Molecular Dissection of the Role of Two Methyltransferases in the Biosynthesis of Phenolglycolipids and Phthiocerol Dimycoserosate in the Mycobacterium tuberculosis Complex*

A few mycobacterial species, most of which are pathogenic for humans, produce dimycocerosates of phthiocerol (DIM) and of glycosylated phenolphthiocerol, also called phenolglycolipid (PGL), two groups of molecules shown to be important virulence factors. The biosynthesis of these molecules is a very complex pathway that involves more than 15 enzymatic steps and has just begun to be elucidated. Most of the genes known to be involved in these pathways are clustered on the chromosome of M. tuberculosis. Based on their amino acid sequences, we hypothesized that the proteins encoded by Rv2952 and Rv2959c, two open reading frames of this locus, are involved in the transfer of methyl groups onto various hydroxyl functions during the biosynthesis of DIM, PGL, and related p-hydroxybenzoic acid derivatives (p-HBAD). Using allelic exchange and site-specific recombination, we produced three recombinant strains of Mycobacterium tuberculosis carrying insertions in Rv2952 or Rv2959c. Analysis of these mutants revealed that (i) the protein encoded by Rv2952 is a methyltransferase catalyzing the transfer of a methyl group onto the lipid moiety of phthiotriol and glycosylated phenolphthiotriol dimycocerosates to form DIM and PGL, respectively, (ii) Rv2959c is part of an operon including the newly characterized Rv2958c gene that encodes a glycosyltransferase also involved in PGL and p-HBAD biosynthesis, and (iii) the enzyme encoded by Rv2959c catalyzes the O-methylation of the hydroxyl group located on carbon 2 of the rhamnosyl residue linked to the phenolic group of PGL and p-HBAD produced by M. tuberculosis. These data further extend our understanding of the biosynthesis of important mycobacterial virulence factors and provide additional tools to decipher the molecular mechanisms of action of these molecules during the pathogenesis of tuberculosis.

Bacteria of the Mycobacterium tuberculosis complex exhibit amazing capacities to infect their host, resist bactericidal responses, and subvert the host immune response. Thus, M. tuberculosis, the causative agent of tuberculosis, has colonized one-third of the population of the world and kills more than two million people annually (1). This pathogenicity has been largely attributed to the unusual mycobacterial envelope. This complex structure has a high lipid content, up to 60% of the dry weight of the bacterium, and contains a large variety of lipids with unusual structures (2). Schematically, from the cytoplasm to the external side of the bacterium, the cell envelope is formed by (i) a plasma membrane, (ii) a cell wall core composed of three covalently attached macromolecules, i.e. the peptidoglycan, the arabinogalactan, and the mycolic acids, which are long chain (C60-C90) fatty acids, capped with a layer of non-covalently linked lipids and glycolipids, and (iii) a capsule of polysaccharide, proteins and lipids (2).
Among the extractable constituents of the cell envelope are two structurally related families of lipids, the diesters of phthiocerol and phenolglycolipids. The former class of lipids is a mixture of long chain ␤-diols, called phthiocerol and relatives, which are esterified by multimethyl-branched fatty acids (3). Depending on the stereochemistry of the chiral centers bearing the methyl branches, the fatty acids are called mycocerosic or phthioceranic acids (4). To date, phthiocerol dimycocerosates (DIM) 1 and diphthioceranates (DIP) have been identified in eight mycobacterial species; DIM have been found in M. tuberculosis, Mycobacterium bovis, Mycobacterium africanum, Mycobacterium leprae, Mycobacterium gastri, and Mycobacterium kansasii, whereas DIP have been found in Mycobacterium marinum and Mycobacterium ulcerans (4). With the exception of M. gastri, all the DIM-or DIP-containing species are pathogenic. The mycobacterial species that produce DIM or DIP may also synthesize structurally related substances, called phenolphthiocerols and relatives, in which phthiocerol is -terminated by an aromatic nucleus, usually glycosylated by a typeor species-specific mono-, tri-, or a tetrasaccharide unit leading to PGL (5,6). More recently, we identified a new group of molecules related to PGL, the glycosylated p-hydroxybenzoic acid methyl esters (p-HBAD), in the culture media of bacteria of the M. tuberculosis complex (7).
The genes involved in the biosynthesis of DIM, PGL, and p-HBAD are clustered on the chromosome of bacteria of the M. tuberculosis complex ( Fig. 1) (8 -11). Five of these genes, ppsA-E, encode a type-I modular polyketide synthase responsible for the synthesis of phthiocerol and phenolphtiocerol by the elongation of a C20-C22 fatty acyl chain or an acyl chain containing a phenol moiety with three malonyl-CoA and two methylmalonyl-CoA units (12). The pks15/1 gene encodes an iterative type-I polyketide synthase catalyzing the elongation of p-HBA to form p-hydroxyphenylalkanoates, which in turn are converted into phenolphthiocerol derivatives by the PpsA-E synthase (7). This gene is mutated in most M. tuberculosis clinical isolates, explaining why most strains of M. tuberculosis are unable to synthesize PGL even though they produce the structurally related DIM and p-HBAD. The mas gene encodes another iterative type-I polyketide synthase responsible for the synthesis of mycocerosic acids after 2-4 rounds of extension of C18-C20 fatty acids with methylmalonyl-CoA units (13). Two additional genes, fadD26 and fadD28 (also named acoas), encode two acyl-adenylate synthases involved in the formation of DIM and PGL presumably by activating the various polyketide synthase substrates (11,14,15). Finally, four genes (drrA, drrB, and drrC encoding an ABC transporter and mmpL7 encoding a transporter of the RND permease superfamily) are also located in this locus and are involved in the translocation of DIM from the cytoplasm to the bacterial cell surface (11). Close examination of this genomic region revealed five putative open reading frames (ORF) encoding proteins possibly involved in the glycosylation and methylation of DIM and PGL (7).
In the accompanying paper (16) we demonstrated that three of these genes (Rv2957, Rv2958c, and Rv2962c) encode proteins with similarities to glycosyltransferases and that these proteins are involved in the sequential elongation of both phenolphthiocerol dimycocerosates and p-hydroxybenzoic acid to form PGL and p-HBAD, respectively. In the present paper we show that the other two genes (Rv2959c and Rv2952) encode methyltransferases. We demonstrate (i) that the protein encoded by Rv2952 catalyzes the methylation of phthiotriol and phenophthiotriol to form phthiocerol and phenophthiocerol, respectively, and (ii) that the Rv2959c product is involved in the O-methylation of the hydroxyl group located at position 2 of the first rhamnosyl residue found in PGL-tb and p-HBAD of M. tuberculosis. We also show that Rv2959c and Rv2958c, which encodes the glycosyltransferase involved in the transfer of the terminal disaccharide moiety of PGL-tb and p-HBAD II (16), form an operon.

EXPERIMENTAL PROCEDURES
Most of the materials and methods used are described in the accompanying article (16). Only the experimental procedures specific to the present work are detailed.
Construction of M. tuberculosis H37Rv Mutants-PCR was carried out in a final volume of 50 l containing M. bovis Bacille Calmette-Guérin (BCG) genomic DNA, 1 mM primer A (Rv2959A or Rv2952A), 1 mM primer B (Rv2959B or Rv2952B) (Table I), 2.5 units of Taq DNA polymerase (Roche Applied Science), and 10% Me 2 SO. The amplification program consisted of 1 cycle of 10 min at 95°C followed by 35 cycles of 30 s at 95°C, 30 s at 57°C, and 3 min at 72°C and a final extension at 72°C for 10 min. The PCR product was analyzed by electrophoresis in 0.8% agarose gel. The resulting fragments, 2805 bp for Rv2952 and 2690 bp for Rv2959c, were gel-purified using the Qiaquick gel extraction purification kit (Qiagen, Courtaboeuf, France). The Rv2952 fragment was digested with SacI and XbaI, and the Rv2959c fragment was digested with NotI and XbaI. The pBlueScript vector was digested with SpeI and KpnI and religated yielding pPET2. The Rv2952c and Rv2959c PCR fragments were inserted into the pPET2 vector between the XbaI and SacI or between the XbaI and NotI restriction sites, respectively, to give pPET9 (Rv2952) and pPET5 (Rv2959c). A kanamycin resistance cassette (km) formed by the ⍀km cassette from pHP45⍀km (17) flanked by two res sites from transposon ␥␦ (18) was inserted between the KpnI and SalII sites of Rv2952, generating a 605-bp deletion, and the two MfeI and BglII sites of Rv2959c, generating a 290-bp deletion. The resulting plasmids were named pPET23 and pPET14, respectively. The PmeI fragments from the various plasmids, containing the disrupted gene constructs, were inserted at the XbaI site of pPR27 (19), generating pPET34 (⌬Rv2952::km) and pPET40 (Rv2959c::km). M. tuberculosis H37Rv was electrotransformed as previously described, and transformants were selected on 7H11 ϩ oleic acid-albumin-dextrose-catalase ϩ km at 32°C (20). Two transformants obtained with each of the plasmids were grown in 5 ml of 7H9 ϩ albumin-dextrose-catalase ϩ km ϩ Tween at 32°C until saturation. Dilutions of this culture were plated on 7H11 ϩ oleic acid-albumindextrose-catalase ϩ km ϩ sucrose and incubated at 39°C for 4 weeks. The colonies were screened by PCR using primers C, D, E, res1, or res2. The amplification program consisted of 1 cycle of 10 min at 95°C followed by 35 cycles of 30 s at 95°C, 30 s at 55°C, 3 min at 72°C, and a final 10 min at 72°C. Two clones giving the pattern corresponding to allelic exchange were retained for further analysis. These strains were renamed PMM28 (⌬Rv2952::km) and PMM28 (⌬Rv2959c::km).
Construction of M. tuberculosis H37Rv ⌬Rv2959c::res Unmarked Mutant-To recover the res-⍀km-res cassette from M. tuberculosis PMM18, we transferred the plasmid pWM19 containing the resolvase gene of transposon ␥␦ under the control of the mycobacterial promotor pBlaF* into this strain (18). The PMM18:pWM19 transformation mixture was resuspended in 5 ml of 7H9 ϩ albumin-dextrose-catalase and incubated for 48 h at 32°C to allow the expression of hygromycin resistance. Transformants were selected directly in the liquid medium by adding hygromycin to the transformation mixture and incubating the culture at 32°C for 12 days. Viable bacteria were then recovered by plating serial dilutions on 7H11 ϩ oleic acid-albumin-dextrose-catalase plates without antibiotics and incubating at 39°C, a non-permissive temperature for pWM19 replication. Twenty-one colonies picked randomly were then tested for their growth on km-containing plates. Thirteen of these colonies were unable to grow on km-containing plates but grew as the control on antibiotic-free plates. Three of these clones were analyzed by PCR using primers res1 ϩ 2959C, 2959E ϩ 2959C, and 2959F ϩ 2959G. The amplification program consisted of 1 cycle of 10 min at 95°C followed by 35 cycles of 30 s at 95°C, 30 s at 55°C, 3 min at 72°C, and a final extension of 10 min at 72°C. The amplification pattern revealed that the res-⍀km-res cassette had been excised in these three clones, leaving a copy of the 132-bp res site within the Rv2959c gene. One clone, named PMM18res, was retained for further analysis.

Disruption of Rv2952 and Rv2959c
Genes in M. tuberculosis H37Rv by Allelic Exchange-The biosynthesis of PGL, p-HBAD, and DIM in bacteria of the M. tuberculosis complex involves the transfer of several methyl groups onto the carbohydrate moieties of PGL and p-HBAD and onto the lipid domains of PGL and DIM. This implies that several unknown methyltransferases catalyze these reactions. Previous studies have shown that many genes involved in the biosynthesis of DIM and PGL are clustered on a 70-kilobase fragment of the M. tuberculosis chromosome; that is, the DIM and PGL locus ( Fig. 1). Therefore, we looked for ORFs encoding putative proteins with similarities to methyltransferase in this locus. Two such ORFs, Rv2952 and Rv2959c, were identified downstream of pks15/1. Interestingly, the protein encoded by Rv2952, but not that encoded by Rv2959c, harbored an amino acid motif conserved in several other mycobacterial methyltransferases shown to be required for the transfer of methyl groups onto fatty acids (21). Thus, we hypothesized that the Rv2959c product is involved in the methylation of the glycosyl moiety of PGL-tb and p-HBAD and that the Rv2952 product methylates the lipid domains of both DIM and PGL-tb.
To establish the exact roles of the proteins encoded by Rv2952 and Rv2959c in the biosynthesis of DIM, PGL-tb, and p-HBAD, we constructed two M. tuberculosis H37Rv mutants in which these genes were disrupted by allelic exchange. Briefly, chromosomal fragments overlapping the 5Ј or 3Ј ends of Rv2952 or Rv2959c were cloned flanking a km resistance cassette into the vector pPR27 (19) yielding plasmids pPET40 (Rv2959c) and pPET34 (Rv2952). These constructs were independently transferred by electroporation into M. tuberculosis H37Rv. Two transformants obtained with each plasmid were grown in liquid broth at 32°C until stationary phase and allelic exchange mutants were selected by plating serial dilutions of the cultures on solid medium containing kanamycin and sucrose at 39°C. PCR analysis of 10 colonies obtained with each construct on the counterselective plates using C ϩ D, C ϩ res2, or E ϩ res1 primers revealed that several clones gave an amplification pattern consistent with an allelic replacement of the wild-type allele of the different genes by the disrupted allele (Fig. 2). Two clones, named PMM18 (⌬Rv2959c::km) and PMM28 (⌬Rv2952::km), were retained for further studies.
Biochemical Phenotypes of Methyltransferase Mutants-To examine the effects of the mutations in the Rv2952 and Rv2959c genes on p-HBAD, DIM, and PGL production, we first transferred the plasmid pPET1 into the H37Rv strain and in the two mutants, PMM18 and PMM28. This plasmid carries a functional pks15/1 gene from M. bovis BCG because the wildtype H37Rv strain is unable to produce PGL-tb due to a frameshift mutation in pks15/1 (7).
The p-HBAD compounds are mostly found in the culture supernatant of the H37Rv strain of M. tuberculosis, whereas DIM and PGL-tb remain associated with the bacterial cells. Therefore, the mutants and the wild-type strain (either transformed with the plasmid pPET1 or not transformed) were grown in liquid culture, and lipids were extracted from both the culture supernatant and the bacterial pellets.
TLC analysis of the organic solvent extracts from the culture supernatants obtained with the various strains revealed the presence of one glycoconjugate spot in the lipid extract of H37Rv. It corresponds to the previously described p-HBAD II . This compound is the major p-HBAD produced by M. tuberculosis H37Rv. A minor compound p-HBAD I corresponding to 2-Omethyl-rhamnosyl-␣-p-hydroxybenzoic acid methyl ester is sometime visible but was undetectable in this experiment. The same pattern of glycoconjugate production was observed with the PMM28 mutant strain (Fig. 3A). In contrast, no glycoconjugates were detected in the organic-solvent extracts from the PMM18 mutant (Fig. 3A). Hence, the putative methyltransferase encoded by Rv2959c, but not that encoded by Rv2952, plays a role in the biosynthesis of p-HBAD.
To determine the roles of the putative methyltransferase in the biosynthesis of PGL-tb, lipids from the bacterial cells were examined. TLC analysis of these lipids showed that the PMM28:pPET1 mutant produced two glycoconjugates, products B and C (Fig. 3B). These molecules have slightly different mobilities to PGL from M. bovis BCG (2-O-methyl-rhamnosylphenolphthiocerol dimycocerosates) and from M. tuberculosis In the case of PMM18:pPET1 mutant, a glycoconjugate (product A, Fig. 3B) with a lower mobility than that of PGL-tb was observed. These results strongly suggest that both Rv2952 and Rv2959c are involved in PGL-tb biosynthesis.
Because PGL and DIM are structurally related, we investigated the roles of the putative methyltransferases in the biosynthesis of DIM. As expected, the control strain (H37Rv: pPET1) produced both DIM A (phthiocerol dimycocerosates) and DIM B (phthiodiolone dimycocerosate) (Fig. 3B). The mutation in the Rv2959c gene did not affect the production of DIM, as both forms were detected in similar amounts to those found in the wild-type strain (data not shown). In contrast, the mutation in the Rv2952 gene seemed to abolish the synthesis of DIM A completely but did not affect the production of DIM B both in PMM28 and PMM28:pPET1 (Fig. 3C). When large amounts of lipid extracts were loaded on TLC plates, an additional spot (product D) was detected in the organic solvent extracts of PMM28:pPET1 in comparison with that of the wild-type strain. Genes encoding proteins involved in the translocation of these molecules are in black. The roles of the various genes, established previously or during the course of this work, are indicated. The vertical arrows indicate ORFs for which no ortholog has been found in the M. leprae genome at the same position. Among these genes Rv2943, Rv2944, and Rv2961 encode putative polypeptides with similarities to proteins from mobile elements that are unlikely to be involved in the biosynthesis of DIM, PGL, or p-HBAD.
Thus, these preliminary analyses demonstrated that mutations in Rv2959c and Rv2952 affected the production of p-HBAD, PGL-tb, and DIM but in different ways. The protein encoded by Rv2959c seems to be required for the biosynthesis of both p-HBAD and PGL-tb but not DIM, whereas that encoded by Rv2952 appears to be required for the biosynthesis of PGL-tb and DIM A in M. tuberculosis but not p-HBAD. Because p-HBAD and PGL-tb share a common glycosyl-phenolic domain absent from DIM, whereas DIM and PGL-tb exhibit a common lipid domain absent from p-HBAD, these results support our hypothesis, based on the amino acid sequences of the two putative methyltransferases, that the Rv2959c product is involved in the methylation of the glycosyl moiety and that the Rv2952 product is involved in the methylation of the lipid domain.
Structural Analysis of the Compounds Accumulated in the M. tuberculosis H37Rv ⌬Rv2952::km Mutant-The mutations in both genes resulted in the accumulation of new compounds (Fig. 3) that may be biosynthetic intermediates of p-HBAD, PGL-tb, and DIM. To test this hypothesis, we purified these substances by chromatography on Florisil and analyzed them further.
First, products B and C, which accumulated in the M. tuberculosis PMM28:pPET1 mutant, were analyzed by MALDI-TOF mass spectrometry. The spectrum of glycoconjugate B showed a series of pseudomolecular ion (MϩNa ϩ ) peaks at 1850, 1864,1878,1892,1906,1920,1934,1948,1962,1976,1990, and 2004 m/z (Fig. 4B) (the major peaks are underlined). The same peaks were observed in the mass spectrum of the purified PGL-tb from the M. tuberculosis H37Rv:pPET1 strain (Fig. 4A), but the mass values of the major pseudomolecular ion peaks corresponding to glycolipid B from the PMM28:pPET1 mutant were 14 mass units lower than those in the spectrum of PGL-tb. This suggests that the PMM28:pPET1 mutant produces a PGL-like substance that may differ from PGL-tb by the absence of a methyl group. This hypothesis was supported by the 1 H NMR analysis of the glycolipid B, which showed that the 1 H NMR spectrum of the purified glycoconjugate B from PMM28:pPET1 was very similar to that of PGL-tb (Fig. 5A) (7,22). Two deshielded doublets were observed at 6.97 and 7.10 ppm and corresponded to proton resonances of the phenolic group of PGLs (7,22). Three anomeric proton resonances were seen in the 1 H NMR spectrum of the glycolipid B from PMM28:pPET1 at 5.50 ppm (1H) and 5.15 ppm (2H) (Fig. 5A). Their chemical shift values were identical to those of PGL-tb (7,22). In addition, four singlets were observed in the region of the resonances of sugar-linked methoxyl (OCH3) protons at 3.5-3.7 ppm. Again, the chemical shift values of the methoxyl proton resonances of product B (Fig. 5A) were identical to those found located on the trisaccharide portion of PGL-tb (7,22). These data strongly suggest that compound B has the same carbohydrate moiety as PGL-tb. The occurrence of polymethylenic (CH2) units in the glycolipid was deduced from the presence of a broad signal resonance at 1.25 ppm. The resonances of the expected terminal methyl (CH 3 ) protons were observed at 0.8 -1.0 ppm, whereas those attributable to methyl branches located on carbons 2 of fatty acyl residues were seen at 1.14 ppm. The resonance of the methine (CH) proton of the esterified ␤-glycol was seen at 4.83 ppm (Fig. 5A, signal a). Interestingly, the proton resonance typical of the methoxyl groups of the phthiocerol and phenolphthiocerol moiety of DIM A and PGL, respectively, expected at 3.32 ppm (singlet 3H) (7,22,23), was absent from the 1 H NMR spectrum of glycolipid B (Fig. 5A). This observation was consistent with the absence of a signal at 2.85 ppm (multiplet, 1H) assignable to the resonance of the methine proton of the carbon bearing this methoxyl group (7,22,23). Instead, a broad signal was observed at 3.31 ppm (1H) in the 1 H NMR spectrum of glycolipid B. This signal may correspond to a proton resonance of the carbon bearing a hydroxyl group (Fig.  5A, signal c). Altogether these data indicate that the glycolipid B from the PMM28:pPET1 mutant strain corresponds to a tri-O-methyl-fucosyl-(␣133)-rhamnosyl-(␣133)-2-O-rhamnosyl-␣-phenolphthiotriol dimycocerosates.
The same structural analyses were repeated with the glycoconjugate C purified from the PMM28:pPET1 mutant. MALDI-TOF mass spectrometry analysis of the product revealed a series of pseudomolecular ion (MϩNa ϩ ) peaks, at 2088, 2102, 2116, 2130, 2144, 2158, 2172, 2186, 2200, 2214, 2228, 2242, and 2256 m/z (Fig. 4C). These mass peaks were 238 mass units higher than those observed in the mass spectrum of the glycolipid B (Fig. 4B). Furthermore, the analysis of the 1 H NMR spectrum of glycoconjugate C revealed the PGL nature of the glycolipid C in that the spectrum (Fig. 5B) was identical to that of glycoconjugate B (Fig. 5A) except for an additional broad signal at 4.7 ppm (1H), which may correspond to the proton resonance of a carbon bearing an ester group (Fig. 5B, signal c), and the absence of the signal assigned to the resonance of the methine proton of the carbon bearing the hydroxyl group of the phenolphthiotriol moiety of glycolipid B (Fig. 5A, signal c). These data indicate that the free OH group of phenolphthiotriol found in glycoconjugate B is esterified in compound C by a fatty acid the acyl moiety of which has 239 mass units, e.g. a palmitoyl residue.
As well as producing two new glycoconjugates, the PMM28: pPET1 mutant synthesized a new apolar lipid, product D, the structure of which was solved by both MALDI-TOF mass spectrometry and 1 H NMR spectroscopy analyses. The mass spectrum revealed a series of pseudomolecular ion (MϩNa ϩ ) peaks at 1334, 1348, 1362, 1376, 1390, 1404, 1418, 1432, and 1446 m/z (data not shown). These mass values were 14 mass units lower than those observed in the mass spectrum of DIM A from wild-type strain (11). Comparison of the 1 H NMR spectra of product D and DIM A from the wild-type H37Rv strain showed that all the signal resonances detected with DIM A were present (11) in product D accumulated in PMM28:pPET1 (Fig. 5C), with the notable exception of two signals at 3.32 ppm (singlet, 1H) and 2.85 ppm (multiplet, 1H). These latter signals correspond, respectively, to the proton resonances of the methoxyl group and the carbon bearing this group (11). Remarkably, a broad signal at 3.31 ppm was observed in the spectrum of product D (1H). This signal was absent from the spectrum of DIM A (11) but present in that of the triglycosylphenolphthiotriol dimycocerosates (product B) (Fig. 5A). This proton resonance may correspond to that of a hydroxyl group in place of the methoxyl group of DIM A. Altogether these results indicate that PMM28:pPET1 synthesizes a new apolar lipid corresponding to phthiotriol dimycocerosates. Accordingly, structural analyses of the three compounds that accumulated in the PMM28:pPET1 strain revealed that the insertional mutation in the Rv2952 gene affected the conversion of the phenolphthiotriol and phthiotriol into their methoxylated forms usually found in PGL-tb and DIM A.
Structural Characterization of the PGL from M. tuberculosis H37Rv ⌬Rv2959c::km Mutant Strain-We used first MALDI-TOF mass spectrometry to determine the structure of the ad- ditional compound that accumulated in PMM18:pPET1 strain, compound A. The MALDI-TOF mass spectrum showed a series of pseudomolecular ion (MϩNa ϩ ) peaks at 1516, 1530, 1544,  1558, 1572, 1586, 1600, 1614, 1628, and 1642 m/z (Fig. 6A). The spectrum was similar to that of the PGL-tb from the H37Rv: pPET1 strain (Fig. 4A), but the mass values of the major pseudomolecular ion peaks were 348 mass units lower. This mass difference corresponds to that calculated for the tri-Omethyl-fucosyl-rhamnosyl moiety (334 mass units) plus that corresponding to the loss of a methyl group (14 mass units). Consistently, the mass values observed for compound A correspond to those obtained with the 2-O-methyl-rhamnosyl-phe-nolphthiocerol dimycocerosates produced by PMM24:pPET1 (16) minus 14 mass units (corresponding to the mass of a methyl group). Accordingly, we speculated that the sugar moiety of the glycoconjugate A accumulated in PMM18:pPET1 is an unmethylated rhamnosyl residue. To firmly establish the nature of the sugar residue, we performed an acid hydrolysis of the glycolipid A followed by trimethylsilylation and gas chromatography-mass spectrometry analysis. The mass spectra of the persilylated sugar derivative showed a (M ϩ ϩNH 4 ϩ ) mass peak at 470 m/z (in chemical ionization mode) and fragmentation peaks at 217, 204, 191, 147, and 73 m/z (in the electron impact mode). Identical mass spectra were obtained with a FIG . 5. Partial 1 H NMR spectra of purified glycolipid B (A), glycolipid C (B), and apolar lipid D (C) from PMM28:pPET1 (⌬Rv2952::km). Spectra were recorded in CDCl 3 (A and C) and CDCl 3 /CD 3 OD (95:5, v/v) at 500.13 Mhz. The structures of the analyzed compounds are shown above the spectra, and the protons corresponding to the main signals are indicated. p, pЈ ϭ 3-5; n, nЈ ϭ 16 -18; m2 ϭ 15-17; m1 ϭ 20 -22; R ϭ CH 2 -CH 3 or CH 3 ; RЈ ϭ CH 3 -(CH 2 ) 14 -CO. standard persilylated 6-deoxysugar, e.g. rhamnose. Finally, gas chromatography analysis of the sugar derivative from the acid hydrolysis of compound A identified rhamnose as the only sugar constituent of this new glycoconjugate (data not shown). These results suggest that the methyltransferase encoded by the Rv2959c gene catalyzes the transfer of the methyl group on rhamnosyl-phenolphthiocerol dimycocerosates to form 2-O-methyl-rhamnosyl-phenolphthiocerol dimycocerosate (also called mycoside B).
Production of a M. tuberculosis H37Rv Mutant Strain with an Unmarked Mutation within Rv2959c-The detection of rhamnosyl-phenolphthiocerol dimycocerosates in the lipid extract of PMM18:pPET1, but not that of other glycoconjugates, such as the tri-O-methyl-fucosyl-di-rhamnosyl phenolphthiocerol found as a minor constituent of M. tuberculosis Canetti (24), suggests that the transfer of the terminal disaccharidyl unit found in PGL-tb was also impaired in the mutant strain. We reasoned that either the glycosyltransferase(s) involved in the transfer of the two terminal sugars (16) do not recognize the unmethylated rhamnosyl-phenolphthiocerol dimycocerosates as a substrate, or alternatively, the enzyme(s) is not produced in the PMM18:pPET1 mutant. The second hypothesis was supported by a close examination of the H37Rv genome sequence, which suggested that Rv2959c is part of an operon that includes Rv2958c, which is involved in the transfer of the terminal two sugars onto mycoside B to form PGL-tb (see Ref. 16). Indeed, both genes are encoded by the same DNA strand, and the intergenic region contains only 101 nucleotides. Therefore, the insertion of the km cassette into Rv2959c may exert a polar effect on the expression of the downstream glycosyltransferase gene Rv2958c.
To study the possible polar effect of the km insertion into Rv2959c, we constructed a mutant with an unmarked mutation in the same gene. For this construct, we took advantage of the features of the cassette used to produce the PMM18 mutant. This cassette is formed by a kanamycin resistance gene flanked by two transcriptional terminators and two res sites from transposon ␥␦ (18). We previously demonstrated that, when the resolvase from transposon ␥␦ is expressed in a mycobacterial strain containing this res-⍀km-res cassette, site-specific recom-bination occurs between the two res sites, excising the cassette and leaving behind a single res sequence. Therefore, to produce an unmarked mutation within the Rv2959c gene, we transferred pWM19 (a thermosensitive plasmid allowing the expression of tnpR, the resolvase gene of transposon ␥␦ (18)) into PMM18. The transformants were plated on antibiotic-free plates and incubated at 39°C to cure the thermosensitive plasmid. Thirteen of the 21 clones tested were sensitive to km. PCR analysis of these clones using primers res1 ϩ 2959C, 2959E ϩ 2959C, or 2959F ϩ 2959G revealed that recombination occurred between the two res sites of the res-⍀km-res inserted within the Rv2959c gene in PMM18 (Fig. 2). Therefore, these clones contained a deletion of 290 nucleotides within the Rv2959c gene and an insertion of one res sequence at the same locus (Fig. 2). One of these clones, named PMM18res, was retained for further analysis. Because no transcription terminator has been found within the res sequence, the occurrence of this sequence within the Rv2959c gene should not prevent the transcription of the Rv2958c gene.
Biochemical analysis of the M. tuberculosis H37Rv ⌬Rv2959c::res Mutant Strain-To examine the effect of the unmarked mutation within Rv2959c on PGL-tb production, the plasmid pPET1 was transferred into the PMM18res strain. Lipids were then extracted with organic solvents from the bacterial cells of a PMM18res:pPET1 liquid culture.
TLC analysis of these lipids revealed two glycoconjugates spots (Fig. 7). The compound that exhibited the lowest mobility on TLC corresponded to the unmethylated rhamnosylphenolphthiocerol dimycocerosates (product A) previously characterized in the lipids from the PMM18:pPET1 mutant. The second glycoconjugate, product E, exhibited a higher mobility than product A (Fig. 7). This product was purified and analyzed by MALDI-TOF mass spectrometry and 1 H NMR spectroscopy. The mass spectrum showed a series of pseudomolecular ion (MϩNa ϩ ) peaks at 1850,1864,1878,1892,1906,1920,1934,1948,1962,1976,1990, and 2004 m/z (Fig. 6B). These mass values are 14 mass units (corresponding to the mass of a methyl group) lower than those of PGL-tb (Fig. 4A). To determine the nature of the sugar residues of the glycoconjugate, we subjected product E to an acid methanolysis followed by trimethylsilylation and gas chromatography analysis of the TMS derivatives of the methanolysates. This resulted in the identification of rhamnose, but not 2-O-methylrhamnose, in the methanolysis products of compound E (data not shown), indicating that the glycoconjugate under study corresponded to tri-O-methylfucosyl-di-rhamnosylphenolphthiocerol dimycocerosates.
Therefore, these results demonstrated that (i) Rv2959c and Rv2958c are part of the same operon and that the insertion of the res-⍀km-res cassette within Rv2959c affects the expression of Rv2958c and that (ii) Rv2959c encodes a methyltransferase involved in the transfer of a methyl group onto position 2 of the rhamnosyl residue directly linked to the phenolphthiocerol dimycocerosate moiety of PGL-tb. DISCUSSION It is becoming increasingly clear that the genes involved in the biosynthesis and translocation of DIM, PGL, and p-HBAD, a group of molecules that includes important mycobacterial virulence factors, are clustered on the chromosome of M. tuberculosis ( Fig. 1) (7,8,11,12,25). We, thus, investigated this chromosomal locus further, especially the role of two ORFs, Rv2952 and Rv2959c, putatively encoding two proteins with similarities to methyltransferases. For this purpose we constructed several recombinant strains of M. tuberculosis containing insertions of either a km cassette or a short DNA sequence carrying no transcriptional terminators within Rv2952 or Rv2959c. Biochemical and structural analyses of the various recombinant strains demonstrated that Rv2959c encodes the methyltransferase responsible for the methylation of position 2 of the first rhamnosyl residue of PGL-tb and p-HBAD. Additionally, we demonstrated that Rv2952 encodes the methyltransferase catalyzing the transfer of a methyl group onto the free hydroxyl group of DIM A and PGL precursors, i.e. phthiotriol and phenolphthiotriol dimycocerosates. This latter finding is consistent with several published observations. First, Rv2952 encodes a protein that contains the same sequence motifs as other methyltransferases involved in the methylation of hydroxyl groups found in fatty acids, such as Mtf2 in Mycobacterium smegmatis, or polyketides, such as RapM or SorM in Streptomyces (21). Second, a mutation in the Rv2952 gene abolishes the synthesis of DIM A and PGL-tb in M. tuberculosis but does not affect the production of the related molecule DIM B, which upon reduction of the keto function and subsequent methylation of the resulting hydroxyl group would lead to DIM A. Finally, the M. tuberculosis ⌬Rv2952::km mutant produces phthiotriol dimycocerosates and tri-O-methyl-fucosyl-(133)-rhamnosyl-(133)-2-O-rhamnosyl-phenolphthiotriol dimycocerosates, the two expected intermediates in the biosynthesis of DIM A and PGL-tb, respectively. In addition, we identified a new lipid of the PGL family that accumulated in the M. tuberculosis ⌬Rv2952::km mutant. Although we did not fully solve the structure of this compound in the present study, our results clearly indicated that this lipid is a triglycosyl-triacylated phenolphthiotriol. It is unlikely that this molecule is an intermediate in the synthesis of PGL-tb as the hydroxyl group of the phenolphthiotriol, which is usually methylated in PGL-tb, is esterified in the new molecule. It is tempting to speculate that this compound is a byproduct of PGL-tb biosynthesis and that it accumulates in the ⌬Rv2952::km mutant due to the lack of methylation of phenolphthiotriol dimycocerosates.
Concerning the methylation of the saccharide domain of PGL-tb, we showed that the enzymes catalyzing the O-methylation of position 2 of the rhamnosyl residue linked to the phenol group in PGL-tb are encoded by Rv2959c. We also demonstrated that this gene is part of an operon including a gene, Rv2958c, encoding the glycosyltransferase involved in the transfer of additional sugar residues onto position 3 of this rhamnosyl unit (16). These results are also consistent with the enzyme encoded by Rv2959c being involved in the methylation of p-HBAD. Indeed, the ⌬Rv2959c::km mutant strain no longer synthesizes p-HBAD I or p-HBAD II. The unmethylated p-HBAD are probably too polar to be recovered from the culture supernatant in the organic phase after extraction using the protocol described by Bligh and Dyer (26), and as a consequence, are probably lost during the extraction procedure. However, this result supports the model proposed in the accompanying paper, according to which the same enzymatic machinery is involved in the formation of the saccharide moieties of p-HBAD and PGL-tb. The accumulation of unmethylated rhamnosyl-phenolphthiocerol dimycocerosates and tri-O-methylfucosyl-di-rhamnosyl-phenolphthiocerol dimycocerosates in the ⌬Rv2959c::res mutant demonstrates that other methyltransferases are involved in the O-methylation of the terminal fucosyl residue of PGL-tb. This was not unexpected because it was unlikely that the same methyltransferase enzyme would be involved in the O-methylation of various positions of different sugar units. Indeed, different methyltransferases are required for the methylation of three positions of the same rhamnosyl residue during the biosynthesis of glycopeptidolipids in M. smegmatis (27). The novel glycoconjugates that accumulated in the ⌬Rv2959c::res mutant are probably biosynthetic intermediates in the formation of mycoside B and PGL-tb and were previously identified in a M. tuberculosis wild-type strain as minor products (23,24). Because no obvious gene candidate that may encode the methyltransferase(s) responsible for the O-methylation of the terminal fucosyl residue is present on the cluster of genes studied here, further studies are needed to identify this gene(s).
The two methyltransferases encoded by the Rv2952 and Rv2959c genes are highly conserved in M. leprae (83 and 77% identity, respectively, over the entire protein with their orthologs ML0130 and ML0127). For the enzyme involved in the methylation of the lipid domain of PGL and DIM, this was not unexpected as this domain is similar in these molecules in both species (4). However, this is not the case with the saccharide domain, which is a 3,6-di-O-methyl-glucosyl-di-O-methylrhamnosyl-3-O-methylrhamnoside in M. leprae, i.e. the methyl group at position 3 of the first rhamnosyl residue is located at a position different in PGL-tb. This suggests that, despite the very high sequence similarity between the two proteins encoded by Rv2959c and ML0130, the two enzymes have different specificities in terms of positions of the hydroxyl groups of the rhamnosyl residue that would be methylated.
In this work and in the accompanying study (16) we have provided new insight into the biosynthesis of DIM, PGL, and p-HBAD. We have characterized the roles of five new enzymes involved in the formation of the saccharide domains of PGL-tb and p-HBAD and in the modification of the lipid moieties of PGL-tb and DIM A. However, the enzymes involved in several other biosynthetic steps remain to be identified. For instance, it has been suggested that p-hydroxybenzoic acid is the precursor of PGL (2), but the source of this molecule in mycobacteria remains to be established. Additional work is clearly required to complete our understanding of these biosynthetic pathways. Nevertheless, it is noteworthy that more than 1.5% of the M. tuberculosis genome is devoted to the synthesis and trans-location of this group of molecules. The maintenance of such complex biosynthetic pathways, even in M. leprae, which has undergone massive gene decay, strongly suggests that these molecules play essential roles for the biology of mycobacteria. These roles remain to be clarified, but there is considerable evidence suggesting that they are related to pathogenicity. The various recombinant M. tuberculosis strains producing defined variants of DIM, PGL, and p-HBAD constructed during the course of the present study and the accompanying work (16) provide valuable tools for the characterization of the molecular mechanisms of action of these molecules in pathogenicity.