Lipoarabinomannan biosynthesis in Corynebacterineae: the interplay of two α(1→2)-mannopyranosyltransferases MptC and MptD in mannan branching

Lipomannan (LM) and lipoarabinomannan (LAM) are key Corynebacterineae glycoconjugates that are integral components of the mycobacterial cell wall, and are potent immunomodulators during infection. LAM is a complex heteropolysaccharide synthesized by an array of essential glycosyltransferase family C (GT-C) members, which represent potential drug targets. Herein, we have identified and characterized two open reading frames from Corynebacterium glutamicum that encode for putative GT-Cs. Deletion of NCgl2100 and NCgl2097 in C. glutamicum demonstrated their role in the biosynthesis of the branching α(1→2)-Manp residues found in LM and LAM. In addition, utilizing a chemically defined nonasaccharide acceptor, azidoethyl 6-O-benzyl-α-D-mannopyranosyl-(1→6)-[α-D-mannopyranosyl-(1→6)]7-D-mannopyranoside, and the glycosyl donor C50-polyprenol-phosphate-[14C]-mannose with membranes prepared from different C. glutamicum mutant strains, we have shown that both NCgl2100 and NCgl2097 encode for novel α(1→2)-mannopyranosyltransferases, which we have termed MptC and MptD respectively. Complementation studies and in vitro assays also identified Rv2181 as a homologue of Cg-MptC in Mycobacterium tuberculosis. Finally, we investigated the ability of LM and LAM from C. glutamicum, and C. glutamicumΔmptC and C. glutamicumΔmptD mutants, to activate Toll-like receptor 2. Overall, our study enhances our understanding of complex lipoglycan biosynthesis in Corynebacterineae and sheds further light on the structural and functional relationship of these classes of polysaccharides.


Introduction
The Corynebacterineae, a sub-order of the Actinobacteria, represent an unusual group within Gram-positive bacteria, with a distinctive cell wall architecture. Prominent members are the human pathogens, Corynebacterium diphtheriae, Mycobacterium tuberculosis and Mycobacterium leprae (Bloom and Murray, 1992). Typically, the cell walls of Corynebacterineae contain mycolic acids (m), arabinogalactan (AG) and peptidoglycan (P), which are covalently linked to each other to form the mycolylarabinogalactan-peptidoglycan (mAGP) complex McNeil et al., 1990;Besra et al., 1995;Brennan, 2003;Dover et al., 2004). In addition, specialized glycophospholipids, phosphatidyl-myoinositol mannosides (PIMs) and lipoglycans, lipomannan (LM) and lipoarabinomannan (LAM), are found in the outer leaflet of mAGP (Hill and Ballou, 1966;Brennan and Ballou, 1967;1968;Brennan and Nikaido, 1995;Morita et al., 2004). The non-pathogenic bacillus, Corynebacterium glutamicum, also belongs to this class of bacteria, and is widely used in the industrial production of amino acids (Eggeling and Bott, 2005). Use of this easily cultivatable bacterium, together with the fact that the cell walls of Corynebacterineae share identical basic structures and building blocks, have contributed significantly to decipher the biosynthesis of their complex cell walls. For instance, in both C. glutamicum and M. tuberculosis the orthologous acyl-CoA carboxylase genes (Gande et al., 2007) together with Pks (Gande et al., 2004), are key to mycolic acid biosynthesis, and the glycosyltransferases AftA (Alderwick et al., 2006a), and AftB (Seidel et al., 2007a) introduce specific arabinofuranose (Araf) residues into AG, and AftC in both AG and LAM of Mycobacterium smegmatis (Birch et al., 2008;. A speciation-specific difference is the presence of the glycosyltransferase RptA in C. glutamicum that introduces decorating rhamnose residues into AG (Birch et al., 2009), whereas a structurally very similar protein introduces galactosamine residues into AG of M. tuberculosis (Skovierova et al., 2010).
A particular interesting group of cell wall components are the lipoglycans LM and LAM that are based on PIMs. They play poorly defined roles in Corynebacterineae but mycobacterial LAM has been implicated in many of the key aspects of the pathogenesis of tuberculosis and leprosy, such as induction of phagocytosis, phagosomal alteration and acquired T cell-mediated immunity (Briken et al., 2004). Mycobacterial LAM has attached to the PIM base, an elongated a(1→6) linear, a(1→2) branched mannan 'core', of approximately 30 mannopyranose (Manp) residues, and linked to its terminus a branched D-arabinan domain. In mycobacteria, this large arabinan domain consists of approximately 70 arabinofuranose (Araf) residues and is assembled through several distinct motifs (Chatterjee et al., 1991;Alderwick et al., 2007;Birch et al., 2010), formed via specific a(1→5), a(1→3) and b(1→2) arabinofuranosyltransferases Zhang et al., 2003;Alderwick et al., 2005;Birch et al., 2010), and can be capped to various degrees with either short a(1→2) mannopyranosyl chains, as is the case in M. tuberculosis (Chatterjee et al., 1992), or in M. smegmatis by inositol phosphate , or completely devoid of caps as found in Mycobacterium chelonae (Guerardel et al., 2002). There is very limited structural information on LAM in other Corynebacterineae, except for C. glutamicum (Tatituri et al., 2007a,b). In this organism the core a(1→6) linear a(1→2) branched LM (see Fig. 9), which is structurally akin to mycobacterial LM, is substituted by t-Araf residues to form a modified LAM-like molecule (Tatituri et al., 2007a,b).
The current model of lipoglycan biosynthesis supported by biochemical and genetic studies, follows a linear pathway from PI→PIM 2→LM→LAM . As with AG synthesis selected enzymatic and structural features are in part shared within the Corynebacterineae. Glycosylation of phosphatidylmyo-inositol (PI) by different a-mannopyranosyltransferases and acylation by acyltransferase(s) results in the synthesis of mono-and di-acylated PIMs (Kordulakova et al., 2002;2003;Kremer et al., 2002;Morita et al., 2006;Lea-Smith et al., 2008;Mishra et al., 2009). Previ-ous studies have identified the enzymes involved in the early and late stages of LM and LAM biosynthesis in Corynebacterineae. For instance, NCgl1505/Rv1459c (MptB) and NCgl2093/Rv2174 (MptA), synthesize the proximal and distal a(1→6) mannan backbone in LM and LAM (Kaur et al., 2007;Mishra et al., 2007;2008a). However, the enzymatic steps involved in the biosynthesis of mannan branching of LM and LAM, still remain to be studied in detail in Corynebacterineae (Kaur et al., 2006;Sena et al., 2010).
In the current study we have examined two open reading frames from C. glutamicum, which encode for putative glycosyltransferase family C (GT-C) members (Liu and Mushegian, 2003). On the basis of mutant studies, chemical and enzymatic data we report that NCgl2100 (Cg-MptC) and its M. tuberculosis homologue Rv2181 function as a(1→2)-mannopyranosyltransferases, and NCgl2097 functions as a second a(1→2)mannopyranosyltransferase (Cg-MptD), both of which are required for complete mannose-branching found in LM and LAM in Corynebacterineae.
The Cg-mptC encoded protein is a long hydrophobic polytopic membrane protein of 812 amino acid (aa) residues. The first half of the protein (1-417 aa), shares 37% identity (55% similarity) with Mt-MptC and 38% identity (57% similarity) to Cg-MptD, whereas the second half (417-812 aa), has no counterpart in various Mycobacte-rium species. The protein appears to be a fusion of two membrane proteins, functional as an a(1→2)mannopyranosyltransferase as described below, and termed in this study Cg-MptC. Cg-MptD is a hydrophobic polytopic membrane protein of 436 amino acid residues with 11 transmembrane helixes (TMH). After TMH-1 and TMH-7, larger loop regions are present probably localized in the periplasm. Cg-MptC, Cg-MptD and Mt-MptC, together with Mt-PimE, an a(1→2)mannopyranosyltransferase demonstrated to be involved in PIM 6 synthesis (Morita et al., 2006), share a similar size, degree of hydrophobicity and regions of high identity (Fig. 1B), suggesting closely related functions.

Construction of deletion mutants and growth phenotype
The construct pK19mobsacBDCg-mptC was made containing 12 nucleotides (nt) of the 3′ end of NCgl2100 together with the genomic upstream sequences, and 36 nt of the 5′ end together with genomic downstream sequences (Table S1). This non-replicative vector was used to transform C. glutamicum to kanamycinresistance (Kan r ) indicating chromosomal integration. Sucrose-resistant (Suc r ) clones were selected in a second round of positive selection, indicating loss of the vector-encoded sacB function (Schäfer et al., 1994).
From 12 Kan s and Suc r clones analysed via PCR, seven had lost NCgl2100, whereas in the remaining five the genomic wild type situation was restored. One clone with deleted NCgl2100 was selected and will be referred to as C. glutamicumDmptC (Fig. S1) In an analogous manner, pK19mobsacBDCg-mptD was constructed and used to select from the wild type of C. glutamicum by double cross-over events as described above for deletion of Cg-mptD, yielding C. glutamicumDmptD (Fig. S1). Starting from C. glutamicumDmptC and using pK19mobsacBDCg-mptD the double deletion mutant C. glutamicumDmptCDmptD was also constructed (Fig. S1).
The mutants were inoculated into liquid BHI and CGXII medium and their growth patterns studied. There was no statistical difference observed in three independent cultivations for each of the strains. The growth rates in CGXII were 0.39 Ϯ 0.02 h -1 for wild-type C. glutamicum, C. glutamicumDmptC and C. glutamicumDmptD, and 0.37 Ϯ 0.02 h -1 for C. glutamicumDmptCDmptD (data not shown). The growth rates on BHI medium were 0.66 Ϯ 0.07 h -1 for all three strains and illustrate that the genes are not essential for growth as observed for other GT-Cs involved in mAGP synthesis like, Cg-emb, Cg-aftA and Cg-aftB (Alderwick et al., 2005;2006a;Seidel et al., 2007a). Purification and general characteristic features of lipoglycans from C. glutamicumDmptC and C. glutamicumDmptD Lipoglycans from wild-type C. glutamicum, C. glutamicum DmptC and C. glutamicumDmptD with complemented strains C. glutamicumDmptC pEKEx2-Cg-mptC, C. glutamicumDmptC pEKEx2-Rv2181, C. glutamicumD-mptD pEKEx2-Cg-mptD and C. glutamicumDmptD pEKEx2-Rv2181, were extracted from delipidated cells by ethanol/water extraction followed by hot-phenol water treatment and enzymatic degradation. The crude lipoglycan extract was subjected to hydrophobic interaction and gel permeation chromatography allowing the separation of Cg-LM (Cg-LM-A and Cg-LM-B) and Cg-LAM (Tatituri et al., 2007b;Lea-Smith et al., 2008;Mishra et al., 2008b). Because the mannan backbone structures of Cg-LM-A and Cg-LM-B are similar as revealed through previous matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS), and gas chromatography-mass spectrometry (GC-MS) analyses of partially methylated alditol acetates (Mishra et al., 2008b), subsequent analyses focused on the Cg-LM as a mixture. The extracted lipoglycans were examined for their size and mobility on 15% SDS-PAGE ( Fig. 2A). The crude extracts from wild-type C. glutamicum showed the presence of Cg-LM and Cg-LAM (Tatituri et al., 2007a;Lea-Smith et al., 2008;Mishra et al., 2008b) (Fig. 2A), while the crude lipoglycans from mutant strains, C. glutamicumDmptC and C. glutamicumDmptD, showed subtle differences in their migration on SDS-PAGE ( Fig. 2A). Purified Cg-LAM from C. glutamicumDmptC and C. glutamicumDmptD ( Fig. 2A) migrated faster on SDS-PAGE in comparison with those from wild-type C. glutamicum, indicating that these lipoglycans were smaller in size with low molecular weights. In addition, fractionation using gel permeation chromatography also revealed a degree of heterogeneity in Cg-LM and A. The left-hand and middle panels illustrate lipoglycans extracted from wild-type, mutant and complemented strains analysed using SDS-PAGE and visualized using a Pro-Q emerald glycoprotein stain (Invitrogen). The right-hand panel illustrates purified Cg-LAM and Cg-LM from C. glutamicumDmptC and C. glutamicumDmptD, respectively, in comparison with wild-type C. glutamicum lipoglycans. Lipoglycan profiles are represented with standard molecular weight markers of glycoproteins of 180, 82, 42 and 18 kDa. B. MALDI-TOF-MS spectra of Cg-LM and Cg-LAM from different strains of C. glutamicum. MALDI-TOF-MS spectra were acquired in the linear negative mode with delayed extraction using 2,5-dihydrobenzoic acid as a matrix.
Cg-LAM, as evident when fractions were pooled. Complementation of C. glutamicumDmptC and C. glutamicumD-mptD by pEKEx2-Cg-mptC and pEKEx2-Cg-mptD, respectively, restored the wild-type phenotype, as illustrated by the slower and co-migration on SDS-PAGE in comparison with wild-type C. glutamicum ( Fig. 2A). In contrast, cross-complementation experiments, C. glutamicumDmptC by pEKEx2-Cg-mptD, and C. glutamic-umDmptD by pEKEx2-Cg-mptC, failed to restore a wildtype phenotype and reversion to a slower migration of products on SDS-PAGE (Fig. S2), as observed in our previous experiments described above (Fig. 2A).
The molecular weights of the lipoglycans were investigated by MALDI-TOF-MS. The Cg-LAM from wild-type C. glutamicum exhibited a broad unresolved peak centred at m/z 15 000 (approximately 85 Manp/Araf glycosyl residues), and Cg-LM at m/z 5500 (approximately 25 Manp residues), indicating a molecular weight of approximately 15 and 5.5 kDa for the major molecular species of these lipoglycans (Fig. 2B). Interestingly, Cg-LAM from C. glutamicumDmptC ( Fig. 2A) peaked at m/z 13 800 (approximately 78 Manp/Araf glycosyl residues), indicating a decrease of around 1.2 kDa for the Cg-LAM isolated from C. glutamicumDmptC as compared with wild-type Cg-LAM (Fig. 2B). In addition, Cg-LM from C. glutamic-umDmptC ( Fig. 2A) exhibited a broad unresolved peak centred at m/z 4400 (approximately 20 Manp glycosyl residues), indicating a molecular weight of approximately 4.4 kDa, a decrease of 1.1 kDa in comparison with wildtype C. glutamicum (Fig. 2B). Similarly, the lipoglycans, Cg-LM and Cg-LAM from C. glutamicumDmptD ( Fig. 2A) are smaller and centred around m/z 13 000 (approximately 73 Manp/Araf glycosyl residues), and 4700 (approximately 20 Manp glycosyl residues) respectively ( Fig. 2B). The differences in size of Cg-LAM and Cg-LM from the mutant strains, suggest that there is a difference in a common component of both mutant strains, which also reflects the earlier observation on SDS-PAGE ( Fig. 2A).
Glycosyl composition of purified lipoglycans from C. glutamicumDmptC and C. glutamicumDmptD Purified lipoglycans from wild-type C. glutamicum, C. glutamicumDmptC and C. glutamicumDmptD were analysed as alditol acetates and their glycosyl composition was determined by gas chromatography (GC) (Fig. S3). GC analysis of alditol acetates prepared from wild-type Cg-LAM revealed a molar ratio of Araf : Manp of 0.47:1.0. The mutant Cg-LAM from C. glutamicumDmptC and C. glutamicumDmptD yielded a significant reduction in Manp content concomitant with a relative increase in the amount of Araf. The C. glutamicumDmptC yielded a mutant Cg-LAM with an Araf : Manp ratio of 0.59:1.0, while Cg-LAM from C. glutamicumDmptD, yielded an Araf : Manp ratio of 0.7:1.0. The data suggest that both Cg-MptC and Cg-MptD are involved in the synthesis of the mannan portion of Cg-LAM together with Cg-MptA and Cg-MptB 2008a).
In vitro a(1,2)-mannopyranosyltransferase assay utilizing a novel blocked nonasaccharide acceptor of 3, donor thioglycoside 4 was synthesized starting from the reported 5,6-O-benzylidene protected thioglycoside 5 (Cumpstey et al., 2004) (Fig. 5B). In a straight forward reaction sequence, thioglycoside 5 was treated with benzoyl chloride in pyridine to access the thioglycoside 6, which on regioselective ring opening with Et 3SiH gave 6-OBn thioglycoside 7. The 1 H NMR and 13 C NMR spectral studies confirmed the structure of 7. An acetate protecting group was installed at the 4-OH position in thioglycoside 7 by reacting with acetic anhydride in pyridine resulting in the thioglycoside donor 4. An acetate group was specifically chosen to easily characterize the reaction product during coupling of 4 and 2 by NMR. The donor thioglycoside 4 and 2 were coupled in the presence of NIS and triflic acid at -20°C, resulting in nonasaccharide 3 after chromatographic purification. The observed singlet signal in the 1 H NMR spectrum of 3 at d 1.88 ppm was assigned to the three protons of the acetyl protecting group, and the resonances at d 4.48 and 4.28 ppm (doublets, J = 11.9 Hz), for one proton each to the benzyl protecting group, supporting the structure of 3. The 13 C NMR spectrum further confirmed the structure of 3, with the presence of signals at 73.45 and 20.81 ppm, which were assigned to the CH2 of the benzyl and CH3 of the acetyl carbons respectively. The ester protecting groups in oligosaccharide 3 were removed in an overnight reaction using sodium methoxide in CH3OH. The reaction was neutralized using H + Amberlite resin and the reaction product purified using Bio-Beads SM-4 using a CH3OH-H2O mobile phase and 0-60% gradient resulting in the desired nonomannoside 1. The 1 H NMR spectrum of 1 showed the presence of benzyl signals at d 4.64 and 4.61 ppm (doublet, J = 11.7 Hz). In the 13 C NMR spectrum, the benzyl carbon was observed at d 73.32 ppm and CH2N3 at d 51.79 ppm. Finally, the structure of 1 was further supported by HR-MS, which showed a [M + Na] + peak at 1685.55.
The structural features of the acceptor allowed its specific isolation and purification of in vitro reaction products synthesized by distinct a(1→2)-mannopyranosyltransferases using a strong anion exchange cartridge, thus removing from the in vitro assay unused C50polyprenol-phospho-[ 14 C]-mannose (PP-[ 14 C]-M) and other in vitro synthesized products, such as higher [ 14 C]-PIMs, [ 14 C]-LM and [ 14 C]-LAM. In addition, the nonreducing sugar of 1 (which we have abbreviated as Acc-Man9) is blocked at the O-6 position by a benzyl ether, thus preventing further elongation by endogenous activities through MptA and MptB via a(1→6)mannopyranosyltransferases. Overall, the utilization of Acc-Man9 and the in vitro assay provides a novel route to evaluate distinct a(1→2)-mannopyranosyltransferases in Corynebacterineae (Fig. 6A). The in vitro transfer of [ 14 C]-Manp from PP-[ 14 C]-M onto the Acc-Man9 acceptor was examined using membrane preparations from wildtype and mutant strains of C. glutamicum (Fig. 6B). In this improved assay, membrane preparations from wild-type C. glutamicum transferred 3072 Ϯ 132 cpm of [ 14 C]-Manp A. Lipoglycans were extracted from wild-type and mutant strains and were analysed using SDS-PAGE and visualized using a Pro-Q emerald glycoprotein stain (Invitrogen). Lipoglycan profiles are represented with standard molecular weight markers of glycoproteins of 180, 82, 42 and 18 kDa. B. 2D 1 H-13 C-NMR HMQC characterization of Cg-LAM from C. glutamicum as described in Tatituri et al. (2007b). C. 2D 1 H-13 C-NMR HMQC characterization of Cg-LAM from C. glutamicumDmptCDmptD. NMR spectra of Cg-LAMs were recorded in D2O at 313 K. from PP-[ 14 C]M onto the synthetic Acc-Man9 via an a(1→2)-mannopyranosyltransferase ( Fig. 6B and inset). Assays repeated in the absence of Acc-Man9 afforded background activity 102 Ϯ 12 cpm, demonstrating the efficiency of the assay (Fig. 6B). Membrane preparations from both C. glutamicumDmptC and C. glutamicum-DmptD used in above assays afforded a(1→2)mannopyranosyltransferase activities, but in each case the relative level of activity in comparison with wild-type C. glutamicum membranes was reduced (Fig. 6B).

Functional identification of the mycobacterial homologue of Cg-MptC
As described above (Fig. 1), Cg-MptC showed highest similarity of 55% to Rv2181 of M. tuberculosis, while Cg-MptD showed weaker, but significant similarity of 47.9% to Rv2181. To assay for a complementary function of Rv2181 in C. glutamicum, plasmid pEKEx2-Mt-Rv2181 was constructed. Upon introduction into C. glutamicumD-mptC, the wild-type phenotype of C. glutamicum was restored ( Fig. 2A) illustrating that Mt-MptC is functional and utilizes the corynebacterial substrates. In addition, Rv2181 was very specific for C. glutamicumDmptC as complementation of the mutant phenotype in C. glutamic-umDmptD was not obtained (Fig. 2A). The complementation of C. glutamicumDmptC with pEKEx2-Mt-Rv2181 also restored the wild-type glycosyl linkage phenotype in this strain (Fig. 7, and see Fig. 3A and B). Furthermore, use of membrane preparations from C. glutamicumD-mptCDmptD pEKEx2-Mt-Rv2181 and the synthetic Acc-Man 9 (Fig. 6A) resulted in the incorporation of 1811 Ϯ 87 cpm of [ 14 C]-Manp into the [ 14 C]-mannan product (Fig. 6B). Altogether, these data strongly suggest that Rv2181 is a homologue of Cg-mptC and both encode for a(1→2)-mannopyranosyltransferase activities, involved in the synthesis of the a(1→2)-Manp branches of LM and LAM in Corynebacterineae.

Activation of Toll-like receptor 2 (TLR2) by wild-type and mutant Cg-LM and LAM
Both LM and LAM are bioactive molecules thought to be involved in modulating the host immune response (Briken et al., 2004). Bacterial lipoglycans are recognized by a variety of host immune receptors including TLR2. In fact, mycobacterial LMs are well known to be potent activators of this latter receptor hence providing them strong proinflammatory activity (Vignal et al., 2003;Quesniaux et al., 2004;Gilleron et al., 2006). On the other hand,  (Cooper et al., 2002).
LAMs are only poorly active; a feature attributed to the presence of the arabinan domain that masks the 'bioactive' mannan core (Vignal et al., 2003;Nigou et al., 2008;Birch et al., 2010). To investigate the ability of Cg-LM and Cg-LAM to activate TLR2 to assess possible consequences of Cg-mptC and Cg-mptD inactivation in this process, purified LM and LAM from C. glutamicumDmptC and C. glutamicumDmptD, were used to stimulate HEK293 cells expressing TLR2. To prevent interference of co-purified lipopeptides, the glycolipids were first pretreated with hydrogen peroxide (Morr et al., 2002;Geurtsen et al., 2009;Birch et al., 2010). As shown in Fig. 8, wild-type Cg-LAM was unable to activate TLR2. This result is consistent with an earlier study in which it was shown that LAMs substituted with single arabinosyl residues were unable to induce the production of tumour necrosis factor-a in human THP-1 monocytes . Interestingly, LAMs isolated from the C. glutamicumDmptC and C. glutamicumDmptD mutant strains demonstrated an increased activity as evidenced by the enhanced TLR2-dependent IL-8 production by HEK293 cells (Fig. 8). This suggests that partial removal of a(1→2)-Manp units, increases the ability of Cg-LAM to activate TLR2. In contrast to mycobacterial LMs, wild-type Cg-LM did not show any activity towards TLR2. Furthermore, mutation of Cg-mptC and Cg-mptD did not alter this behaviour and the mutant LMs remained inactive (Fig. 8). These results demonstrate that wild-type Cg-LAM and Cg-LM, are both TLR2 inactive. Furthermore, partial removal of a(1→2)-Manp branching may promote TLR2 activation, but this is only the case for Cg-LAM.

Discussion
Apart from sharing a similar cell wall architecture, C. glutamicum possesses similar genetic loci responsible for cell wall biosynthesis like M. tuberculosis. As a model, it has proven extremely useful in the study of essential genetic elements of M. tuberculosis (Alderwick et al.,  After 24 h of stimulation (37°C), supernatants were harvested and analysed for IL-8 content using ELISA. The results represent the mean IL-8 production Ϯ SD from three independent experiments (each experiment was performed in triplicate). Double asterisks indicate highly significant (P < 0.001) differences as compared to the unstimulated control cells.
2006a,b). In particular, it has enabled us to use it as a suitable model for the identification and functional study of mycobacterial genes involved in mycolic acid (Gande et al., 2004;, arabinogalactan (Alderwick et al., 2005;2006b;Seidel et al., 2007a,b;Birch et al., 2008; and LAM biosynthesis (Gibson et al., 2003;Mishra et al., 2007;2008a,b;Tatituri et al., 2007a,b). Previously, we reported the identification of enzymes involved in early and late stages of LM and LAM biosynthesis in Corynebacterineae 2008a,b;.
SDS-PAGE and MALDI-TOF-MS analysis of different lipoglycan species, including Cg-LM and Cg-LAM from C. glutamicumDmptC and C. glutamicumDmptD, suggested that there is a difference in a similar component of all these species. Two-dimensional thin-layer chromatography (2D-TLC) analysis supplemented with negative MALDI-TOF-MS of extracted glycolipids from mutants failed to show any differences in the PIM precursors (data not shown), and the presence of MptA and MptB in these mutants, negates the possibility of a change in the a(1→6) mannan backbone of these lipoglycans, leaving the only possibility of a change in a(1→2)-Manp branching, suggesting that MptC and MptD are involved in syn-thesis of these a(1→2)-Manp units in these lipoglycans. In addition, glycosyl linkage analysis revealed a reduction in the relative abundance of t-Manp and 2,6-Manp with a concomitant increase in the abundance of 6-Manp in Cg-LM and Cg-LAM from these strains. The cumulative chemical analysis of C. glutamicumDmptC and C. glutamicumDmptD agrees with the view that Cg-MptC and Cg-MptD are involved in synthesis of a(1→2)-Manp units in Cg-LM and Cg-LAM in C. glutamicum. This view was strengthened by the analysis of a C. glutamicumDmptCD-mptD double mutant and the specificity of MptC and MptD in cross-complementation studies (Fig. S2). In summary, the data clarified the occurrence of residual t-a-Manp, a(1→2)-Manp and t-a-Araf units in the Cg-MptC and Cg-MptD mutants, as the singular mutants only partially removed 2,6-Manp branching. The absence of these units in the C. glutamicumDmptCDmptD mutant provides evidence that these units are present as t-a-Manp-(a1→2)-Manp and t-a-Araf-(a1→2)-Manp motifs (Fig. 9). Furthermore, specific in vitro assays utilizing a chemically synthesized nonasaccharide acceptor demonstrate that both Cg-mptC and Cg-mptD encode for a(1→2)mannopyranosyltransferases, responsible for the synthesis of the a(1→2)-Manp units on the mannan backbone of LM and LAM. From the earlier studies with Rv2181 (Kaur et al., 2008), and the fact that Cg-MptC and Rv2181 are homologues, it is tempting to speculate that Cg-MptC adds a(1→2)-Manp residues to the proximal backbone of Cg-LM, or one could imply that Rv2181 in vitro has MptC function. A plasmid encoding Rv2181 was used to transform the C. glutamicum mutants, and Rv2181 was able to restore the wild-type phenotype in only the C. glutamic-umDmptC mutant, and possessed a(1→2)mannopyranosyltransferase activity in the defined in vitro acceptor assay reported in this study.
Whereas the LM/LAM backbone is potentially rather similar within the Corynebacterineae, as indicated also from the relatedness of the a(1→6)-mannopyranosyltransferases (Fig. 1), it is clear from structural, immunological and genetic evidence, that the more elaborated structures and decorations of these glycolipids are variable. Known structural variations include large arabinan domains in M. tuberculosis (Chatterjee et al., 1991;McNeil et al., 1994), Man-capping of arabinan domains in M. tuberculosis (Chatterjee et al., 1992) and their PI capping in M. smegmatis . Accordingly, also immunological properties of LM/LAM are different (Briken et al., 2004). This variability is naturally reflected at the genetic level. An example is the capping GT-C transferase of M. tuberculosis encoded by Rv1635c (Appelmelk et al., 2008), which is not present in C. glutamicum. The fact that MptC in C. glutamicum is a fusion protein carrying a second domain with unknown function, also illustrates that mature LM and LAM glycolipids are species-specific and accordingly the corresponding biosynthesis genes more prone to rapid evolution. Duplication or fusion of genes may indicate that the types of enzymes involved in the later steps of LM and LAM synthesis are highly dynamic and of evolutionary change within the Corynebacterineae (Xie et al., 2003).
A key question in lipoglycan biosynthesis is the division of LM and LAM abundance in the cell wall. The separation of LM and LAM in C. glutamicum, biosynthetically probably occurs after the initial formation of the a(1→6) mannan backbone by MptB, followed by a(1→2)-Manp branching by MptC (or MptD). The product of these two enzymes would then be the preferred substrate for further a(1→6)-Manp elongation by MptA, followed by a(1→2)-Manp branching by MptD (or MptC) (Fig. 9). Additional GT-C glycosyltransferases remain to be characterized involved in further modification of the a(1→6)-mannan backbone. These include how the dimanoside (via MptE?) and Araf-Manp side-chains are synthesized (AftE?) and are summarized in Fig. 9.
Both LAM and LM are immunomodulatory molecules (Briken et al., 2004). Whereas LAM has an important role during intracellular survival, for instance by preventing phagosome-lysosome fusion (Fratti et al., 2003), LM has strong pro-inflammatory activity through its ability to activate TLR2 (Vignal et al., 2003;Quesniaux et al., 2004;Gilleron et al., 2006). Previous research has demonstrated that the pro-inflammatory activity of LM is at least partly based upon its acylation status (Gilleron et al., 2006;Doz et al., 2007). PI-based LAM and LM species can differ in their degree of acylation, each carrying between one and four acyl chains. Investigation of the immune activating properties of differentially acylated LM species demonstrated that LM should at least be triacylated to be pro-inflammatory, with species carrying less acyl chains exhibiting immunosuppressive activity in a mechanism independent of TLR2 (Doz et al., 2007).
Here we investigated different Cg-LAM and LM species for their ability to activate TLR2. Similar to mycobacterial LAMs, wild-type Cg-LAM was found to be inactive (Fig. 8). This finding is consistent with previous observations that LAM species carrying only t-Araf substitutions, such as LAMs isolated from Rhodococcus ruber and Turicella otitidis, are largely inactive . Comparison of wild-type Cg-LAM to the LAMs isolated from the C. glutamicumDmptC and C. glutamicumDmptD mutant strains demonstrated that the mutant LAMs had gained the ability to activate TLR2 (Fig. 8). The explanation for this increased activity is currently not known. Interestingly, wild-type Cg-LM, like wild-type Cg-LAM, but unlike mycobacterial LM was found to be completely TLR2-inactive (Fig. 8). The reason for the discrepancy is unclear, but may be related to structural differences in their membrane anchors. In mycobacteria, LAMs and LMs share a common PI anchor, on which further assembly occurs to give mature LM and LAM. However, in C. glutamicum, although Cg-LAM contains a normal PI anchor, two different pools of Cg-LM (Cg-LM-A and Cg-LM-B) dominate the majority of lipoglycan composition of the corynebacterial cell wall (Tatituri et al., 2007a;Lea-Smith et al., 2008;Mishra et al., 2008b). The large majority of the Cg-LM species (~95%), Cg-LM-B, contains an anchor based upon a di-acylated a-D-glucopyranosyluronic acid-(1→3)glycerol moiety, and only a minor fraction contains the more classical anchor based on PI, Cg-LM-A (Mishra et al., 2008b). Despite the different lipid anchors, the mannan backbone structures of Cg-LM-A and Cg-LM-B are similar as revealed through GC-MS analysis of partially methylated alditol acetates (Mishra et al., 2008b). This means that the majority of Cg-LM species will contain a di-acylated GlcAGroAc 2 anchor, which is a feature that has been demonstrated to be incompatible with activation of TLR2 (Doz et al., 2007).
Furthermore, a complete investigation of the structural and functional relationship of these lipoglycans during host-pathogen interaction requires the availability of mutants defective in their respective biosynthetic pathways. The availability of complete genome sequences of several mycobacteria and related actinomycetes, like C. glutamicum, and the development of novel tools for genetic manipulation has opened up these possibilities (Cole et al., 1998).

Experimental procedures
Bacterial strains, cell culture and growth conditions Corynebacterium glutamicum and E. coli DH5amcr were grown in LB broth at 30°C and 37°C respectively. For C. glutamicum complex medium BHI and minimal medium CGXII was used (Eggeling and Bott, 2005). Kanamycin and ampicillin were used at a concentration of 50 mg ml -1 for selection of recombinants. Samples for lipid analysis were prepared by harvesting cells at an OD of 10-15, followed by a saline wash and freeze drying. HEK293 cells transfected with TLR2 (Flo et al., 2002) were kept in DMEM (Invitrogen) containing 10% FCS, 100 U ml -1 penicillin, 100 mg ml -1 streptomycin, 0.5 mg ml -1 G418, 2 mM L-glutamine and 110 mg l -1 pyruvate.

Construction of plasmids and strains
To delete Cg-mptC and Cg-mptD the deletion vectors pK19mobsacBDCg-mptC, and pK19mobsacBDCg-mptD were constructed. In each case cross-over PCR was used with genomic DNA as template and two different PCR's with primer pairs AB and CD (Table S1). The resulting PCR products served as template for primer pairs AD. The PCR product contained 12 nucleotides (nt) of the 3′ end of the respective gene together with genomic upstream sequences, and 36 nt of the 5′ end together with genomic downstream sequences. The inserts of all plasmids used in this work were confirmed by sequencing. Genes were deleted in by first introducing plasmids prepared from E. coli via electroporation into C. glutamicum and then selection for sucrose resistance in a procedure as described (Schäfer et al., 1994). Starting from C. glutamicumDmptC and using pK19mobsacBDCg-mptD the double deletion mutant C. glutamicumDmptCDmptD was made. The starting strain used was the type strain C. glutamicum ATCC13032. Chromosomal deletions were confirmed using primer pairs AD, as well as the additional new primer pairs EF, hybridizing outside of the regions used for plasmid constructions.
To assay for complementation of chromosomal deletions four expression plasmids were used. To construct pEKEx2-Cg-mptC primer pairs Ex-2100-RBS-for and Ex-2100-rev were used to amplify the 5′ half of NCgl2100, the PCR product treated with SalI/EcoRI and ligated with similar treated pEKEx2. To construct pEKEx2-Cg-mptD primer pairs Ex-2097-RBS-for and Ex-2097-rev were used to amplify NCgl2097, the PCR product treated with SalI/EcoRI and ligated with similar treated pEKEx2. To construct pEKEx2-Mt-Rv2181, chromosomal DNA of M. tuberculosis served as template, and Rv2181 was amplified using primer pairs Ex-Rv2181-for and Ex-Rv2181-rev, the former providing the sequence CTGCAG as a ribosome binding site. The amplified product was treated with PstI and EcoRI and ligated with pEKEx2. Plasmids were used to transform the respective mutants via electroporation to kanamycin resistance, and recombinants confirmed by plasmid preparation.

Lipid extraction and analysis
Polar and apolar lipids were extracted as described by Dobson et al. (1985). Briefly, 6 g of dry C. glutamicum cells was treated in 220 ml of methanolic saline (20 ml 0.3% NaCl and 200 ml CH3OH) and 220 ml of petroleum ether for 2 h (Dobson et al., 1985). The suspension was centrifuged and the upper layer containing apolar lipids was separated. An additional 220 ml of petroleum ether was added, mixed and centrifuged as described above. The two upper petroleum ether fractions were combined and dried. For polar lipids, 260 ml CHCl3/CH3OH/0.3% NaCl (9:10:3, v/v/v) was added to the lower aqueous phase and stirred for 4 h. The mixture was filtered and the filter cake re-extracted twice with 85 ml of CHCl3/CH3OH/0.3% NaCl (5:10:4, v/v/v). Equal amounts of CHCl3 and 0.3% NaCl (145 ml each) were added to the combined filtrates and stirred for 1 h. The mixture was allowed to settle, and the lower layer containing the polar lipids recovered and dried. The polar lipid extract was examined by 2D-TLC on aluminium backed plates of silica gel 60 F254 (Merck 5554), using CHCl3/CH3OH/ H2O (65:25:4, v/v/v) in the first direction and CHCl3/CH3COOH/CH3OH/ H2O (40:25:3:6, v/v/v/v) in the second direction. The thin-layer chromatographic plates sprayed with the appropriate staining solution to detect the presence of lipids, glycolipids or phospholipids as described (Dobson et al., 1985).

Extraction and purification of lipoglycans
Lipoglycans were extracted from delipidated cells as previously described (Ludwiczak et al., 2001). Briefly, cells were broken by sonication (MSE Soniprep 150, 12 micron amplitude, 60 s ON, 90 s OFF for 10 cycles, on ice) and the cell debris refluxed five times with 50% C2H5OH at 68°C, for 12 h intervals. The cell debris was removed by centrifugation and the supernatant containing lipoglycans, neutral glycans and proteins dried. This dried extract was then treated with hot phenol-H2O. The aqueous phase was dialysed and dried, followed by extensive treatments with a-amylase, DNase, RNase chymotrypsin and trypsin. This fraction was dialysed to remove the low MW break-down products formed after the enzyme treatment, thus yielding the crude lipoglycan fraction.
The crude lipoglycan extract was dried and resuspended in buffer A (50 mM ammonium acetate and 15% propan-1-ol) and subjected to Octyl Sepharose CL-4B chromatography (2.5 ¥ 50 cm). The column (2.5 ¥ 50 cm) was washed initially with four column volumes of buffer A to ensure removal of neutral glycans followed by buffer B (50 mM ammonium acetate and 50% propan-1-ol). The eluent was collected and concentrated to approximately 1 ml and precipitated using 5 ml of C2H5OH. The sample was dried using a Savant Speedvac and then resuspended in buffer C [0.2 M NaCl, 0.25% sodium deoxycholate (w/v), 1 mM EDTA and 10 mM Tris-HCl, pH 8] to a final concentration of 20 mg ml -1 . The sample was gently mixed and left to incubate for 48 h at room temperature. The sample was then loaded onto a Sephacryl S-200 (2.5 cm ¥ 50 cm) column previously equilibrated with Identification and characterization of novel a(1,2)-mannopyranosyltransferases 1253 buffer C. The sample was eluted with 400 ml of buffer C at a flow rate of 3 ml h -1 , collecting 1.5 ml fractions. The fractions were monitored by SDS-PAGE using either a silver stain or a Pro-Q emerald glycoprotein stain and individual fractions pooled and dialysed extensively against buffer D (10 mM Tris-HCl, pH 8, 0.2 M NaCl, 1 mM EDTA) for 72 h with frequent changes of buffer. The samples were further dialysed against deionized water for 48 h with frequent changes of water, lyophilized and stored at -20°C.

MALDI-TOF-MS and NMR analysis
The matrix used was 2,5-dihydroxybenzoic acid at a concentration of 10 mg ml -1 , in a mixture of H2O/C2H5OH (1:1, v/v), 0.1% trifluoroacetic acid (TFA). Samples (0.5 ml) at a concentration of 10 mg ml -1 were mixed with 0.5 ml of the matrix solution. Analyses were performed on a Voyager DE-STR MALDI-TOF instrument (PerSeptive Biosystems, Framingham, MA, USA) using linear mode detection. Mass spectra were recorded in the negative mode using a 300 ns time delay with a grid voltage of 80% of full accelerating voltage (25 kV) and a guide wire voltage of 0.15%. The mass spectra were mass assigned using external calibration. NMR spectra of LAM samples were recorded on a Bruker DMX-500 equipped with a double-resonance ( 1 H/X)-BBi z-gradient probe head. All samples were exchanged in D2O, with intermediate lyophilization, and then dissolved in 0.7 ml D2O and analysed at 313 K. The 1 H and 13 C NMR chemical shifts were referenced relative to internal acetone at 2.225 and 34.00 ppm respectively. All details concerning NMR sequences and the experimental procedures were described previously (Tatituri et al., 2007b).

Glycosyl compositional and linkage analysis
Lipoglycans were hydrolysed using 250 ml of 2 M TFA at 120°C for 2 h, reduced with 100 ml of NaBD4 (10 mg ml -1 in C2H5OH:1 M NH4OH), and the resultant alditols per-Oacetylated using acetic anhydride (100 ml) at 120°C for 1 h, before examination by GC (Tatituri et al., 2007b). GC was performed using a Thermoquest Trace GC 2000 and a DB225 column (Supelco) with samples injected in the splitless mode. The oven was programmed to hold at an isothermal temperature of 275°C for a run time of 15 min. Glycosyl linkage analyses were performed as described previously (Tatituri et al., 2007b). Briefly, lipoglycan samples were per-O-methylated using dimethyl sulfinyl carbanion, hydrolysed using 2 M TFA, reduced using NaB2H4, and per-O-acetylated before analysis by GC-MS (Tatituri et al., 2007b). GC-MS was carried out on a Finnigan Polaris GCQ PlusTM using a BPX5 (Supleco) column. The injector temperature was set at 50°C, held for 1 min and then increased to 110°C at 20°C min -1 . The oven was held at 110°C then ramped to 290°C at 8°C min -1 and held for 5 min to ensure all the products had eluted from the column. All the data were collected and analysed using Xcaliber (v.1.2) software.

General methods for chemical synthesis of Acc-Man9 (compound 1)
Reactions were performed under a dry argon atmosphere. Whenever necessary, compounds and starting materials were dried by azeotropic removal of water with toluene under reduced pressure. The reactions were monitored by TLC on pre-coated silica gel (60F 254) plates (0.25 mm) from E. Merck and visualized using UV light (254 nm) and/or heating after spraying with (NH4)2SO4 solution (150 g ammonium sulphate, 30 ml H2SO4, 750 ml H2O). Flash column chromatography was carried out on Isco Teledyne CombiFlash Rf200 chromatographic system using pre-packed columns. 1 H and 13 C NMR spectra were recorded at 400 MHz and 100 MHz, respectively, on a Varian 400-MR spectrometer. Coupling constants (J) are reported in Hz and chemical shifts (d) in ppm relative to a residual solvent peak or an internal standard (TMS). The HR-MS spectra were recorded on a Agilent 6140 liquid chromatography mass spectrometry instrument.