Modification of glycopeptidolipids by an O-methyltransferase of Mycobacterium smegmatis

All mycobacteria synthesize a unique cell envelope that contains a core complex of covalently linked peptidoglycan, arabinogalactan and mycolic acids (Brennan & Nikaido, 1995 ). The mycolic acids form the inner leaflet of an outer asymmetric lipid bilayer. The outer leaflet of this bilayer is composed of several distinct classes of glycolipids and phospholipids that contribute to the surface properties of mycobacterial cells. Glycopeptidolipids (GPLs) are a major outer-layer glycolipid of several species of mycobacteria. The GPLs are important surface antigens in the opportunistic pathogens of the Mycobacterium avium complex, and are also abundant outer-layer components in saprophytic species such as Mycobacterium smegmatis (Daffé et al., 1983 ). Studies have shown that GPLs of M. avium are immunosuppressive and induce some cytokines, thus affecting the hosts response to infection (Chatterjee & Khoo, 2001 ; Horgen et al., 2000 ). Most mycobacterial GPLs share a common lipopeptide core, in which the peptide headgroup, D-Phe-D-alloThr-D-Ala-L-alaninol, is modified (on the amino-terminal D-Phe) with a mixture of amide-linked 3-hydroxy and 3-methoxy C₂₆₃₄ fatty acids (Aspinall et al., 1995 ; Belisle et al., 1993 ; Daffé et al., 1983 ). This lipopeptide core is substituted with 6-deoxytalose (6-dTal), which may be variably O-acetylated, and rhamnose, which may be variably O-methylated (Patterson et al., 2000 ). In M. avium GPLs, the 6-dTal may be further extended with haptenic oligosaccharides, resulting in highly antigenic serovar-specific GPLs (ssGPLs) (Aspinall et al., 1995 ; Belisle et al., 1993 ; Daffé et al., 1983 ). The GPLs of M. smegmatis, which are not glycosylated at 6-dTal, are non-serovar-specific GPLs and represent the biosynthetic precursor of the ssGPLs of M. avium (Eckstein et al., 1998 ). The M. avium genomic region encoding glycosylation of serovar-2-specific GPLs has been identified (Belisle et al., 1991 , 1993 ; Mills et al., 1994 ). Deletion of the ser2 gene cluster results in GPL-deficient mutants that have rough colony morphology (Eckstein et al., 2000 ).

Several M. smegmatis genes involved in the synthesis of the GPL core have now been characterized, and shown to encode a peptide synthetase (Mps), a rhamnose O-methyltransferase (Mtf1), an acetyltransferase (Atf1) and a potential transporter protein (TmptC) (Billman-Jacobe et al., 1999 ; Patterson et al., 2000 ; Recht & Kolter, 2001 ; Recht et al., 2000 ). Disruption of all these genes results in bacteria with a rough colony morphology. In this paper we describe the M. smegmatis GPL biosynthetic gene cluster and identify the methyltransferase that catalyses the conversion of the fatty acid from the 3-hydroxy to the 3-methoxy form.

Bacterial strains and plasmids.
Wild-type M. smegmatis mc²155 (Snapper et al., 1990 ) and strains derived from it were grown on LB medium at 37 °C. Kanamycin and streptomycin were added to a final concentration of 20 µg ml^-1 when required. Sucrose was used at 10% (w/v) for negative selection against clones expressing sacB (Pelicic et al., 1996 ). Liquid cultures were shaken at 180 r.p.m. and Tween 80 (0·02%, v/v) was added to the medium. A list of the bacterial strains and plasmids used is provided in Table 1.

Table 1. Bacterial strains and plasmids used

Cloning and genetic manipulation.
Cloning procedures were performed according to standard protocols (Sambrook et al., 1989 ). Electrotransformation was performed using a Bio-Rad Gene Pulser and Escherichia coli cells were prepared for transformation according to the manufacturers recommendation. M. smegmatis was prepared for transformation according to Jacobs et al. (1991) . Genomic DNA was extracted from M. smegmatis as previously described (Billman-Jacobe et al., 1999 ). Probes for Southern blots were labelled with digoxigenin (DIG)-labelled dNTPs using a Roche DIG labelling kit, and membranes were developed according to the manufacturers instructions. Restriction endonucleases and DNA modification enzymes were obtained from Pharmacia. Sequencing was carried out on a thermal cycler using dye terminator reactions that were analysed using an ABI DNA sequencer. BLAST analysis was used for similarity searches and prediction of the function of the proteins encoded by ORFs (Altschul et al., 1990 ).

Disruption of mtf2.
The plasmid, pHBJ45, which was used to make a gene-disruption cassette, was pUC18 containing the entire mtf2 ORF and flanking sequences on a 3·6 kb EcoRIHindIII fragment (Patterson et al., 2000 ). To disrupt mtf2, the plasmid was linearized by digestion with BglII and treated with T4 DNA polymerase. A streptomycin-resistance gene was excised from pUC19Ω (Prentki & Krisch, 1984 ) by digestion with HindIII and EcoRI and treated with T4 DNA polymerase. The streptomycin-resistance cassette was ligated to pHBJ45 and the recombinant plasmid was transformed into E. coli. The resulting plasmid, pHBJ370, contained mtf2::str on a 5·6 kb fragment which was then excised by HindIIIEcoRI digestion and cloned into HindIIIEcoRI-digested pK18mobsacB (Schafer et al., 1994 ) to create pHBJ371 for gene replacement.

For complementation of mtf2 mutants, a 1·7 kb SalI fragment containing the entire mtf2 ORF was cloned into the SalI site of a mycobacterial shuttle vector, pMV261 (Stover et al., 1991 ). Recombinant plasmids with the insert cloned in both orientations were isolated to test if expression of the cloned gene was independent of the vector-encoded GroEL promoter. The plasmids were named pHBJ389 and pHBJ390 (Table 1).

Extraction of cell-wall-associated GPLs.
Wild-type, mutant and complemented strains were grown in LB broth at 37 °C and cells were harvested by centrifugation (2000 g, 15 min) when in mid-exponential phase. Cells were washed in phosphate-buffered saline and then extracted in 20 vols of chloroform/methanol (2:1, v/v) for 2 h at room temperature with constant agitation. Cell wall lipids were recovered in the chloroform phase after Folch partitioning (Folch et al., 1957 ). The lipid extract was subjected to alkaline methanolysis (0·2 M NaOH in methanol, 2 h, 40 °C) to cleave ester-linked fatty acids. The alkali-stable GPLs were recovered in the upper 1-butanol phase after biphasic partitioning between water-saturated 1-butanol (4 vols) and water (2 vols).

HPTLC analyses and purification of GPLs.
The GPL extracts were analysed by high-performance thin-layer chromatography (HPTLC) using aluminium-backed silica Gel 60 HPTLC sheets (Merck). The HPTLC sheets were developed in chloroform/methanol (9:1, v/v) for chromatographic separation of the GPLs. Individual GPL species were purified using the same solvent system. Silica bands were scraped and extracted twice in chloroform/methanol (2:1, v/v) with sonication. Glycolipids were visualized by spraying with orcinol/H₂SO₄ and charring at 100 °C.

Compositional analysis.
Monosaccharide analyses of crude GPL extracts or purified species were performed after hydrolysis in 2 M trifluoroacetic acid (100 °C, 2 h) and conversion of the released monosaccharides into their corresponding alditol acetate derivatives (Billman-Jacobe et al., 1999 ). The alditol acetate derivatives were separated and analysed by GC-MS on a Hewlett Packard HP-1 column using a Hewlett Packard model 6890 gas chromatograph and a model 5973 mass detector.

Electrospray ionization mass spectrometry (ESI-MS).
HPTLC-purified GPL species were analysed with a Bruker Esquire 3000 electrospray ionization mass spectrometer fitted with a nanospray attachment. GPLs were analysed in negative-ion mode.

Sequence analysis of the GPL gene cluster
The GPL biosynthetic cluster of M. smegmatis contains 13 ORFs that are known or thought to be involved in the synthesis or transport of GPLs (Fig. 1). The cluster is bounded upstream by nitrile hydratase genes (not shown) and downstream by a copy of IS1096 (Cirillo et al., 1991 ). The mps gene, encoding the peptide synthetase, is located immediately downstream of IS1096 (Billman-Jacobe et al., 1999 ). The known or proposed functions of the ORFs in this locus are listed in Table 2 and are described below.

(5K): Fig. 1. Organization of the GPL biosynthetic locus. Genes that have been disrupted are indicated with an asterisk.

Table 2. Characteristics of genes and their deduced products and similarities to other proteins

Genes involved in GPL transport. ORFs tmptA, tmptB and tmptC are highly similar to three ORFs in Mycobacterium tuberculosis and M. avium that are predicted to be membrane transport proteins. M. smegmatis mutants that have a disrupted tmptC do not produce GPLs; however the role of TmptC is not known (Recht et al., 2000 ).

Genes involved in deoxy sugar synthesis. Genes encoding the four enzymes for rhamnose synthesis (glucose-1-phosphate thymidylyltransferase, dTDP-glucose 4,6-dehydrogenase, dTDP-4-dehydrorhamnose 3,5-epimerase and dTDP dehydrorhamnose reductase) are conserved among bacteria (Stevenson et al., 1994 ). Rhamnose biosynthesis has been described in M. tuberculosis and M. smegmatis and genes encoding the enzymes above have been identified in M. tuberculosis (Cole et al., 1998 ; Ma et al., 1997 ). Two genes in the GPL locus possibly encode the enzymes for the first two steps of rhamnose biosynthesis. The rmlA gene product is 92% similar to RmlA of M. tuberculosis, which encodes glucose-1-phosphate thymidylyltransferase and has been shown to convert glucose 1-phosphate to dTDP-glucose (Liu & Thorson, 1994 ; Ma et al., 1997 ). Several glucose-1-phosphate thymidylyltransferases have a conserved motif in the N-terminus which is also observed in the M. smegmatis RmlA polypeptide sequence (LAGGSGTRLHP) (Aguirrezabalaga et al., 2000 ; Merson-Davies et al., 1994 ; Steffensky et al., 2000 ). The second gene that is possibly involved in 6-deoxy sugar synthesis is rmlB. The deduced product of rmlB is proposed to be a dTDP-glucose 4,6-dehydrogenase which would catalyse conversion of dTDP-glucose to dTDP-xylohexulose, the second step in dTDP-deoxy sugar synthesis (Liu & Thorson, 1994 ; Ma et al., 1997 ). The amino acid sequence is most similar to GepiA (86% similarity) of M. avium; however, it is also similar to several possible dTDP-glucose 4,6-dehydrogenases of M. tuberculosis including RmlB2 (75% similarity). The M. smegmatis RmlB has a NAD cofactor-binding motif (TGGAGFIG) at the N-terminus and a second motif (YAATKLAQE) which is conserved in all members of the short-chain dehydrogenase/reductase family of enzymes, of which nucleotide sugar dehydratases are a subfamily (Baker & Blasco, 1992 ; Gerratana et al., 2001 ; Jornvall et al., 1995 ).

Putative glycosyltransferases. There are three proposed glycosyltransferase genes in the GPL biosynthetic locus. These three ORFs, gtf1, gtf2 and gtf3, are similar to putative M. avium glycosyltransferases, GtfA (7585% similarity), GtfB (7178% similarity) and the authentic rhamnosyltransferase RtfA (7275% similarity) (Eckstein et al., 1998 ). Gft1 is also similar to M. tuberculosis Rv1524 (Cole et al., 1998 ).

Acetyltransferase. The gene product of atf1 is thought to be an O-acetyltransferase because an M. smegmatis transposon mutant with a disrupted atf1 gene synthesized nonacetylated GPLs (Recht & Kolter, 2001 ). The deduced protein encoded by atf1 is 70% similar to the putative acetyltransferases of M. avium subsp. paratuberculosis (Bull et al., 2000 ) and M. avium AtfA.

Methyltransferases. There are four ORFs that could encode methyltransferases in the GPL cluster. They all have conserved S-adenosylmethionine-binding motifs and, to varying degrees, display motifs conserved in other methyltransferases (Fig. 2) (Malone et al., 1995 ). M. smegmatis mtf1 has been shown to encode a 3-O-methyltransferase that adds the first methyl group to the terminal rhamnose on the GPL core (Patterson et al., 2000 ). M. smegmatis mtf3 is predicted to encode a protein that is similar to the authentic rhamnosyl O-methyltransferase ElmMIII of Streptomyces olivaceus (Patallo et al., 2001 ) and the putative M. avium O-methyltransferases, MtfC and MtfB. M. smegmatis mtf4 is predicted to encode a protein that is similar to authentic rhamnosyl O-methyltransferases of S. olivaceus ElmMI (55% similarity) and ElmMII (52% similarity) (Fig. 2). These analyses suggest that mtf1, mtf3 and mtf4 are likely to be required for the addition of the three methyl groups to GPL rhamnose or other sugars. This raises the question of the function of the fourth mtf gene, mtf2. This gene is predicted to encode a protein that is similar to two putative M. tuberculosis methyltransferases: Rv2952 (66% similarity) and Rv1523 (65% similarity) (Cole et al., 1998 ). It is also similar to RapM (53% similarity) of the polyketide producer Streptomyces hygroscopicus.

(58K): Fig. 2. Amino acid sequences of motifs strongly conserved with other authentic (*) and proposed methyltransferases: M. smegmatis Mtf1*, Mtf2*, Mtf3 and Mtf4; Strepomyces olivaceus ElmMI*, ElmMII* and ElmMIII*; Mycobacterium avium MtfB, MtfC and MtfD, Mycobacterium tuberculosis Rv1523 and Rv2952; Streptomyces avermitilis AveBVII; Micromonospora griseorubida MycF; Streptomyces fradiae TylF; Polyangium cellulosum SorM; and Streptomyces hygroscopicus RapM. Conserved amino acids are grouped as E, D, Q and N; V, L, I and M; F, Y and W; G, P and A; K and R; and S and T using the standard one-letter abbreviations. Residues conserved in at least seven members of the group are shown in bold, underlined type. Motifs are numbered in accordance with Malone et al. (1995) .

In contrast, M. smegmatis Mtf2 is more like a subgroup of methyltransferase-like proteins which includes RapM, SorM and two putative methyltransferases of M. tuberculosis (group 2 in Fig. 2). RapM and SorM are methyltransferases from species of Streptomyces that make the non-glycosylated polyketides rapamycin and soraphen (Chung et al., 2001 ; Schupp et al., 1995 ). Both of the enzymes methylate hydroxy groups on the macrocyclic rings rather than sugars. On this basis, we predicted that Mtf2 may be involved in O-methylating GPL fatty acids. Other members of group 2 are proposed methyltransferases of M. tuberculosis, Rv2952 and Rv1523. It is pertinent to note that M. tuberculosis does not synthesize GPLs and the roles of the Rv2952 and Rv1523 enzymes are unknown.

Disruption of mtf2 in M. smegmatis
We decided to disrupt the ORF of mtf2 to determine the role of the enzyme encoded by it. The disruption plasmid, pHBJ371 (see above) contained the mtf2 ORF disrupted by insertion of the 2 kb streptomycin-resistance marker cloned into a vector carrying a kanamycin-resistance marker and the sacB gene encoding levansucrase, but it lacked a mycobacterial origin of replication. M. smegmatis mc²155 was transformed with pHBJ371, and resulting transformants selected with streptomycin and kanamycin were then cultured on media containing streptomycin and sucrose to select for mutants that had undergone mtf2 gene replacement. Four kanamycin-sensitive, streptomycin- and sucrose-resistant transformants were characterized further.

Purified genomic DNA was digested with PstI and EcoRI and then transferred to a nylon membrane for Southern hybridization. The membrane was probed with a DIG-labelled probe generated by PCR amplification of mtf2 using the oligonucleotide primers #109 (5'-AGGCTGAAAGACGACGAC-3') and #239 (5'-GATTCGAGCGCATCGA-3'). Fig. 3 shows that the probe bound to a 4·2 kb fragment in mc²155 as expected. Three putative double crossover mutants each had a hybridizing band of 6·2 kb, indicating that gene replacement had occurred. In the remaining clone, the entire pHBJ371 had integrated through a single homologous recombination. The colony morphology of all the mutants was smooth and indistinguishable from the parent strain. A mutant that had undergone replacement of mtf2, Myco493, was selected for further analysis.

(57K): Fig. 3. Southern blot of PstIEcoRI-digested genomic DNA of: lane 1, mc2155; lane 2, a single-crossover mutant; and lanes 36, mutants that had a double crossover to replace mtf2 with mtf2::str. The blot was probed with sequences derived from mtf2 by PCR. Size markers are indicated on the left.

The parental mc²155 and Myco493 strains synthesized approximately equal amounts of GPLs. However, HPTLC analysis of the alkali-treated GPLs revealed that the mutant lacked two (GPL-1 and -3) of the four major GPL molecular species synthesized by the wild-type strain (Fig. 4). Monosaccharide analysis revealed that the upper and lower GPL bands in the mutant profile had the same monosaccharide composition as GPL-2 and GPL-4 in the parental profile, which differ primarily in containing 2,3,4-tri-O-methylrhamnose and 3,4-dimethylrhamnose, respectively (Fig. 5). All purified GPLs contained 6-dTal. These data indicate that the methylation of the rhamnose moiety of the GPLs was not affected by disruption of mtf2 and that the difference in the GPL profile observed by HPTLC may be due to changes in the O-methylation of the amide-linked fatty acid. This was supported by ESI-MS analysis of the two major Myco493 GPLs species. As shown previously, each GPL comprises a number of molecular species due to heterogeneity in the length and degree of unsaturation of the amide-linked fatty acid (Patterson et al., 2000 ). The major molecular ions of the upper and lower bands from Myco493 were identical to those of the parental GPL-2 and -4, and 14 atomic mass units lower than GPL-1 and -3. Taken together with the monosaccharide compositional analyses, these data show that the Myco493 GPLs contain an hydroxyl rather than the methoxyl-substituted fatty acid (Table 3).

(56K): Fig. 4. HPTLC of alkali-stable glycolipids stained with orcinol/H2SO4.

(9K): Fig. 5. GC-MS analysis of the monosaccharides from purified alkali-stable GPLs extracted from M. smegmatis: (a) mc2155 GPL-2, (b) mc2155 GPL-4, (c) Myco493 upper band (GPL-2), (d) Myco493 lower band (GPL-4). A reagent peak was detected at 11 min in all samples and standards.

Table 3. Fatty acid compositions deduced from the molecular masses of purified GPLs in negative ion ESI-MS spectra

Complementation
The mutant was complemented with two versions of a plasmid containing a 1·7 kb SalI fragment containing the mtf2 coding region plus 534 bp of sequence upstream of the mtf2 translational start codon. The fragment was cloned into a mycobacterial shuttle vector, pMV261 (Stover et al., 1991 ), and a clone containing the insert in each orientation was obtained. The presence of either plasmid, pHBJ389 or pHBJ390, in Myco493 restored the ability to make all four GPLs (Fig. 4). Both complementation plasmids had the same effect, suggesting that expression of mtf2 was independent of the vector-encoded promoter. Extracts from complemented Myco493 contained more GPL-1 and -3 than GPL-2 and -4 (Fig. 4). The same effect was observed in mc²155 (data not shown). The plasmid would be expected to be present in approximately five copies per genome (Stover et al., 1991 ) and it is possible that the higher gene dosage led to higher levels of expression of the Mtf2 O-methyltransferase and thus more of the 3-hydroxy fatty acid may have been converted to the 3-methoxy form. ESI-MS of the major bands from the complemented mutant confirmed the hypothesis, as the molecular masses were consistent with the lipid being the methoxy form (Table 3, Fig. 6).

(13K): Fig. 6. Negative-ion ESI-MS spectra of (a) Myco493 GPL-2, (b) Myco493 GPL-4, (c) Myco493/pHBJ389 GPL-1, (d) Myco493/pHBJ389 GPL-3.

All the GPLs of M. smegmatis contain a conserved diglycosylated lipopeptide core that is variably O-methylated on both the terminal rhamnose residue and amide-linked fatty acid. Several genes have been ascribed functions for GPL biosynthesis based on the characterization of mutants generated by transposon mutagenesis (Billman-Jacobe et al., 1999 ; Patterson et al., 2000 ; Recht & Kolter, 2001 ; Recht et al., 2000 ). In this report we describe a locus containing genes encoding enzymes for GPL biosynthesis and assign a function to mtf2. The locus spans 18·6 kb of the M. smegmatis chromosome and contains 13 ORFs including those previously identified (Patterson et al., 2000 ; Recht & Kolter, 2001 ; Recht et al., 2000 ). The locus is interrupted by a copy of IS1096 (Cirillo et al., 1991 ) and then continues with the 17·9 kb peptide synthetase gene, mps (Billman-Jacobe et al., 1999 ).

Genes encoding the four enzymes for rhamnose synthesis are conserved among bacteria and are arranged in an operon in enteric bacteria (Stevenson et al., 1994 ). In all mycobacteria, including non-GPL-expressing species, rhamnose biosynthesis is of interest because this residue links the arabinogalactan portion of the cell wall core to peptidoglycan (McNeil et al., 1990 ). Rhamnose biosynthesis has been delineated in M. tuberculosis and M. smegmatis and genes encoding the key enzymes have been characterized in M. tuberculosis (Cole et al., 1998 ; Ma et al., 1997 ). Unlike the situation in enteric bacteria the mycobacterial genes for rhamnose biosynthesis are dispersed around the chromosome in a similar fashion to that described in Saccharopolyspora spinosa (Madduri et al., 2001 ). Rhamnose is also required for glycosylation of GPLs in M. smegmatis, and genes for glucose-1-phosphate thymidylyltransferase (rmlA) and dTDP glucose 4,6-dehydrogenase (rmlB) occur in the GPL biosynthetic locus. Rhamnose and 6-dTal are derived from identical biosynthetic precursors, with two dTDP-dehydrorhamnose reductases having opposite stereospecificity providing either TDP-rhamnose or TDP-6-dTal (Liu & Thorson, 1994 ). The steps catalysed by RmlA and RmlB will therefore lead to synthesis of intermediates that could be converted to either of the sugars characteristic of GPLs.

Four methyltransferase genes occur in the GPL locus. The methyltransferases were tentatively divided into two groups on the basis of the amino acid sequence similarity in four methyltransferase motifs (Fig. 2). In DNA methyltransferases, motifs I and V contact S-adenosylmethionine, the methyl donor, and motifs IV and VI contribute to the structure of the active site (Malone et al., 1995 ). Fig. 2 shows an alignment of the amino acid sequences that are conserved among selected authentic and putative methyltransferases. Group 1 includes the M. smegmatis methyltransferases (Mtf1, Mtf3 and Mtf4) which are like the authentic methyltransferases, ElmMI, ElmMII and ElmMIII, that methylate rhamnose during synthesis of the antitumour polyketide elloramycin in S. olivaceus (Patallo et al., 2001 ). They are also similar to the putative methyltransferases of the ser2 cluster in M. avium (MtfB, MtfC and MtfD), which may carry out rhamnose modification in GPL synthesis. AveBVII, MycF and TylF methylate the sugar components in the macrolides avermectin, mycinamicin and tylosin, respectively (Fouces et al., 1999 ; Ikeda et al., 1999 ; Inouye et al., 1994 ). If the grouping is significant then it may be used to predict that Mtf3 and Mtf4 are possible rhamnosyl methyltransferases.

In this study, we provide direct evidence that the M. smegmatis mtf2 gene encodes an O-methyltransferase that modifies the hydroxy group of the fatty acids rather than rhamnose. An mtf2 mutant was made by targeted gene replacement. The mutant, Myco493, synthesized GPLs with the same glycopeptide headgroup as the wild-type mc²155 strain. However, HPTLC and ESI-MS analyses revealed that the mutant lacked GPL molecular species with O-methylated fatty acids. Complementation of the Myco493 mutant with mtf2 restored the synthesis of these molecular species. Moreover, overexpression of mtf2 in wild-type bacteria resulted in the increased expression of these molecular species. Collectively, these data suggest that mtf2 encodes an O-methyltransferase that specifically modifies the GPL amide-linked fatty acid. This reaction presumably occurs early in GPL biosynthesis.

We thank Dr Deidreia Tull for assistance with the ESI-MS. This work was supported by a NH&MRC Program Grant and the WHO Special Programme for Research and Training in Tropical Diseases. M.J.M. is a NH&MRC Principal Research Fellow and Howard Hughes International Research Fellow.