complemented strain, demonstrating that the altered oxygenated mycolates are secondary to the cmaA2 mutation (Fig. 3C). Thus, inactivation of cmaA2 causes the accumulation of an unsaturated subpopulation of oxygenated mycolates, demonstrating that cmaA2 is required for the proper cyclopropanation of these lipids.
Inactivation of cmaA2 Abolishes trans-Cyclopropanated Mycolates-Because the cmaA2 mutant has defects in a subpopulation of oxygenated mycolates, we reasoned that cmaA2 may be involved in either the cis or trans cyclopropanation of these molecules. To define the mycolic acid alteration in the cmaA2 mutant, we examined total mycolic acids from wild-type and the cmaA2 mutant by 1 H NMR, a technique that can clearly distinguish between cis-and trans-cyclopropane residues. The cis-and trans-cyclopropane proton resonances contributed by the three mycolic acid classes of wild-type M. tuberculosis are visible in the region of the NMR spectrum shown in Fig. 4A, top panel (3). In this expansion of the region from Ϫ0.4 ppm to 0.8 ppm, the characteristic resonances of cis-cyclopropane hydrogens (Ϫ0.33 ppm 2H, 0.56 ppm 1H) and trans-cyclopropane hydrogens (0.15 ppm 2H, 0.45 ppm 1H) can be distinguished (Fig. 4A, cis-and trans-cyclopropane structures label corresponding peaks). The cis-cyclopropane proton peak at 0.67 ppm (1H) and the trans-cyclopropane proton peak at 0.70 ppm are overlapping. In the wild-type Erdman strain used in this study, the ratio of cis/trans cyclopropane hydrogens is 8:1, lower than in previously examined laboratory strains (13).
The cmaA2 mutant lacks trans-cyclopropane rings, as evidenced by the complete absence of the complex mutiplets at 0.15 and 0.45 ppm in the spectrum shown in Fig. 4B. Importantly, the cis-cyclopropane resonances are unaffected. The TLC data presented above demonstrates that the oxygenated mycolates in the cmaA2 mutant contain a subpopulation of unsaturated mycolates. Accordingly, the NMR spectrum of the total mycolates from the cmaA2 mutant contains a complex multiplet at 5.33 ppm that is not present in wild-type mycolates (Fig. 4, A and B), consistent with the presence of the unsaturated mycolates in the mutant strain.
To further investigate the structure of the altered oxygenated mycolic acids in the cmaA2 mutant, we examined the mycolic acids of wild-type and mutant strains by two-dimensional COSY and TOCSY proton NMR spectroscopy. We first confirmed the previously reported structure of the cyclopropyl groups and their surrounding functional groups in total mycolic acids from wild type (Fig. 4D). According to the two-dimensional TOCSY spectrum, the cis-cyclopropyl hydrogen resonances at Ϫ0.33, 0.56, and 0.67 ppm all belong to the coupled spin network, as do the trans-cyclopropyl hydrogen resonances at 0.15, 0.45, and 0.7 (Fig. 4D). In addition, the trans-cyclopropyl group protons are adjacent to a methyl branch, as evidenced by a TOCSY cross-peak between the trans-cyclopropane proton resonances and a doublet at 0.95 ppm (Fig. 4D).
Two-dimensional TOCSY spectroscopy of purified methoxymycolates from the cmaA2 mutant confirmed the lack of transcyclopropyl protons demonstrated on the one-dimensional spectrum (Fig. 4E). In addition, the unsaturated derivatives of the methoxymycolates seen on TLC contain predominantly trans double bonds, as evidenced by the TOCSY cross-peak between FIG. 3. Radio two-dimensional TLC analysis of cmaA2 mutant mycolic acids. A, TLC system is described in detail in the text. The sample is developed along the left edge without silver impregnation (arrow 1 in A) and then in the second dimension with silver impregnation (arrow 2 in A). 14 C-labeled mycolates from wild-type M. tuberculosis Erdman (A), M. tuberculosis ⌬cmaA2::hyg (B), and M. tuberculosis ⌬cmaA2::hyg attB::pMSG136 (cmaA2) (C) are shown. The arrows in B point to new mycolic acid species with the polarity of methoxy-and ketomycolates that are retarded in the silver dimension the vinyl proton resonance centered at 5.33 ppm and the methyl branch resonance at 0.95 ppm (Fig. 4F and Ref. 8) and a COSY cross-peak between the vinyl protons and a methine proton resonance at 2 ppm (data not shown). cis-Cyclopropanes and cis double bonds in mycolic acids are not adjacent to methyl branches. Accordingly, a weak resonance at 5.39 ppm does not show a TOCSY cross-peak with the methyl branch at 0.95 ppm, indicating a small amount of cis-unsaturated methoxymycolate (Fig. 4F, box at 0.95 ppm) in the cmaA2 mutant methoxymycolates.
The trans cyclopropanation defect in the cmaA2 mutant was somewhat surprising as cmaA2 had previously been shown to catalyze the formation of cis-cyclopropane rings when overexpressed in M. smegmatis (11). Therefore, we considered whether the lack of trans-cyclopropane residues in the cmaA2 mutant could be an indirect effect on another, as yet undefined, cyclopropane synthetase. To investigate this possibility, we purified individual mycolate classes of the cmaA2 mutant by preparative TLC and examined them by proton NMR. Individual mycolate classes were examined for the presence of cyclopropane and methyl branch resonances known or likely to be added by the SAM-dependent methyl transferases of M. tuberculosis. The ␣-mycolate of the cmaA2 mutant was identical to wild-type ␣-mycolate (data not shown). As detailed above, the methoxymycolate of the cmaA2 mutant exhibited characteristic resonances of cis-cyclopropane protons, methyl branch protons adjacent to a methoxyl group (0.85 ppm, doublet), and the allylic methyl branch of the proximal trans double bond (0.95 ppm, doublet, Ref. 8). The ketomycolate also contained all predicted resonances except for the trans-cyclopropane residues. Therefore, as assessed by proton NMR of individual my-colate classes from the cmaA2 mutant, the only cyclopropane or methyl branch missing from the mycolic acids of the mutant is the trans-cyclopropane ring.
Expression of cmaA2 in M. smegmatis-The data presented above show that cmaA2 is the trans-cyclopropane synthetase of M. tuberculosis. To confirm that cmaA2 produces cis-cyclopropane rings in M. smegmatis as had been previously reported (11), we introduced cmaA2 into M. smegmatis on a multicopy plasmid under its own promoter. NMR examination of total mycolic acids from this strain revealed cis-cyclopropane proton resonances but not trans-cyclopropane proton resonances (data not shown). Mmas1 appears to catalyze the isomerization of the proximal cis double bond in oxygenated mycolates with the introduction of an allylic methyl branch (13). As this isomerization is necessary for trans-cyclopropane formation, we investigated whether cmaA2 would produce trans-cyclopropane rings in M. smegmatis when introduced with mmaA1. When coexpressed with mmaA1, cmaA2 still catalyzed only cis-cyclopropane formation (data not shown).

DISCUSSION
The mycolic acid methyl transferases of M. tuberculosis are a large family of highly homologous proteins that modify the mycolic acids of the cell wall with cyclopropane rings and methyl branches. Although cyclopropanated fatty acids are found in many bacteria (17), M. tuberculosis has evolved an elaborate enzymatic system of cyclopropane synthetases not found in any other bacteria. In this work we have shown that one of these transferases, cmaA2, is a trans-cyclopropane synthetase for oxygenated mycolates and that the other members of this gene family cannot compensate for the loss of cmaA2.

cmaA2 Encodes a trans-Cyclopropane Synthetase
All of the members of this gene family share striking amino acid sequence similarity. The sequence alignment of these proteins shown in Fig. 5 demonstrates that the individual cyclopropane synthetases share substantial amino acid identity over most of their length and that the sequence divergence between the members is limited to several distinct regions. Despite this striking sequence conservation, each member of this gene family appears to have a distinct catalytic function that cannot be compensated by another member of the family. Specifically, we have shown previously that inactivation of pcaA abolishes proximal cyclopropanation of the ␣-mycolate molecule despite intact cmaA2, mmaA2, and cmaA1 genes. It is interesting to note in the sequence alignment that CmaA2 contains an 8amino acid segment at amino acids 152-160 that is not present in any of the other methyl transferases. As CmaA2 is the only trans-cyclopropane synthetase of the group, this eight amino acid segment may be important for catalysis or substrate binding. In addition, a phylogenetic tree derived from these sequences demonstrates that there are three distinct groups within this gene family that are consistent with the known or cmaA2 Encodes a trans-Cyclopropane Synthetase suspected functions of these proteins (Fig. 5). The first group contains MmaA3 and MmaA4, two proteins that introduce the methoxy group in the distal position of the methoxymycolates (7-9, 18, 19). The second group contains CmaA2, MmaA1, and UmaA1. MmaA1 is likely responsible for the isomerization of the proximal cis double bond to a trans double bond in the meromycolate chain with simultaneous introduction of an allylic methyl branch (13). Because overexpression of MmaA1 in M. tuberculosis produced an excess of both trans unsaturated and trans-cyclopropanated mycolic acids, MmaA1 action is presumably an early step in trans-cyclopropane synthesis. It is therefore logical that CmaA2 is within the same phylogenetic subfamily. UmaA1 has no known function at present. The last group contains PcaA, CmaA1, and MmaA2. All of these enzymes are known or putative cis-cyclopropane synthetases. PcaA synthesizes the proximal cis-cyclopropane ring of the ␣-mycolates (6), CmaA1 produces a distal cis-cyclopropane ring in the ␣-mycolate of M. smegmatis (15), and MmaA2 likely synthesizes the proximal cis-cyclopropane ring of the methoxymycolates (8,9). Three-dimensional structural studies of these proteins may help elucidate the basis for their substrate specificity.
Several explanations are possible for the ability of cmaA2 to produce cis-cyclopropanes in M. smegmatis. Given the high sequence identity within this gene family, it is possible that cmaA2 can inefficiently catalyze cis-cyclopropane synthesis when highly overexpressed. Alternatively, the substrate specificity of these enzymes may be determined in part by physical association in multienzyme complexes. Although this possibility has not been examined experimentally, these enzymes catalyze the sequential modification of the meromycolate chain of mycolic acids and therefore could associate in multienzyme complexes to achieve efficient modification of a mycolic acid subclass. Therefore, it is possible that in M. smegmatis, CmaA2 cannot associate with other methyl transferases and the correct CmaA2 substrate is not available.
The significance of trans-cyclopropanated oxygenated mycolic acids for M. tuberculosis pathogenesis is unknown. However, previous work has shown that clinical strains of M. tuberculosis have higher trans-cyclopropane content than extensively passaged laboratory strains, suggesting that in vivo growth either dynamically enhances trans-cyclopropane formation or favors subpopulations of M. tuberculosis with higher trans-cyclopropane content (13). These results are consistent with the high proportion of trans-cyclopropane rings in the wild-type M. tuberculosis strain used in this study as this strain was recently passaged through animals and has not been passaged significantly in vitro. The results presented here define cmaA2 as the trans-cyclopropane synthetase of M. tuberculosis. Further examination of the cmaA2 mutant in animal models of infection will broaden our understanding of the role of individual cyclopropane residues in general, and of trans-cyclopropane residues in particular, in M. tuberculosis pathogenesis.