University of Birmingham Structural characterization of a partially arabinosylated lipoarabinomannan variant isolated from a Corynebacterium glutamicum ubiA mutant

Arabinan polysaccharide side-chains are present in both Mycobacterium tuberculosis and Corynebacterium glutamicum in the heteropolysaccharide arabinogalactan (AG), and in M. tuberculosis in the lipoglycan, lipoarabinomannan (LAM). Herein, we show by quantitative sugar and glycosyl linkage analysis that C. glutamicum possesses a much smaller LAM version, Cg-LAM, characterised by single t -Ara f residues linked to the (cid:1) (1 (cid:1) 6)-linked mannan backbone. MALDI-TOF MS showed an average molecular weight of 13 800-15 400 Da for Cg-LAM. The biosynthetic origin of Ara f residues found in the extracytoplasmic arabinan domain of AG and LAM is well known to be provided by decaprenyl-monophosphoryl- D -arabinose (DPA). However, the characterisation of LAM in a C. glutamicum :: ubiA mutant devoid of prenyltransferase activity and devoid of DPA-dependent arabinan deposition into AG, revealed partial formation of LAM, albeit with a slightly altered molecular weight. These data suggest that in addition to DPA utilisation as an Ara f donor, alternative pathways exist in Corynebacterianeae for Ara f delivery, possibly via an unknown sugar nucleotide.


INTRODUCTION
The Corynebacterianeae represent a distinct and unusual group within Gram-positive bacteria, with the most prominent members being the human pathogens Mycobacterium tuberculosis and Mycobacterium leprae (Bloom & Murray, 1992). In addition, nonpathogenic bacteria also belong to this taxon, such as Corynebacterium glutamicum, which is used in the industrial production of amino acids (Sahm et al., 2000) . These bacteria collectively belong to the same sub-order, and share a similar genome, cell envelope and corresponding cell wall biosynthetic enzymes (Dover et al., 2004;Stackebrandt et al., 1997).
For LAM biosynthesis we initially proposed the following biosynthetic pathway PI PIM LM LAM (Besra et al., 1997) in M. tuberculosis, which is now largely supported by biochemical and genetic evidence (Gurcha et al., 2002;Kordulakova et al., 2002;Kremer et al., 2002;Schaeffer et al., 1999). PimA catalyses the addition of Manp provided by GDP-Mannose to the 2-position of the myo-inositol of PI to form PIM 1 (Kordulakova et al., 2002), whereas PimB might be responsible for the addition of a second Manp to the 6-position to yield Ac 1 PIM 2 (Schaeffer et al., 1999;Tatituri et al., 2007). PimC has been demonstrated to allow further mannosylation to Ac 1 PIM 3 (Kremer et al., 2002) and more recently, PimE has been shown to involved in the biosynthesis of Ac 1 PIM 5 (Morita et al., 2006). It has been proposed that PIM 4 is the direct precursor of LM, characterised by a linear (1 6)-linked mannan backbone linked with (1 2) mannopyranose side chains generated through a mannosyltransferase encoded by Rv2181 (Kaur et al., 2006). LM is then further glycosylated with arabinan to produce LAM, and finally 'mannose-capped' to produce ManLAM, a process initiated by the capping enzyme encoded by Rv1635c (Dinadayala et al., 2006. For the addition of arabinose residues into mycobacterial LAM, EmbC is required (Berg et al., 2005;Shi et al., 2006;Zhang et al., 2003), whereas AftA, EmbA and EmbB perform arabinan polymerisation in AG (Alderwick et al., 2006b;Belanger et al., 1996;Escuyer et al., 2001). In contrast, C. glutamicum possesses a singular Emb, and an AftA orthologue, which are used in AG biosynthesis (Alderwick et al., 2005;Alderwick et al., 2006b;Seidel et al., 2007). It is interesting to note that the arabinan domains of AG and LAM utilise several different Araf linkages, which suggests that additional arabinofuranosyltransferases must be required to form AG and LAM, and still remain to be identified in Corynebacterianeae.
It has been shown that the activated Araf sugar donor in Corynebacterianeae is decaprenyl-monophosphoryl-D-arabinose (DPA) (Wolucka et al., 1994). It was also proposed that decaprenyl-monophosphoryl-D-ribose (DPR) could be an additional precursor involved in arabinan synthesis, via a 2'-epimerase (Wolucka & de Hoffmann, 1995). Recently, it was shown that 5-phosphoribofuranose pyrophosphate (pRpp) is converted to decaprenyl-monophosphoryl-5-phosphoribose (DPPR) via a 5-phospho--Dribose-1-diphosphate-decaprenyl-phosphate 5-phosphoribosyltransferase, (UbiA; Rv3806c and NCgl2781 in M. tuberculosis and C. glutamicum, respectively) (Huang et al., 2005), which is then dephosphorylated to form DPR. DPR is then enzymatically oxidised to form a keto-intermediate, DPX, followed by an enzymatic reduction to DPA (Mikusova et al., 2005). Additionally, disruption of ubiA in C. glutamicum, the first enzyme involved in the biosynthesis of DPA, resulted in a complete loss of cell wall arabinan (Alderwick et al., 2006a). To date, no sugar nucleotides of Araf have been identified. The key discovery that Araf residues in AG arise from DPA, also raised doubts whether mycobacteria neither produce nor require Araf sugar nucleotides. Herein, we now clearly show that an ubiA disrupted mutant of C. glutamicum still produces a modified LAM variant (Cg-LAM), which is arabinosylated even in the absence of the sugar donor DPA (Alderwick et al., 2006a). Taken together, the data suggests that alternative pathways may exist in Corynebacterianeae, independent of DPA utilisation, specifically involved in arabinosylation events in Cg-LAM biosynthesis.

METHODS
Strains, construction of plasmids and culture conditions. C. glutamicum ATCC 13032 (the wild type strain, and referred for the remainder of the text as C. glutamicum) and Escherichia coli DH5 mcr were grown in Luria-Bertani (LB) broth (Difco) at 30 °C and 37 °C, respectively. The C. glutamicum::ubiA mutant was grown on complex medium brain heart infusion (BHI) (Difco). Kanamycin and ampicillin were used at a concentration of 50 g ml -1 . The vector used for inactivation of C. glutamicum::ubiA was pCg::ubiA. It contained an internal ubiA-fragment of 321 bp amplified with the primer pairs ATC TTC AAC CAG CGC ACG ATC and AAT ATC GAT CAC TGG CAT GTG, which was ligated into the SmaI site of the non-replicative vector pK18mob to yield pCg::ubiA. To enable chromosomal inactivation of ubiA, pCg::ubiA was introduced into C. glutamicum by electroporation (Alderwick et al., 2005).

Extraction and purification of lipoglycans.
Purification procedures were adapted from protocols established for the extraction and purification of mycobacterial lipoglycans (Nigou et al., 1999). Briefly, 10 g of cells grown to and absorbance at 600 nm of 1.0, were delipidated at 60 o C by using 500 ml CHCl 3 /CH 3 OH (1:1, v/v). The delipidated cells were resuspended in deionized water and disrupted by probe sonication (MSE Soniprep 150, 12 micron amplitude, 60 s ON, 90 s OFF for 10 cycles, on ice) and the cell debris refluxed 5 times with 100 ml of C 2 H 5 OH/H 2 0 (1:1, v/v) at 68 °C, for 12 h intervals. The cell debris was removed by centrifugation and the supernatant containing lipoglycans, neutral glycans and proteins dried. The dried supernatant was then treated with a hot 80 % (w/w) phenol-H 2 O, biphasic wash at 70 °C for 1 h, followed by several protease treatments and the lipoglycan fraction recovered following extensive dialysis against deionized water (Nigou et al., 1999).
The crude lipoglycan extract was resuspended in buffer A (50 mM ammonim acetate and 15 % propan-1-ol) and subjected to Octyl-Sepharose CL-4B hydrophobic interaction chromatography (2.5 cm x 50 cm) as previously reported (Leopold & Fischer, 1993). The column was initially washed with 4 column volumes of buffer A to ensure removal of neutral glycans followed by an increasing gradient of propan-1-ol ranging from 25-65 % keeping the concentration of ammonium acetate constant. The eluates were collected and extensively dialysed against deionized water and concentrated to approximately 1 ml and precipitated using 5 ml of C 2 H 5 OH and the sample freeze-dried using a Savant SpeedVac.
The freeze-dried sample containing the retained material from the hydrophobic interaction column was then re-suspended in buffer B (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 200 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 column (2.5 cm x 50 cm) previously equilibrated with buffer B. The sample was washed with 400 ml of buffer C (0.2 M NaCl, 1 mM EDTA and 10 mM Tris-HCl, pH 8) at a flow rate of 3 ml h -1 , collecting 1.5 ml fractions using a Bio-Rad auto-sampler. The fractions were monitored by SDS-PAGE using either a silver stain utilising periodic acid and silver nitrate (Hunter et al., 1986) or a Pro-Q emerald glycoprotein stain (Invitrogen) and individual fractions pooled and dialysed extensively against buffer C for 72 h with frequent changes. The samples were further dialysed against deionized water for 48 h with frequent changes of deionized water, lyophilised and stored at -20ºC.

Glycosyl composition and linkage analysis of lipoglycans by alditol acetates.
Lipoglycans from wild type C. glutamicum and the C. glutamicum::ubiA mutant were hydrolysed using 2 M trifluoroacetic acid (TFA), reduced with NaBD 4 , and the resultant alditols were per-O-acetylated and examined by Gas Chromatography (GC) (Tatituri et al., 2007). Lipoglycans (2 mgs) were per-O-methylated using dimethyl sulfinyl carbanion as described previously (Alderwick et al., 2005;Besra et al., 1995;Daffé et al., 1990). In this procedure, lipoglycan samples (2 mgs) were resuspended in 0.5 ml dimethyl sulphoxide (anhydrous) and 100 μl of 4.8 M dimethyl sulfinyl carbanion. The reaction mixture was stirred for 1 h and then iodomethane (50 μl) was slowly added and the suspension stirred for 1 h, this process was repeated for a total of three times. The reaction mixture was then diluted with an equal volume of water, dialysed against deionized water and dried. The resulting retentate was applied to a pre-equilibrated C 18 Sep-Pak cartridge and initially washed with 10 ml water, 10 ml 20 % acetonitrile, and the Finnigan Polaris/GCQ PlusTM, as described previously (Besra et al., 1995;Daffé et al., 1990 ppm, respectively. All the details concerning NMR sequences used and experimental procedures were described in previous studies (Gilleron et al., 1999;Gilleron et al., 2000;Nigou et al., 1999). Sequazyme peptide mass standards were used as external calibrants (Applied Biosystems, Foster City, CA).

RESULTS
Disruption of Cg-ubiA. The ubiA gene product was shown in prior work to synthesise DPPR, which is converted to DPA, thus supplying the substrate for the "priming" arabinosyltransferase AftA (Alderwick et al., 2006b). These initial Araf residues "prime" the galactan backbone for further attachment of (1 5) linked Araf residues. These reactions require the arabinofuranosyltransferase activities of Mt-EmbA and Mt-EmbB, or Cg-Emb, respectively, and also the utilisation of DPA (Alderwick et al., 2005;Escuyer et al., 2001;Seidel et al., 2007), to eventually result in mature AG. In light of our present studies demonstrating the occurrence of Cg-LAM we inactivated the mycobacterial orthologue of C. glutamicum, NCgl2781, by transforming the wild type to kanamycin resistance conferred by the vector borne aph gene product of pCg::ubiA with a view of generating a clear C. glutamicum LM-only phenotype. The vector, as previously described (Alderwick et al., 2005), was integrated into the chromosomal ubiA gene, thus disrupting ubiA, as confirmed by two independent PCR analyses with two different primer pairs (Alderwick et al., 2005). As expected, the resulting strain C. glutamicum::ubiA exhibited a strong reduced growth (Alderwick et al., 2006a).
Purification of lipoglycans from C. glutamicum and C. glutamicum::ubiA. Analysis of the crude phenol-extracted lipoglycans from C. glutamicum (Fig. 1A) and C.
glutamicum::ubiA (Fig. 1B) visualised on SDS-PAGE revealed two closely migrating lipoglycans. A two-step purification protocol was performed to fractionate the crude lipoglycans. In the first step, the neutral glycans and nucleic acids were eliminated by using an octyl-Sepharose CL-4B column (Amersham Biosciences). In the second step, the lipoglycans were individually purified using a Sephacryl S-200 column (Amersham Biosciences). Fractions containing the lipoglycans were monitored by SDS-PAGE stained with either silver nitrate or Pro-Q emerald glycoprotein stains and pooled accordingly to afford lower (1) and upper (2) lipoglycans ( Fig. 1A and B).  (Tatituri et al., 2007).

General structural features of lipoglycans from
Accordingly, the 1D 1 H-NMR anomeric region (Fig. 4A)  Our previous experiments demonstrated that the C. glutamicum::ubiA mutant failed to synthesise DPA (Alderwick et al., 2006a) and possessed an arabinan deficient cell wall phenotype (Alderwick et al., 2005). As a consequence, we anticipated that the Cg-LAM from the mutant would be devoid of arabinose. Surprisingly, glycosyl compositional analysis of Cg-LAM from the C. glutamicum::ubiA mutant contained arabinose, albeit substantially reduced, and mannose in a ratio (4.5:95.5) (Fig. 2B). In addition, per-Omethylation analysis of Cg-LAM from the mutant indicated the presence of t-Araf, t-Manp, 2-Manp, 6-Manp and 2,6-Manp, although, the relative abundance of t-Araf residues was reduced (Fig. 3B). In agreement with the glycosyl compositional and linkage analysis data indicating a reduced amount of t-Araf units, the 1D 1 H anomeric region of Cg-LAM from the glutamicum::ubiA mutant (Fig. 4C) was simpler than that observed for Cg-LAM from wild type C. glutamicum (Fig. 4A). Resonance I 1 corresponding to a t--Araf unit was dramatically reduced, suggesting two overlapping t-Araf signals. This was further confirmed by 1 H-13 C HMQC analysis (Fig. 4D). Indeed, the relative intensity of the correlation at H 1 C 1 5.20/112.2 (I 1 ) was much weaker and correlation at H 1 C 1 5.13/112.0 (II 1 ) of the third t-Araf residue was absent.
The native Cg-LAMs from wild type C. glutamicum and the C. glutamicum::ubiA mutant were analysed by MALDI TOF MS. MALDI spectra were performed in the linear negative mode. The mass spectra show broad unresolved molecular ion envelopes due to the heterogeneity of the Cg-LAMs and the relatively low resolution of this type of mass spectrometry. The native Cg-LAM from wild type C. glutamicum shows an average molecular weight of around 15 400 Da (Fig. 5A). The native Cg-LAM from the C.
glutamicum::ubiA mutant shows a somewhat lower average molecular weight in the region of 13 900 Da (Fig. 5B). Samples of Cg-LAM from wild type C. glutamicum and Altogether, these data indicate that the C. glutamicum::ubiA mutant produces a slightly truncated Cg-LAM with a severe reduction in t-Araf content but maintaining and possibly extending its mannan core. In addition, the presence of UbiA and the biosynthesis of DPA, appears essential for the higher arabinosylation of Cg-LAM.
Moreover, these findings suggest that arabinose deposition in the C. glutamicum::ubiA mutant is a result of incorporation of t-Araf through another arabinose substrate, and one which is not DPA dependent.

DISCUSSION
Understanding the biosynthesis of mycobacterial arabinan is paramount to identifying potential new drug targets for the treatment of tuberculosis. However, as discussed earlier a complete understanding of these pathways is far from clear. In this respect, C.
glutamicum is a useful tool in understanding mycobacterial cell wall biosynthesis, since this organism possesses the core structural elements in Corynebacterianeae with few gene duplications, and deletion of orthologous genes which is often lethal in mycobacterial species is possible (Alderwick et al., 2005;Alderwick et al., 2006b;Gande et al., 2004). Specifically, the deletion of the DPA requiring arabinosyltransferase Emb in C. glutamicum resulted in a strongly abrogated arabinan domain of AG (Alderwick et al., 2005), and deletion of AftA also disabled attachment of the remaining singular t-Araf residues to galactan (Alderwick et al., 2006b). Furthermore, the loss of DPA synthesis in a C. glutamicum::ubiA mutant also resulted in the full loss of the arabinan domain in mature AG (Alderwick et al., 2005). In contrast, Cg-LM and Cg-LAM isolated from the corresponding C. glutamicum emb and aftA mutants were similar to the wild type glycans (data not shown). An absence of Araf residues was expected in Cg-LAM of the C. glutamicum::ubiA mutant. Strikingly, a total absence of arabinan was not observed (Fig. 2), thus requiring an alternative source and mechanism of addition of this sugar which may be attributed to the presence of an uncharacterised nucleotide sugar donor in

Corynebacterineae.
In contrast to mycobacteria, parasites do synthesise and utilise a sugar nucleotide to provide an activated arabinose sugar donor. Schneider et al. (1994) characterised GDP--D-Arap, the precursor of D-Ara in Leishmania major lipophosphoglycan. In trypanosomatids, this sugar nucleotide is believed to be synthesised in a two step process through the combined activities of an Ara-1-kinase and GDP-Ara pyrophosphorylase, whereby arabinose is phosphorylated to form arabinose 1-phosphate and then activated to the nucleotide level by GDP-D-Arap pyrophosphorylase (Mengeling & Turco, 1999;Schneider et al., 1995). Compared with the many genes encoding glycosyltransferases in prokaryotic and eukaryotic systems, very little is known about those that encode arabinopyranosyltransferases. Dobson et al. (2003) identified two genes (SCA1/2) encoding arabinosyltransferases mediating scAra capping, and recently it was demonstrated by heterologous expression that Leishmania SCA1 encodes an arabinopyranosyltransferase (Goswami et al., 2006).
Glycosyltransferases belonging to one of the families, GT-A and GT-B, or both, may be responsible for the initial addition of t-Araf residues on to Cg-LAM in the absence of DPA. These enymes may be present on the cytoplasmic side of the membrane and probably aid in the transfer of t-Araf residues from a putative sugar nucleotide donor.
These results could explain the presence of t-Araf in Cg-LAM, even though DPA biosynthesis has been completely abrogated in the C. glutamicum::ubiA mutant (Alderwick et al., 2005). The mechanism as to how these Araf residues are added is not clear. One possible explanation is that Araf residues from a nucleotide precursor are added in a step-wise fashion to Cg-LM on the cytoplasmic side to give motif-A proximal to PI, and then this complex polysaccharide is then probably flipped on to the periplasmic side. It is thought that the flipping mechanism is a prerequisite for arabinosylation to occur outside of the membrane, where the arabinan and mannan domains are formed through specific GT-C glycosyltransferases, by the addition of either t-Araf and t-Manp residues, from DPA or decaprenyl-monophorphoryl-D-mannose (DPM) to give motif-B, C and D ( Fig. 4E and Fig. 6) (Berg et al., 2007;Seidel et al., 2007;Tatituri et al., 2007).
The biosynthetic origin of mycobacterial Araf residues is generally a poorly understood area. To date, no sugar nucleotides of Araf have been identified, and as a result, a number of theories regarding the generation of arabinan in mycobacteria have been put forward (Huang et al., 2005;Mikusova et al., 2005;Scherman et al., 1995;Scherman et al., 1996). Depending on the organisation of the arabinan chains, it is speculated that five or six arabinofuranosyltransferases are required for arabinan biosynthesis in mycobacteria.
In this regard, it is interesting to note that an M. smegmatis embC mutant (Escuyer et al., 2001;Zhang et al., 2003) was found to be devoid of LAM but from the chemical analysis, the resulting LM precursor, possessed 2-3 Araf residues. The possibility exists that these residues are initially added via a non-DPA dependent glycosyltransferase and requires further investigation.   derivatives were analysed by GC/MS as described previously (Alderwick et al., 2005).   glycosyltransferase (Liu & Mushegian, 2003). After transportation across the membrane by an unknown "flippase" enzyme, further elaboration of the lipoglycan then occurs through the addition of mannose and arabinose units catalysed by several GT-C glycosyltransferases utilising DPM and DPA, respectively (Berg et al., 2007;Seidel et al., 2007;Tatituri et al., 2007).