MtrP, a putative methyltransferase in Corynebacteria, is required for optimal membrane transport of trehalose mycolates

Pathogenic bacteria of the genera Mycobacterium and Corynebacterium cause severe human diseases such as tuberculosis (Mycobacterium tuberculosis) and diphtheria (Corynebacterium diphtheriae). The cells of these species are surrounded by protective cell walls rich in long-chain mycolic acids. These fatty acids are conjugated to the disaccharide trehalose on the cytoplasmic side of the bacterial cell membrane. They are then transported across the membrane to the periplasm where they act as donors for other reactions. We have previously shown that transient acetylation of the glycolipid trehalose monohydroxycorynomycolate (hTMCM) enables its efficient transport to the periplasm in Corynebacterium glutamicum and that acetylation is mediated by the membrane protein TmaT. Here, we show that a putative methyltransferase, encoded at the same genetic locus as TmaT, is also required for optimal hTMCM transport. Deletion of the C. glutamicum gene NCgl2764 (Rv0224c in M. tuberculosis) abolished acetyltrehalose monocorynomycolate (AcTMCM) synthesis, leading to accumulation of hTMCM in the inner membrane and delaying its conversion to trehalose dihydroxycorynomycolate (h2TDCM). Complementation with NCgl2764 normalized turnover of hTMCM to h2TDCM. In contrast, complementation with NCgl2764 derivatives mutated at residues essential for methyltransferase activity failed to rectify the defect, suggesting that NCgl2764/Rv0224c encodes a methyltransferase, designated here as MtrP. Comprehensive analyses of the individual mtrP and tmaT mutants and of a double mutant revealed strikingly similar changes across several lipid classes compared with WT bacteria. These findings indicate that both MtrP and TmaT have nonredundant roles in regulating AcTMCM synthesis, revealing additional complexity in the regulation of trehalose mycolate transport in the Corynebacterineae.

The synthetic pathways and enzymatic machineries that produce these species have been extensively studied and reviewed (4,(8)(9)(10)(11). In Mycobacteria, two independently-synthesized fatty acids, the short ␣ branch (C 24 -C 26 ) and long meromycolate chain (C 50 -C 60 ), produced by the single FAS-I enzyme and FAS-II complex (12,13), respectively, are modified and then condensed by the cytoplasmic polyketide synthase, Pks13 (14), to form a 2-alkyl 3-keto fatty acid that is then reduced by CmrA to generate the mature, hydroxylated mycolic acid (15). These fatty acids are conjugated to trehalose to form TMM, which is then bound and flipped to the periplasm by the integral membrane lipid transporter MmpL3 (16,17), the target of several classes of new antimycobacterial inhibitors (18,19). Whereas structurally-diverse classes of MmpL3 inhibitors have been reported, several of these compounds were subsequently found to indirectly inhibit MmpL3 by disrupting the transmembrane electrochemical proton gradient (20). However, recent studies have identified two compounds, BM212 (a 1,5-diarylpyrrole compound) and PIPD1 (a piperidinol-containing molecule), that inhibit MmpL3 via direct interactions, advancing our understanding of the mechanism of action of this transporter (16,21). Periplasmic mycolyltransferases of the Ag85ABC complex (22) subsequently utilize TMM in the periplasmic space as the mycolic acid donor for the synthesis of the mycolated-AG layer, trehalose dimycolate (TDM), or mycolated glucose or glycerol.
Most cell wall components are essential for the growth of pathogenic mycobacteria, hampering efforts to functionally characterize genes involved in cell wall assembly. In contrast, Corynebacteria such as Corynebacterium glutamicum retain viability after loss of conserved cell wall components and are thus excellent model systems for investigating cell wall synthesis in Corynebacterineae (15,(23)(24)(25)(26)(27)(28). Recently, we identified a new trehalose monocorynomycolate (TMCM) glycolipid species in C. glutamicum, which contains an acetyl group on the corynomycolate chain. This species, termed acetyl-trehalose monocorynomycolate (AcTMCM), required expression of the putative acetyltransferase, TmaT (NCgl2759) (Fig. 1). Genetic deletion of tmaT significantly reduced the rate of corynomycolate transport across the inner membrane and subsequent conversion to h2TDCM, indicating that AcTMCM is the major substrate transported by C. glutamicum CmpL1/4 (NCgl0228/ NCgl2769) (Fig. 1) (28).
Here, we show that optimal transport of hTMCM is also regulated by a methyltransferase. Specifically, we show that disruption of a putative methyltransferase gene, NCgl2764, which is located in the tmaT locus, also leads to the accumulation of hTMCM on the cytoplasmic site of the inner membrane and defects in the synthesis of h2TDCM. Comprehensive lipidomic analysis of WT and mutant C. glutamicum (29) suggested that NCgl2764 is required for synthesis of AcTMCM. The mutant also accumulated trehalose monoketocorynomycolates (ketoTMCM), had supressed synthesis of phosphatidylglycerol (PG), and alanylated diacylglycerols (Ala-DAG), and lacked a class of acyl-PG-like lipids, mirroring the general lipid phenotype of a tmaT mutant. Mutation analysis provided further support for NCgl2764 having methyltransferase activity, revealing unexpected complexity in the regulation of corynomycolate transport across the inner membrane of these bacteria.

Identification of the corynebacterial ortholog of M. tuberculosis Rv0224c
We have previously identified a genetic locus in C. glutamicum that harbors a number of genes involved in regulating cell wall synthesis. This locus is conserved in all Corynebacterineae (15,25,30) and contains NCgl2759 (tmaT, Rv0228) (28), NCgl2760 (Rv0227c) (31), and NCgl2764 (Rv0224c) (Fig. 2). NCgl2764 is functionally uncharacterized but is predicted to encode a protein of 257 residues, sharing 49% identity and 86% similarity with M. tuberculosis Rv0224c (Fig. S1). NCgl2764/ Rv0224c is annotated as a cytoplasmic methyltransferase based on weak similarity with a range of methyltransferases and the presence of conserved residues essential for activity (see below).

Loss of the NCgl2764 gene affects growth of C. glutamicum
To investigate the function of this gene/protein, a two-step recombination strategy was used to create the NCgl2764 mutant (Fig. 3). Sequences flanking the gene were PCR-amplified and cloned into the suicide vector pK18mobsacB (32), which carries a kanamycin resistance gene (aph) and Bacillus subtilis sacB gene conferring sensitivity to sucrose. This plasmid was introduced into C. glutamicum 13032, and a NCgl2764 deletion mutant, designated ⌬NCgl2764, was isolated, and gene deletion was confirmed by PCR analysis (Fig. 3, A and B). The ⌬NCgl2764 mutant formed small colonies relative to the wildtype (WT) parent on agar plates and displayed an extended lag-phase in liquid BHI medium, although it ultimately achieved a final cell density comparable with the WT parent strain (Fig. 3C). A similar growth phenotype was observed following loss of tmaT (i.e. ⌬NCgl2759) but not other genes (i.e. ⌬NCgl2760) in this locus (28,31).

Complementation of C. glutamicum ⌬NCgl2764
Complementation and control strains were created to determine whether the growth defect was due solely to disruption of NCgl2764. To complement the mutant, ⌬NCgl2764 was transformed with pSM22:NCgl2764, a plasmid carrying a full-length NCgl2764 gene plus 203 bp of upstream sequence. The empty pSM22 vector (32) was also introduced into ⌬NCgl2764 as a control. Growth curves indicated that introduction of pSM22: NCgl2764 into the ⌬NCgl2764 mutant restored normal lag phase and growth, whereas the empty pSM22 vector did not (Fig. 3C).

Antibiotic susceptibility testing
Corynebacterial mutants with defects in cell wall synthesis and integrity characteristically display increased sensitivity to antibiotics. To test the condition and permeability of the cell wall in the ⌬NCgl2764 mutant, the strains were cultured on agar media and exposed to paper discs impregnated with a vari-Putative cell wall methyltransferase Figure 2. NCgl2764 is encoded within a genetic locus highly conserved in Corynebacterineae. The tmaT locus of Corynebacterineae is shown. Likely ortholog genes in the four species are shown using the same color. The focus of this study NCgl2764 (boxed) and its likely mycobacterial orthologs are shown in bold. MSMEG_0310 is reported to be a pseudogene. Previously studied genes are NCgl2759 (tmaT (28)) and NCgl2760 (31), whereas the remaining genes remain uncharacterized. Products of the FAS-I and FAS-II pathways are condensed and then attached to trehalose, producing keto-TMCM in the inner leaflet of the IM. Reduction by CmrA produces hTMCM, which may be acetylated to AcTMCM by the integral membrane acetyltransferase TmaT, facilitating transport by CmpL1/4. It is hypothesized that an unidentified deacetylase removes the acetyl group after transport, reconstituting hTMCM in the outer leaflet of the IM. Trehalose corynomycolyl transferase then produce h2TDCM from two hTMCM molecules or transfer the corynomycolates to acceptors on AG. Other acylation variants of hTMCM, AcTMCM, and hTDCM have been identified (29) but are not shown here for clarity. Results of this study place NCgl2764 (MtrP) at the same step as TmaT in the pathway.

Putative cell wall methyltransferase
ety of antibiotics. The sensitivity of the strains was assessed by the size of zones of inhibition (ZOI) around the discs. For lowmolecular-weight antibiotics (Ͻ500 g/mol), the ⌬NCgl2764 mutant displayed significantly larger ZOIs than the WT parent (p Ͻ 0.05) and complemented strains (Fig. 4). The increases in zone diameter were particularly apparent for the ␤-lactam class antibiotics amoxicillin, cefoxitin, and imipenem, suggesting increased access of the drugs to the peptidoglycan layer. Increased sensitivity to drugs targeting cytoplasmic processes (RNA and protein synthesis) was also observed. Clustering of the NCgl2764 gene with cell wall biosynthesis genes, together with the altered growth rate and increased antibiotic sensitivity in the ⌬NCgl2764 strain, suggested a role for the encoded protein in cell wall biosynthesis.

Loss of NCgl2764 affects synthesis of trehalose corynomycolates
HPTLC analysis of lipid extracts of WT and ⌬NCgl2764 bacteria revealed major differences in the abundance of several different cell wall lipid classes (Fig. 5A). In particular, the ⌬NCgl2764 mutant accumulated a lipid species during logarithmic growth (15,28) that comigrated with hTMCM on HPTLC and was confirmed to be this glycolipid by MALDI-TOF-MS (Fig. S2). The accumulation of hTMCM in ⌬NCgl2764 was associated with a concomitant decrease in levels of h2TDCM, relative to WT (Fig. 5A). Levels of both species were restored following complementation of the mutant with pSM22:NCgl2764, but not with empty pSM22. These data suggest that NCgl2764 is required for efficient conversion of hTMCM to h2TDCM.
To determine whether the M. tuberculosis ortholog, Rv0224c, could functionally complement the mutant, we constructed a pSM22:Rv0224c plasmid and introduced it into the ⌬NCgl2764 mutant. hTMCM and h2TDCM levels in the Rv0224c complemented line were unchanged compared with the original mutant (Fig. S3), suggesting that Rv0224c is unable to substitute for NCgl2764. Although the two proteins share significant sequence similarity, including two glycine residues essential for methyltransferase function (see below and Fig. S1), the Rv0224c homolog may lack sequence motifs required for protein-protein interactions essential for activity in C. glutamicum. C, growth curves of WT C. glutamicum, the ⌬NCgl2764 mutant, and complementation strains in liquid BHI medium. Each strain was grown to saturation and then diluted 1:100 in fresh BHI medium. Triplicate cultures were sampled at the times indicated to determine the optical density (OD) at a wavelength of 600 nm. The curves for the WT and ⌬NCgl2764 ϩ pSM22:NCgl2764 (complementation) strains overlap, as do the ⌬NCgl2764 and ⌬NCgl2764 ϩ empty pSM22 strains, indicating that adding back the NCgl2764 gene reverses the mild growth defect in the mutant.

Putative cell wall methyltransferase Labeling studies
To measure rates of glycolipid turnover in WT and mutant lines, bacteria were pulse-labeled with [ 14 C]acetate, then suspended in BHI medium (chase), and sampled at various time points. Cell wall lipids were then separated by HPTLC, and labeled species were detected by autoradiography (Fig. 5B). Label was rapidly incorporated into hTMCM and phosphatidylglycerol (PG) species during the pulse and subsequently into h2TDCM in the 5-min chase in WT cells. In contrast, the label was not incorporated into h2TDCM in the ⌬NCgl2764 strain throughout the 60-min chase period, demonstrating a significant delay in its synthesis. Interestingly, in contrast to the orcinol-stained samples (Fig. 5A), no accumulation of [ 14 C]hTMCM was observed in the ⌬NCgl2764 mutant, implying that the high steady-state levels of hTMCM ( Fig. 5A) reflect reduced rate of turnover of hTMCM rather than excess production of this glycolipid. Interestingly, incorporation of 14 C-label into PG lipids was also reduced in the ⌬NCgl2764 mutant, similar to the phenotype of the tmaT mutant (29) and consistent with the lipidomics analysis (see below).

Altered dynamics of hTMCM turnover and h2TDCM synthesis in a ⌬NCgl2764 mutant
To investigate the dynamics of hTMCM conversion to h2TDCM in the mutant, glycolipids were extracted at specific growth stages and analyzed by HPTLC. Cells were cultured in BHI and sampled at OD 600 0.5-10 after 24 h (stationary phase). Samples were extracted and normalized by pellet wet weight prior to HPTLC analysis (Fig. 5C). In the WT and complemented ⌬NCgl2764 mutant, h2TDCM levels steadily increased from early log phase through to the stationary phase, peaking by OD 600 ϳ4. hTMCM levels also peaked at OD 600 ϳ4 and then reduced to be undetectable by stationary phase. In the ⌬NCgl2764 mutant, an accumulation of hTMCM was observed by OD 600 ϳ2 (early log phase) before reducing at OD 600 ϳ7, and these species were still detectable in stationary phase. h2TDCM synthesis was delayed in the mutant, not reaching maximal levels until OD 600 ϳ7. These findings suggested that synthesis of h2TDCM from hTMCM is delayed in the ⌬NCgl2764 strain, consistent with a defect in the rate of transport of hTMCM to the periplasmic space in logarithmic phase.
Because hTMCM is known to be the donor of corynomycolates for formation of the AG-linked corynomycolate outer membrane, we also extracted and analyzed covalently-bound corynomycolates from the delipidated pellets. No significant differences were observed between WT and the mutant at the various time points (Fig. S4), suggesting that the limited pool of transported corynomycolates are channeled to AG at the expense of h2TDCM synthesis. Unaltered levels of AG-linked corynomycolates were also observed in a tmaT mutant (28).

Global changes in the cell envelope composition in the C. glutamicum NCgl2764 mutant
To further investigate whether hTMCM does indeed accumulate on the cytoplasmic leaflet of the IM in the ⌬NCgl2764 mutant, IM and OM lipid extracts of the WT and mutant strains were analyzed by LC-ESI-QTOF-MS in a positive ionization mode (29). Lipid species identified by MS/MS from WT and the ⌬NCgl2764 mutant are shown in Table 1. After removal of isotopologs, dimer species, isomers, and isobars, 139 and 111 lipid species were identified with high confidence in the IM and OM of the NCgl2764 mutant, respectively. Compared with WT bacteria, the IM fraction of the NCgl2764 mutant contained elevated levels of hTMCM, ketoTMCM, and hGMM species. Conversely, levels of PG and Ala-DAG species were decreased, whereas AcTMCM and acyl-PG like species were either absent or present at very low levels. These changes were also observed in the tmaT (NCgl2759) mutant lipidome (Table 1) (29).
As NCgl2764 encodes a putative methyltransferase, we looked for evidence that major TMCM species might be meth- ϩ ) and 373 Da (two hexoses with one being methylated, with NH 4 ϩ ), respectively, instead of 197 and 359 Da. However, neither of these molecular or fragment ions were detected, suggesting that TMCM is not directly methylated or that only a very small proportion of the steady-state pool of TMCM carries this modification. Changes in the relative abundance of individual lipid molecular species in WT and mutant lines were verified by quantitation of the relative ion (lipid) intensities within four replicate samples of each line. Lipids identified in the MS1 analysis are summarized in Fig. 6, A-F. Both MS1 and MS/MS results confirm the global lipidome changes and an overall shared phenotype between the NCgl2764 and tmaT (29) mutants.
Collectively, these findings suggest that NCgl2764 is required, together with TmaT, for the synthesis of AcTMCM.
To investigate whether these genes may function at the same step, we constructed a double mutant by deleting the NCgl2764 gene in the tmaT mutant. The resultant strain grew similarly to the individual mutants (data not shown), and analysis of cell wall glycolipids by HPTLC with orcinol staining revealed an identical lipid profile to the individual NCgl2764 and NCgl2759 knockout lines (Fig. 7, lane 6). Specifically, the double mutant exhibited a significant accumulation of hTMCM at the expense of h2TDCM and reduced synthesis of PG lipids. These data strongly suggest that NCgl2764 and TmaT regulate the same step or pathway in these bacteria.

Lipoglycan analysis of a NCgl2764 mutant
Our previous studies on genes of this locus revealed that one of them, NCgl2760, was involved in lipoglycan biosynthesis (31). To determine whether loss of NCgl2764 also impacts the synthesis of other cell wall components, lipomannan (LM) and lipoarabinomannan (LAM) were extracted from delipidated cell pellets and analyzed by PAGE and LC-TOF-MS. PAGE analysis suggested that the overall abundance and degree of In contrast, no labeled h2TDCM was detected in the ⌬NCgl2764 samples even after 60 min, consistent with a significantly slower rate of h2TDCM synthesis in this strain. C, delayed conversion of hTMCM to h2TDCM in a ⌬NCgl2764 mutant. Glycolipids were extracted from cells harvested at OD 600 of 0.5, 2, 4, 7, and 10 and from a stationary phase culture (ϳ24 h growth). HPTLC analysis of glycolipids showed that the accumulation of hTMCM peaks during mid-log growth phase in the analyzed strains. In samples from the ⌬NCgl2764 strain, hTMCMs accumulate to a higher level and do so earlier than in the WT or the complemented mutant. The dynamics of h2TDCM synthesis are inverted relative to hTMCM and reach their highest point in the stationary phase where no hTMCM is detected. The identities of glycolipids were based on previously published reports. Gl-X, mannosyl-glucuronic acid diacylglycerol; AcPIM2, Man 2 -acyl-PI.

Putative cell wall methyltransferase
heterogeneity of these lipoglycans were similar in both the WT and ⌬NCgl2764 mutant (Fig. S5). Similarly, LC-ESI-TOF-MS analysis using a newly-developed profiling method also indicated minimal change in the LM/LAM structure in WT and the ⌬NCgl2764 mutant bacteria. The latter approach allows detection of LM and LAM molecular species that have been assembled on different lipid anchors. In WT bacteria, the glycan chains of detected LM species contain between 1 and 38 mannose residues, and six different lipid anchors, comprising PI with 2-3 acyl chains (53 molecular species total, 998 -6997 Da after deconvolution; Fig. S6). Similarly, the LM from the ⌬NCgl2764 mutant contained between 1 and 44 mannose residues and six lipid anchors (61 molecular species, 998 and 8002 Da). Minor differences were observed between WT and mutant bacteria in the degree of polymerization of one of the lipid anchor platforms. The plots in Fig. S6 show the masses (m/z, after deconvolution) of all detected LM/LAM species, from low to high molecular weight, and the predicted mannan chain length. Fig. S5 of our previous study (31) explains, step by step, the development of Fig. S6.

Mutagenesis of NCgl2764
To investigate whether the putative methyltransferase activity of NCgl2764/Rv0224c is functionally important, the ⌬NCgl2764 mutant was complemented with plasmids encoding NCgl2764 proteins in which specific amino acids in the methyltransferase catalytic site had been mutated. The amino acid sequences of NCgl2764 and its mycobacterial orthologs were compared with two confirmed methyltransferases, MT0146/CbiT and protein arginine methyltransferase 1 (PRMT1 (33)), for which crystal structures have been solved and active-site residues identified (34,35). Two glycine residues are conserved in nearly all analyzed S-adenosylmethionine (SAM)-dependent methyltransferase sequences and are known to be involved in methyl donor binding (35). These correspond to Gly-69 and Gly-71 in NCgl2764, which are also conserved in Rv0224c and ML2584 (Fig. S1). We hypothesized that substitution of either of those crucial glycine residues with alanine would result in loss of a methyltransferase function, if present, and failure to complement the ⌬NCgl2764 phenotype.
The desired mutations were engineered into the pSM22: NCgl2764 plasmid previously used to complement the ⌬NCgl2764 strain (see above). Transformation of the deletion mutant with the pSM22:NCgl2764_G69A or pSM22: NCgl2764_G71A constructs, but not the unmodified pSM22: NCgl2764 plasmid encoding the native protein, resulted in strains with the delayed growth phenotype (data not shown) and with glycolipid profiles identical to ⌬NCgl2764 (Fig. 7,  lanes 1-5). This finding is in agreement with the annotation of NCgl2764 as a methyltransferase, although we cannot discount the possibility that NCgl2764 may have other functions.

Discussion
We have previously identified a conserved gene locus in mycobacteria and corynebacteria that is required for cell wall assembly (28,29,31). Most of the genes in this locus are essential in M. tuberculosis, hampering their functional characterization in this pathogen (36,37). In contrast, these genes are not essential in C. glutamicum, reflecting the capacity of this bacterium to tolerate major defects in cell wall synthesis. One of the first genes characterized in this locus, C. glutamicum NCgl2759 (ortholog of M. tuberculosis Rv0228, designated tmaT), encodes an acetyltransferase that acetylates the mycolic acid moiety of hTMCM, facilitating transport of this glycolipid across the inner membrane (28), presumptively by the MmpL3 transporter. Consistent with these findings, TmaT was recently found to be the only protein of the mycolic acid pathway to interact with MmpL3 in a two-hybrid screen (38). It remains unclear whether acetylation is required for recruitment of AcT-MCM to the membrane domains containing the transporter and/or for transporter recognition and transmembrane movement. The next gene in this locus, NCgl2760 (Rv0227c), has been shown to be required for synthesis of mature cell wall lipoglycans LM and LAM (31), and surprisingly, Rv0227c also interacts with MmpL3 (38). Although LM/LAM biosynthesis lacks any shared enzymes with the hTMCM pathway, both pathways are likely to be coordinately regulated to ensure balanced synthesis of major cell wall components.
In this study, we show that NCgl2764 is also involved in cell wall synthesis. The encoded protein, denoted MtrP, shows similarity to protein methyltransferases. Phenotypic characterization of the ⌬NCgl2764 deletion mutant indicated that this protein is involved in trehalose corynomycolate metabolism, rather than LM/LAM synthesis. The mutant was found to share striking cell wall similarities with the ⌬tmaT mutant previously characterized by our group. Both mutants failed to produce AcTMCM, accumulated hTMCM, and ketoTMCM in the IM fraction and exhibited markedly reduced kinetics of synthesis of h2TDCM (28). This phenotype is consistent with reduced Table 1 Detected

Lipid (sub) class
Wildtype ⌬NCgl2759 ⌬NCgl2764   IM  OM  IM  OM  IM  OM   1  hTMCM  18  20  26  15  23  18  2  AcTMCM  3  8  0  0  0  0  3  keto-TMCM  3  0  10  3  8  5  4 Acyl-hTMCM 7 5 12 5 9 7 5 Acyl- AcTMCM  1  0  0  0  0  0  6  h2TDCM  28  28  30  25  27  29  7  Ac1-hTDCM  3  3  3  1  0  0  8  hGMM  5  Putative cell wall methyltransferase transport of hTMCM across the inner membrane by C. glutamicum transporters CmpL4 (NCgl0228) and CmpL1 (NCgl2769). The latter has significant sequence similarity to mycobacterial TMM transporter MmpL3, the direct or indirect target of several molecules with antimycobacterial activity (16, 20, 21, 39 -41). Deletion of both NCgl2764 and tmaT resulted in a phenotype that was essentially identical to that of the single gene knockout lines, supporting the notion that both proteins act sequentially or on the same step within the same pathway. In addition to the reduced conversion of hTMCM to h2TDCM in the ⌬NCgl2764 strain, a reduced synthesis of PG and Ala-DAG species and the complete absence of acyl-PGlike lipids were observed. Again, this lipid profile was also observed in our ⌬tmaT mutant (29) and likely reflects global changes in the synthesis of other cell wall components to compensate for the accumulation of hTMCM/ketoTMCM in the IM and loss of TDCM in the OM. Changes in LM mannan chain length were also detected in the NCgl2764 mutant but were minor relative to the significant changes in other lipid classes and are probably also a secondary effect.
NCgl2764 and its mycobacterial orthologs are annotated as possible methyltransferases based on sequence similarity with a range of SAM-dependent enzymes. Mutagenesis of two glycine residues equivalent to those in PRMT1 known to bind the methyl donor resulted in loss of NCgl2764 activity. Although these analyses support the annotation of NCgl2764 as a methyltransferase, further studies are needed to confirm enzyme activity. We have been unable to find any evidence that TMCM lipids are methylated in WT bacteria, suggesting that NCgl2764 may be involved in regulating the activity of TmaT. This conclusion is supported by the finding that loss of NCgl2764 phenocopies loss of NCgl2759 (TmaT) and the double knockout line. NCgl2764 could either directly methylate TmaT and/or regulate its function through protein-protein interactions. Protein methylation has been shown to modulate the activity of histones, leading to epigenetic effects through repression or

Putative cell wall methyltransferase
activation of gene expression (42). Interestingly, histones can also be acetylated (43), and the histone acetyltransferases responsible are themselves often methylated, usually at arginine residues within glycine/arginine-rich "GAR" sequences (44,45). Alternatively, MtrP and TmaT may form heterodimeric or larger protein complexes. Disruption of one protein could destabilize the other, leading to the similar lipid phenotypes of our mutants. Further studies are required to explore this hypothesis and clarify the relationship between these proteins.
The MtrP ortholog in M. tuberculosis H37Rv, Rv0224c, is reported to be essential for in vitro growth based on several high-density mutagenesis and sequencing studies (36,37,46). Although the cell wall defect in the NCgl2764 mutant was not reversed by introduction of a plasmid carrying the Rv0224c gene, the sequence similarity between the two proteins would suggest that Rv0224c may perform a similar function in this pathogen, making it a potential target for studies aimed at developing new antimycobacterial agents or at making bacterial strains more sensitive to existing agents, a phenotype seen in our mutant.

Bacterial strains and culturing conditions
Escherichia coli DH5␣ was grown in Luria-Bertani (LB) medium at 37°C with aeration. C. glutamicum ATCC 13032 was grown in brain heart infusion (BHI) medium (Oxoid) or LBHIS (LB, BHI, sorbitol) at 30°C with aeration. When necessary, ampicillin was added to a final concentration of 100 g ml Ϫ1 and kanamycin at 50 g ml Ϫ1 .

Genetic manipulation of bacteria
E. coli plasmid DNA was isolated using the High Pure plasmid isolation kit (Roche Applied Science), and C. glutamicum genomic DNA was extracted using the Illustra DNA extraction kit (GE Healthcare), according to the manufacturer's instructions. DNA manipulations and molecular biology techniques were performed as described (28).

Bioinformatic identification and analysis of NCgl2764
The corynebacterial ortholog of M. tuberculosis Rv0224c was found using the BLASTp (47) algorithm. Mycobacterial DNA and protein sequences were obtained from Mycobrowser (RRID:SCR_018242). Amino acid sequence alignments were generated using Clustal Omega (RRID:SCR_001591).

Extraction of covalently-bound corynomycolates
Dried, delipidated cell pellets were base-treated to cleave AG-bound corynomycolates. Pellets were suspended in 2 ml of 0.1 M KOH in ethanol (30°C, 3 h), and samples were neutralized with 0.5 M acetic acid prior to centrifugation at 3000 ϫ g for 10 min. Cell pellets were dried by evaporation using a Savant SpeedVac Plus SC11A concentrator and then resuspended by sonication in 400 l of water-saturated 1-butanol, and mycolates were recovered after phase partitioning in 1-butanol/water (2:1 v/v) and collection of the upper 1-butanol layer.

Pulse-chase radiolabeling
Bacterial strains were grown in BHI to mid-log phase (OD 600 ϳ5) at which time the cells were harvested by a 10-min centrifugation at 13,000 ϫ g, washed with HEPES-saline (pH 7.4), re-pelleted, and weighed. Pellets were resuspended in HEPESsaline to 0.25 g per ml and incubated for 5 min. Aliquots (200 l) were transferred to new tubes, and 2 Ci of [ 14 C]acetate was added to each sample (except for the "cold" control) followed by incubation for 5 min. The "pulse" sample was then collected and harvested by brief centrifugation and resuspension of the pellet in 500 l of chloroform/methanol (2:1 v/v). The "chase" samples were pelleted, and the medium was replaced with prewarmed BHI, followed by incubation at 30°C for the designated chase time. Chase samples were collected after 5, 15, 30, and 60 min and harvested by brief centrifugation and resuspension of the pellet in 500 l of chloroform/methanol (2:1 v/v). Following a 1-h extraction, the supernatants were collected by a 5-min centrifugation at 4000 rpm, and the pellets were further extracted for 1 h with 500 l of chloroform/methanol/water (1:2:0.8 v/v). The glycolipid-containing supernatants were collected by a 5-min centrifugation at 4000 rpm and dried under a N 2 stream. The samples were then suspended in 200 l of water-saturated 1-butanol and 100 l of butanol-saturated water, and the butanol phase was collected and dried after vigorous vortexing. The sample was then suspended in 20 l of 1-butanol, and 2 l was applied to a HPTLC plate and developed as described above.

Extraction and analysis of lipoglycans
LM and LAM were extracted and purified as described previously (31), separated by PAGE, and stained using a SilverSnap kit (Pierce). Alternatively, purified LM/LAM was suspended in 10 mM ammonium carbonate/bicarbonate buffer (pH 8.16), and 4 l was injected by direct infusion into an Agilent 6220 Accurate-Mass TOF-LC/MS system (Agilent Technologies). The run was performed in negative ionization mode, with a mass range of 100 -3000 Da. The reference nebulizer was set to 20 p.s.i.g. with a detection window of 100 ppm, a minimum height of 1000 counts, an acquisition rate of 0.63 spectra/s, and an acquisition time of 1589.4 ms/spectrum. The gas temperature was set to 325°C with a drying gas of 7 liters/min, dual ESI 3500 V, fragmentor 150 V, and skimmer 65 V. The flow rate was set to 0.25 ml/min, and a solvent consisting of 0.1 M formic acid, acetonitrile (1:1, v/v) was used to wash the lines for 25 min. Data were analyzed using MassHunter, and the length of the mannan chain was calculated, assuming the presence of a PI lipid anchor containing C16:1/18:1/16:0 fatty acids. Analyses were performed in triplicate.