The singular Corynebacterium glutamicum Emb arabinofuranosyltransferase polymerises the α(1 → 5) arabinan backbone in the early stages of cell wall arabinan biosynthesis

The arabinan-containing polysaccharides, arabinogalactan (AG) and lipoarabinomannan (LAM), are key cell wall components of the Corynebacterineae, which include Corynebacteria, Norcadia and Mycobacteria. Both AG and LAM contain elaborate arabinan domains composed of distinct structural motifs. Mycobacterial EmbA, EmbB and EmbC, collectively known as the Emb proteins, have been identified as arabinosyltransferases (ArafTs), which are targeted by the front-line anti-tubercular drug ethambutol. Previous studies have established that EmbA and EmbB play a role in the synthesis of the characteristic terminal hexa-arabinosuranosyl motif, whilst EmbC is involved exclusively in the biosynthesis of LAM. Herein, we have investigated the role of the singular Emb protein from Corynebacterium glutamicum through the detailed biochemical and chemical analysis of a double ΔaftAΔemb mutant, where the priming Cg-AftA protein, which generates the substrate for Cg-Emb has been deleted. Analysis of its cell wall revealed a complete absence of arabinose resulting in a truncated cell wall containing only a galactan backbone accompanied with complete loss of cell wall bound mycolates. In vitro cell-free assays using C. glutamicumΔaftA, C. glutamicumΔemb, C. glutamicumΔaftAΔemb and C. glutamicumΔaftBΔaftD and two synthetic acceptors, which mimick the arabinofuranose (Araf) “primed” galactan chain, demonstrated that Cg-Emb is able to transfer an Araf residue to the C5 of the Araf positioned on the synthetic acceptor(s). These results indicate that Cg-Emb acts as an α(1 → 5) ArafT and elongates the arabinan core during the early stages of arabinan biosynthesis in C. glutamicum.


Introduction
Tuberculosis (TB) is an infection of global significance which was responsible for over 1.3 million deaths in 2016 alone (World Health Organization, 2017).The causative pathogen, Mycobacterium tuberculosis, results in the highest number of fatalities from a single infectious agent (World Health Organization, 2017).Whilst, TB treatments have been available since the 1960s, the prevalence of the disease remains high.The long duration and unpleasant side effects of treatment regimens result in poor patient compliance, which in part may have contributed to the rise of multi-drug resistant (MDR) and extensively drug resistant (XDR) strains of M. tuberculosis.In order to address the global demand for effective TB treatments new chemotherapeutic targets are actively being explored (Boot et al., 2018).The hallmark of mycobacterial species is their impermeable, lipid-rich cell envelope that contributes to their resilience and intrinsic resistance to a range of common antibiotics.As a result, the synthesis of the intricate cell wall matrix provides excellent drug targets for some of the widely used TB drugs, including ethambutol (EMB) and isoniazid (Banerjee et al., 1994;Belanger et al., 1996;Telenti et al., 1997).The mycobacterial cell wall consists of three main components that together form the mycolylarabinogalactan-peptidoglycan (mAGP) complex: a mycolic acid layer formed of long chain fatty acids, a highly branched arabinogalactan (AG) and a peptidoglycan (PG) backbone.This macromolecular structure is further interspersed with glycolipids, such as lipomannan (LM) and lipoarabinomannan (LAM), both of which have crucial structural and immunomodulatory roles (Mishra et al., 2011).
The Emb ArafTs -EmbA, EmbB and EmbCare the known targets of the front-line anti-TB drug ethambutol (Belanger et al., 1996;Telenti et al., 1997).Deletion studies have confirmed that impairment of different emb genes results in different effects on cell wall biosynthesis in M. smegmatis.Specifically, disruption of both embA and embB affected the terminal-Ara 6 motif that serves as a template for subsequent mycolylation (Escuyer et al., 2001).Both EmbA and EmbB are predicted to act as α(1 → 3) ArafTs, but no direct evidence for this has been uncovered (Bhamidi et al., 2008).Whilst, EmbA and EmbB play a role in AG biosynthesis, disruption of embC results in reduced synthesis of LAM, and thus EmbC is suggested to add Araf residues exclusively to the arabinan domain of LAM (Shi et al., 2006).Interestingly, C. glutamicum, possesses only one emb gene that is non-essential and is involved in arabinan synthesis, the deletion of which results in a viable but slowgrowing mutant with a highly truncated AG-glycan possessing single arabinose residues attached to the galactan backbone (Alderwick et al., 2005).The ability to create arabinan-deficient strains of C. glutamicum makes it an ideal candidate for deletion studies and research into AG biosynthesis (Alderwick et al., 2006b;Alderwick et al., 2006a).
In the present study, we have investigated the biosynthesis of AG in C. glutamicum by generating a double deletion mutant of aftA and emb, which through biochemical and chemical analyses has revealed that Cg-Emb acts as an α(1 → 5) ArafT and transfers Araf residues from DPA to the Araf-primed galactan chain in C. glutamicum.

Materials and methods
Synthesis of MJ-13-77 and MJ-14-01 All reagents were purchased from commercial sources without further purification, while reaction solvents were purified using a PURE-SOLV-400 system (Innovative Technology Inc., Newburyport, MA).All reactions were carried out in oven-dried glassware under a positive pressure of argon and monitored by TLC Silica Gel 60 F 254 (0.25 mm, E. Merck) unless otherwise indicated.Plates were visualized under UV light and/or stained with a solution of p-anisaldehyde or 5% H 2 SO 4 in ethanol.Column chromatography was performed using Silicycle UltraPure silica gel (SiliaFlash® P60, 40-63 μm, Cat# R12030 B).The ratio between silica gel and crude product ranged from 100:1 to 20:1 (w/w).Optical rotations were measured in a microcell (10 cm, 1 mL) at 22 ± 2 °C and are in units of degree•mL/(g•dm).Organic solutions were concentrated under vacuum at temperature below 50 °C on a rotary evaporator. 1 H NMR spectra were recorded at 500 MHz, and chemical shifts were referenced to CDCl 3 (7.26ppm), or D 2 O (4.78 ppm). 1 H NMR data are reported as though they are first order and the peak assignments were made on the basis of 2D-NMR ( 1 H-1 H COSY and HMQC) experiments. 13C NMR spectra were recorded at 125 MHz, and 13 C chemical shifts are referenced to CDCl 3 (77.23)or external acetone (31.07,D 2 O).Electrospray mass spectra were recorded on samples suspended in mixtures of THF with CH 3 OH and added NaCl.
To a solution of compound 3 (Completo and Lowary, 2008) (1.34 g, 4.6 mmol) in acetone (40 mL) and 2,2-dimethoxypropane (20 mL) at room temperature was added p-TSA•H 2 O (13 mg, 0.07 mmol) and the solution was stirred for 90 min.The reaction was then quenched by adding four drops of triethylamine and the mixture was then concentrated on a rotary evaporator.The syrupy residue was then dissolved in pyridine (10 mL) and cooled to 0 °C followed by the addition of benzoyl chloride (1.3 mL, 11.2 mmol) dropwise.The reaction mixture was allowed to warm to room temperature and stirred overnight.The reaction was quenched by the addition of chilled water (50 mL) and extracted with CH 2 Cl 2 (60 mL).The CH 2 Cl 2 layer was washed with 10% aqueous copper sulfate solution (30 mL × 4), water (40 mL).The separated CH 2 Cl 2 layer was dried (Na 2 SO 4 ), and concentrated.The syrupy residue obtained was dissolved in a solution of acetic acid: water: THF (3:1.5:1.5;50 mL) and heated at 55-60 °C for 9 h.The reaction mixture was then directly concentrated on rotary evaporator at 50 °C and the residue was purified by column chromatography (6.5:3.5 hexanes-EtOAc) to yield 4 (1.4 g, 61% over three steps) as a thick syrup.
Bacterial strains and growth conditions C. glutamicum ATCC 13032 (the wild type strain referred for the remainder of the text as C. glutamicum) and the recombinant strains were cultivated at 30 °C in either a rich BHI medium (Difco) or a salt medium CGXII as described previously (Eggeling and Bott, 2005).Samples for cell wall analysis were prepared by harvesting cells at an optical density of 10-15 followed by a saline wash and freeze drying.

Construction of plasmids and strains
In order to generate the double deletion mutant C. glutamicumΔaftAΔemb, the deletion vector pK19mobsacBΔaftA (NCgl0185) was constructed as previously described (Alderwick et al., 2006a,b) and introduced into the previously reported C. glutamicu-mΔemb (ΔNCgl0184) strain (Alderwick et al., 2005), to generate C. glutamicumΔaftAΔemb.The chromosomal deletion of NCgl0185 was achieved using two rounds of positive selection as described previously (Schäfer et al., 1994).A similar strategy was employed to successfully generate the double deletion mutant C. glutamicumΔaftBΔaftD (Alderwick et al., 2018a).

Cell wall associated and cell wall bound lipid extraction and analysis
Cells were harvested and equivalent amounts of biomass (100 mg) were extracted using 2 ml of chloroform:methanol:water (10:10:3, v/v/ v) for 4 h at 50 °C.The organic extracts were combined with 1.75 ml of chloroform and 0.75 ml of water.The lower organic phase containing associated lipids was recovered, washed with chloroform:methanol:water (3:47:48, v/v/v) and dried.Samples were resuspended in chloroform:methanol:water (10:10:3, v/v/v) and equivalent aliquots were subjected to thin-layer chromatography (TLC) analysis using silica gel plates (5554 silica gel 60F254, Merck).Alternatively, C. glutamicum cultures were metabolically labelled at mid-logarithmic phase of growth using 1 μCi ml −1 [1,2-14 C]acetate (50-62 mCi mmol −1 (GE Healthcare) for 4 h at 30 °C with shaking, harvested and processed as described above.TLC plates were developed in chloroform:methanol:ammonium hydroxide (80:20:2, v/v/v) and cell wall associated lipids visualised with either molybdophosphoric acid (5% in ethanol; w/v) followed by heating or by autoradiography by exposure of Kodak BioMax MR film.In the latter case, labelled lipids were quantified by phosphorimaging and compared with known standards.The bound corynomycolic acids from the above delipidated extracts were released by addition of 2 ml of 5% (v/v) tetra-butyl ammonium hydroxide, followed by a 12 h incubation at 100 °C.After cooling, water (2 ml), dichloromethane (4 ml) and methyl iodide (100 μl) were added and mixed thoroughly for 30 min.The organic phase was recovered, washed repeatedly with water and resuspended in diethyl-ether.After centrifugation, the clear supernatant containing cell wall bound corynomycolic acid methyl esters (CMAMES) were dried and resuspended in dichloromethane.Equivalent aliquots were subjected to the TLC plates (5554 silica gel 60F254, Merck), developed in petroleum ether/ acetone (95:5, v/v) and lipids visualised using molybdophosphoric acid (5% in ethanol; w/v) followed by heating or by autoradiography using Kodak BioMax MR film.

Isolation of the mAGP complex and glycosyl composition of alditol acetates by gas chromatography
Cells were resuspended in phosphate-buffered saline containing 2% Triton X-100 (pH 7.2), disrupted by sonicaton and centrifuged.The pelleted material was extracted three times with 2% SDS in phosphatebuffered saline at 95 °C for 1 h, washed with water, 80% (v/v) acetone in water, and acetone, and subsequently lyophilised to yield a highly purified cell wall preparation (Besra et al., 1995).The mAGP preparations (5-10 mg) were hydrolysed using 2 M trifluoroacetic acid at 120 °C for 2 h and reduced using 100 μl of sodium borohydride solution (10 mg/ml resuspended in ethanol: 1 M ammonium hydroxide, 1:1).The obtained alditols were per-O-acetylated using 100 μl of acetic anhydride at 100 °C for 1 h before examination by gas chromatography (GC) as described previously (Alderwick et al., 2005;Besra et al., 1995).

Isolation of AG and 1 H/ 13 C-nuclear magnetic resonance spectroscopy
The mAGP preparation was subjected to base saponification to remove mycolic acids using 2% potassium hydroxide in methanol-toluene (1:1) for 48 h.The insoluble residue was recovered by centrifugation at 27,000 g.The sample was washed repeatedly with methanol and the resulting AGP treated with 75 ml of 2 M sodium hydroxide for 16 h at 80 °C.The supernatant, which contained base-solubilised AG, was recovered by centrifugation at 27,000 g for 30 min.The crude AG preparation was neutralised with acetic acid and dialysed to remove residual salts (MWCO 3500).The supernatant was diluted in cold ethanol (80%, v/v) and left at −20 °C overnight to precipitate the base-solubilised AG, which was then recovered by centrifugation and lyophilised.Nuclear magnetic resonance (NMR) spectra of samples were recorded using Bruker DMX-500.Samples were repeatedly exchanged in deuterium oxide (99.9 atom% D) with intermediate lyophilisation 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.Details concerning NMR sequences used and experimental procedures were described previously (Daffé et al., 1990).

Arabinofuranosyltransferase assays using p[ 14 C]Rpp
Membrane and a P60 cell-free wall preparation were prepared to a final concentration of 10-15 mg/ml as described previously (Birch et al., 2008;Lee et al., 1997;Seidel et al., 2007).The branched tetrasaccharide neoglycolipid acceptors, [MJ-14-01], 2 μl from a 20 mM stock solution and decaprenyl phosphate (1 µl of mg/ml) were aliquoted into 1.5 ml Eppendorf tubes and dried.IgePal™ (Sigma-Aldrich) was added (0.1%, v/v) with the appropriate amount of buffer (50 mM MOPS pH 7.9, 10 mM MgSO4, 5 mM β-mercaptoethanol) to a final volume of 80 μl.Samples were sonicated for 15 min to resuspend lipid-linked substrates and then mixed with the remaining assay components: membrane protein and 'P60' fraction (1 mg each) from either C. glutamicum, C. glutamicumΔaftA, C. glutamicu-mΔemb, C. glutamicumΔaftAΔemb, C. glutamicumΔaftBΔaftD, 1 mM ATP, 1 mM NADP, p[ 14 C]Rpp (25,000 cpm) and in some cases EMB (1 mg/ ml).Reaction mixtures were incubated for 1 h at 37 °C, quenched by the addition of 533 μl of chloroform/methanol (1:1, v/v) and mixed for 30 min.Supernatant was recovered following centrifugation at 27,000 g for 30 min and dried under nitrogen.The residue was resuspended in 750 μl ethanol:water (1:1, v/v) and loaded onto 1 ml SepPak ion exchange columns, pre-equilibrated with ethanol:water (1:1, v/v) as described previously (Lee et al., 1997).The column was washed twice with 2 ml of ethanol (100%) and the eluate collected and dried.The sample was resuspended in a mixture of water-saturated n-butanol (3 ml) and water (3 ml), mixed and the organic phase recovered following centrifugation.The aqueous phase was extracted once again with n-butanol (3 ml) and the organic phases pooled.The extracts were further washed using n-butanol-saturated water (3 ml).Finally, the nbutanol fraction was dried and resuspended in 200 μl of n-butanol.The incorporation of [ 14 C]Araf was determined by subjecting samples to TLC using silica gel plates (5735 silica gel 60F254, Merck) developed in isopropanol:acetic acid:water (8:1:1, v/v/v) and visualised by autoradiography employing Kodak BioMax MR films.

Characterisation of in vitro reaction products generated by arabinofuranosyltransferases and C. glutamicum membranes
Large-scale reaction mixtures containing cold pRpp (25 mM, Sigma Aldrich) and neoglycolipid acceptors (MJ-13-77 and MJ-14-01, 8 μl of 20 mM) were mixed and given an initial incubation of 1 h at 37 °C with membranes prepared from C. glutamicum.The assays were replenished with fresh membranes and re-incubated for 1 h at 37 °C with the entire process repeated thrice.Products were extracted from reaction mixtures by n-butanol:water phase separation and subjected to preparative TLC plates developed in isopropanol:acetic acid:water (8:1:1, v/v/v) as described above.Bands of interest were recovered from the plates by extraction with n-butanol.Samples were analysed by matrix-assisted laser desorption/ionisation-time-of-flight mass spectrometry (MALDI-TOF MS) and MS/MS as previously described (Lee et al., 1997;Shi et al., 2008).

Comparison of Emb proteins and construction of a C. glutamicumΔaftAΔemb mutant
In previous studies EMB was shown to specifically inhibit AG biosynthesis, with the precise molecular target being the products of the embCAB loci in M. tuberculosis (Telenti et al., 1997) and embRAB loci in M. avium (Belanger et al., 1996).In comparison, C. diphtheriae and C. glutamicum have only one emb gene (Alderwick et al., 2005;Cerdeño-Tárraga et al., 2003;Kalinowski et al., 2003).This discrepancy is in accordance with the notion that corynebacteria have rarely undergone extensive genome rearrangements and have maintained ancestral genome structures even after the divergence of corynebacteria and mycobacteria, resulting in a low frequency of structural alterations and gene duplications (Alderwick et al., 2005;Nakamura et al., 2003).Interestingly, the single emb gene of C. glutamicum exhibits a higher identity to embC than to embA and embB of mycobacteria (Alderwick et al., 2005).Overexpression studies of the single emb gene from C. glutamicum resulted in an increased resistance of corynebacteria to EMB strongly suggesting that this front-line anti-TB drug inhibits Emb in C. glutamicum (Radmacher et al., 2005).
The growth characteristics of both C. glutamicum and C. glutamicumΔaftAΔemb were studied in brain heart infusion media supplemented with sorbitol for osmotic stabilisation (Eggeling and Bott, 2005).Optical density (OD) measurements revealed normal growth kinetics for the wild-type C. glutamicum and a severe reduction in growth rate for C. glutamicumΔaftAΔemb.Growth of C. glutamicum was completed after 10 h (OD ∼8.0), whereas C. glutamicumΔaftAΔemb scarcely reached an OD of 0.24 after 16 h (Fig. 1B).

Characterisation of cell wall lipids from C. glutamicum and C. glutamicumΔaftAΔemb
Both C. glutamicum and C. glutamicumΔaftaΔemb strains were analysed for AG-esterified corynomycolic acids and cell wall associated glutamicum emb with its adjacent genes and the strategy to delete it using the deletion vector pK19mobsacBΔemb as previously described (Alderwick et al., 2005).The deletion vector carries 12 nucleotides of the 5′-end of emb and 12 nucleotides of its 3′-end thus enabling the in-frame deletion of almost the entire emb gene.The arrows marked as P1 and P2 locate primers used for the PCR analysis to confirm the absence of emb.The C. glutamicu-mΔemb was then employed to subsequently generate C. glutamicumΔaftAΔemb.Previously reported deletion vector pK19mobsacBΔaftA was used to delete aftA in C. glutamicumΔemb (Alderwick et al., 2006b).Distances are not drawn to scale.(B) The consequences of both aftA and emb double deletion on growth of wild type C. glutamicum (WT).lipids from an equivalent starting amount of biomass for each strain due to differences in growth rate.As expected, cell wall bound corynomycolic acids, analysed as CMAMES, were completely abolished in the C. glutamicumΔaftaΔemb mutant indicating a major defect in cell wall biosynthesis in comparison to the cell wall of C. glutamicum (Fig. 2A).Analysis of the cell wall associated lipids in the mutant highlighted an apparent increase in trehalose dicorynomycolates (TDCM) and trehalose monocorynomycolates (TMCM), the equivalent of mycobacterial TDMs and TMMs, respectively (Fig. 2B).This was verified quantitatively by labelling cultures with [ 14 C]-acetate and loading equal amounts of radioactive extractable free lipids from C. glutamicum and C. glutamicumΔaftaΔemb (Fig. 2B).Densitometry based analysis of the C. glutamicumΔaftaΔemb mutant demonstrated a significant increase in TDCMs (275%) and TMCMs (141%), when compared to the wild type strain (Fig. 2B).These cell wall lipid phenotypes are consistent with the previous studies of the single C. glutamicumΔaftA and C. glutamicumΔemb mutants (Alderwick et al., 2006b(Alderwick et al., , 2005)).The continued synthesis of TDCM and TMCM suggests that by perturbing AG biosynthesis, potential mycolation sites have been removed and these mycolate biosynthetic precursors have accumulated.

In vitro arabinofuranosyltransferase activity with membrane extracts from C. glutamicum and mutant strains
Our initial attempts to develop an in vitro assay using purified recombinant Emb from C. glutamicum have thus far proved unsuccessful due to the formation of inclusion bodies.As an alternative, we analysed ArafT activity in a cell-free assay in the presence of exogenous neoglycolipid acceptors, β-D-Galf-( 1 glutamicumΔaftAΔemb.The cell-free assay format was based on previously established ArafT assays (Birch et al., 2008;Seidel et al., 2007) and was designed to directly examine whether the Emb protein from C. glutamicum may transfer a second Araf residue in an α(1 → 5) fashion to the already Araf "primed" galactan chain.Assays were conducted both in the presence and absence of EMB in order to inhibit the single C. glutamicum Emb protein.The [ 14 C]-labelled products were extracted using organic solvents, separated by TLC and detected by autoradiography.Although, the efficiency of MJ-14-01 and MJ-13-77 were lower than our previously described ArafT acceptors (Birch et al., 2008;Seidel et al., 2007), three bands were observed migrating at R f of 0.66, R f of 0.61 and R f of 0.64 and were labelled as Products A, B and C, respectively (see Fig. 6F).Assays conducted using wild type C. glutamicum membranes and acceptor MJ-13-77 produced a TLC autoradiogram with a product profile consisting of two bands annotated as Product A (major) and Product B (minor), respectively (Fig. 6A).For acceptor MJ-14-01, a single band was observed, which was annotated as Product C (Fig. 6A).Addition of EMB to the assay mixture resulted in inhibition of Products A and C, suggesting that the enzyme adding a Araf residue onto these acceptors is EMB sensitive, and therefore most likely is added by the single Emb ArafT in C. glutamicum.These experiments also allowed the minor Product B to be visualised more clearly.A similar experiment was repeated using C. glutamicumΔaftA membranes and MJ-13-77 and MJ-14-01, and a similar profile to that of wild-type C. glutamicum was observed indicating that deletion of Cg-AftA does not participate in the generation of Products A-C (Fig. 6B).Experiments utilising membranes devoid of Cg-Emb activity from C. glutamicumΔemb and MJ-13-77 and MJ-14-01 were performed and both Products A and Product C were not generated (Fig. 6C), strongly indicating that the deleted enzyme Cg-Emb is indeed responsible for the generation of these two [ 14 C]-labelled products observed by TLC-autoradiography.Interestingly, the C. glutamicumΔemb strain was still able to produce Product B utilising MJ-13-77.Addition of EMB to the reaction mixture did not inhibit the synthesis of Product B, thus implying that ArafT adding the Araf residue to the MJ-13-77 is EMB insensitive.The in vitro experiment was repeated using C. glutamicumΔaftBΔaftD membranes and the product profile closely resembled that of the assay carried out using wild type C. glutamicum with and without EMB (Fig. 6D).The data thus suggests that neither AftB with β (1 → 2) (Seidel et al., 2007) or AftD with α(1 → 5) ArafT activities (Alderwick et al., 2018a), respectively, contribute to the Araf addition to either MJ-13-77 or MJ-14-01.The data collectively suggests that the formation of Product B is generated by AftC (Birch et al., 2008), which was shown to possess an α(1 → 3) ArafT activity and is EMB insensitive.Finally, in vitro ArafT assays were performed using C. glutamicumΔaftAΔemb membranes and the acceptors leading to formation of only Product B (Fig. 6E).The combination of visible bands on the TLC is similar to the one produced utilising C. glutamicumΔemb membranes, indicating that formation of both Products A and C is due to Emb enzymatic activity.Overall, the TLC product profiles obtained show that Products A and C were formed as a result of Emb ArafT activity transferring [ 14 C]-Araf to the MJ-13-77 and MJ-14-01 acceptors, respectively.

Mass spectrometry analysis of in vitro generated products A-C
Products A, B and C, synthesised using synthetic acceptors MJ-13-77 and MJ-14-01, non-radiolabelled pRpp and membranes from wildtype C. glutamicum, and in some cases EMB treatment were excised from preparative TLC plates, per-O-methylated and subjected to MALDI-TOF MS analysis.To rule out co-chromatography of the starting material with the products even after separation by TLC, due to similar R f values, both MJ-13-77 and MJ-14-01, were also analysed and served as a marker for any remaining unreacted acceptor (Fig. 7A, B).All three samples containing Products A, B and C have provided similar molecular ion profiles (Fig. 7C-E).Molecular ions of m/z 1099 was observed and corresponded to a pentasaccharide product(s) [Gal 3 Ara 2 - C 8 H 17 + Na] + (Fig. 7C-E) containing a newly-added Araf-(1 → ?) residue.Ions for unreacted MJ-13-77 and MJ-14-01 of m/z 939 [Gal 3 Ara 1 -C 8 H 17 + Na] + were also detected (Fig. 7C-E).In addition, all samples contained molecular ions of m/z 969, suggesting a possible per-O-methylation artefact.Overall, the MALDI-TOF MS analysis confirmed the addition of only a single Araf residue to MJ-13-77 and MJ-14-01, resulting in three pentasaccharide Products A, B and C with composition of Gal 3 Ara 2 -C 8 H 17 .
In order to define the branching pattern of the generated oligomers, the respective ions were subjected to MALDI-TOF MS/MS analysis.The fragmentation patterns of various synthetic acceptors and arabinan oligomers have been validated previously (Lee et al., 2006;Shi et al., 2008;Zhang et al., 2007).The original Gal 3 Ara 1 -C 8 H 17 acceptors provided a series of ions: O,3 A, 2,4 A, C, E, G, O,2 X and 1,4 X (Fig. 8A), consistent with expected linkages and cleavage ions of neoglycolipid acceptors.Both Product A and B provided similar MS/MS spectra with common C and E ions at m/z 463 and 415, respectively, indicating that in both cases an Araf residue has been added onto the existing Araf residue of the MJ-13-77 acceptor (Fig. 8B, C).Lack of arabinosylation at the C-2 of the Araf residue of MJ-13-77 is supported by the absence of ions 2,4 A and O,2 X of m/z 125 and 981, respectively.The Y ion of m/z 765, coupled with ions at m/z 285, 327 and 863, locates the new Araf residue at either C-3 or C-5 of the existing Araf unit of the MJ-13-77 acceptor in Products A and Product B (Fig. 8B, C).Since, earlier studies have shown that α(1 → 5) ArafT activity is EMB sensitive, this would suggest that Product A is probably formed by the addition of a new Araf residue at C-5, while Product B at C-3, as it is ethambutol insensitive (Fig. 6).In contrast, Product C (Fig. 9) is formed by Emb activity yielded the Y, O,3 A and 2,4 A ions at m/z 335, 665 and 693, respectively, indicated that the newly attached Araf residue is linked to the Araf residue of the MJ-14-01 acceptor.The ion of m/z 375 clearly places the Araf residue to the α(1 → 5) linked Araf of the acceptor, which mimics the Araf-"primed" galactan chain.Unfortunately, the diagnostic O,3 A, O,2 X, E and G ions with m/z of 257, 301, 343 and 863, respectively, that   Arabinofuranosyltransferase activity was determined using synthetic acceptors in a cell-free assay with and without ethambutol (EMB).The products of the assay were processed and subjected to TLC using silica gel plates (5735 silica gel 60F254, Merck) in isopropanol:acetic acid:water (8:1:1, v/v/v) with the reaction products visualised by either autoradiography using Kodak BioMax MR films or phosphorimaging using Molecular Imager FX (Bio-Rad).Negative controls represent reactions performed without acceptors and membranes preparations.would define the Araf residue specifically to C-5 were not unique in this spectra (Fig. 9).However, it is most likely to be added at C-5 of the MJ-14-01 acceptor due to its observed sensitivity to EMB.
Considering the TLC, MS analysis and EMB-sensitivity data altogether, it is possible to conclude that the pentasaccharide Product A is formed by the single Emb protein, which acts as an α(1 → 5) ArafT and transfers the Araf residue to the C-5 of the MJ-13-77.In contrast, the generation of pentasaccharide Product B was observed even in the presence of EMB, indicating that the ArafT is EMB insensitive, and mostly likely an α(1 → 3) ArafT (AftC), as other AftTs have been sequentially ruled out through either genetic, MS data or EMB sensitivity data (AftA, AftB and AftD) (Fig. 6A-E).Likewise, Product C results from the addition of an Araf unit to the MJ-14-01 attached at position C-5, which was also EMB sensitive (Fig. 9).Our results establish for the first time that Emb from C. glutamicum acts as α(1 → 5) ArafT and results in the polymerisation of a suitably-primed Araf-galactan chain.

Discussion
The mAGP complex is a key cell wall component of the Corynebacterianeae, which is essential for the growth and viability of M. tuberculosis.This highly complex structure is rich in lipids and sugars that act as a low permeability barrier and contributes to resistance to common antibiotics.It is therefore not surprising that potent anti-TB drugs target the biosynthesis of mAGP, including EMB and isoniazid, both of which inhibit the biosynthesis of AG and mycolic acids, respectively.Nevertheless, cases of MDR and XDR-TB pose a serious challenge to treatments offered by chemotherapy, increasing the need to discover novel drug targets and the development of new therapeutic agents against M. tuberculosis infections.Our current knowledge of ArafTs that synthesise the arabinan domains of both AG and LAM remains somewhat incomplete.Recently, a novel mycobacterial ArafT activity was described for LAM biosynthesis, but the specific enzyme remains to be identified and characterised (Angala et al., 2016).In addition, the catalytic mechanisms and complex protein interactions of how different ArafTs synthesise the bulk of the arabinan, containing α(1 → 5), α(1 → 3) and β(1 → 2) glycosidic linkages in mycobacteria and corynebacteria remain unclear.Structural and functional information on these enzymes would aid the further exploitation of ArafTs as potential drug targets to inhibit the crucial mAGP complex in M. tuberculosis.
Previously described ArafTs include the Emb proteins (Alderwick et al., 2005;Escuyer et al., 2001), AftA (Alderwick et al., 2006b), AftB (Jankute et al., 2017;Seidel et al., 2007), AftC (Birch et al., 2008) and AftD (Alderwick et al., 2018a;Skovierová et al., 2009), which all share certain functional relationships, but also possess distinct ArafTs with their own individual characteristics.A classic example is the sensitivity of Emb proteins from Mycobacterium species and Emb from C. glutamicum towards EMB, and the insensitivity of other ArafTs including AftA, AftB and AftC towards the same drug.The number of ArafTs involved in the assembly of the arabinan domain in AG remains a matter of speculation.However, the current structure of mycobacterial AG suggests at least six different ArafTs that are involved in its biosynthesis.Since, Emb proteins are the targets of EMB, a number of studies here attempted to characterise this role in AG biosynthesis.Individual EmbA and EmbB deletion mutants in M. smegmatis possessed reduced levels of the disaccharide β-D-Araf-(1 → 2)-α-D-Araf in AG resulting in a terminal non-reducing Ara 4 motif instead of the usual Ara 6 motif (McNeil et al., 1991).It was concluded that EmbA and EmbB are responsible for the 3,5 branching in AG and the synthesis of a distinct disaccharide β-D-Araf-(1 → 2)-α-D-Araf (Escuyer et al., 2001).However, direct evidence for site of action i.e. the Ara 6 domain via 3,5 branching in AG or early stages of arabinan biosynthesis has been lacking.It is intriguing also that one can delete successfully Cg-Emb and obtain a slow-growing phenotype with a highly truncated AG-glycan with single arabinose residues attached to the galactan core, whereas M. tuberculosis produces a more matured AG-glycan in both EmbA and EmbB mutants, suggesting possibly that the Emb proteins from C. glutamicum and M. tuberculosis may in fact functional differ in their arabinosyltransferase activities.
In this study, we have characterised a C. glutamicumΔaftAΔemb double deletion mutant.Analysis of its cell wall revealed an AG containing only the galactan backbone of AG with no arabinose residues (Fig. 3).In addition, the mutant led to an increase in non-covalently linked TDCMs and TMCMs and a lack of covalently bound cell wall mycolates indicating the loss of mycolylation sites in AG (Fig. 2A, B).By employing a cell-free assay using C. glutamicum, C. glutamicumΔaftA, C. glutamicumΔemb, C. glutamicumΔaftBΔaftD, C. glutamicumΔaftAΔemb strains, and neoglycolipid acceptors that resemble the Araf-primed galactan of AG, and subsequent analysis of the products formed, we have shown that the transfer of a Araf residue from DPA to the Araf-primed galactan in C. glutamicum is catalysed by the EMB sensitive Emb ArafT with α(1 → 5) activity.
The discovery of Emb from C. glutamicum as an α(1 → 5) ArafT has shed further light on the key ArafTs involved in the synthesis of AG, which may contribute to a more detailed understanding of pathogenicity and persistence of M. tuberculosis.The challenge for the future will be to define the precise role of further ArafTs in Mycobacterium species, scrutinise their catalytic mechanisms and formation of protein complexes in order to synthesise AG.

Fig. 1 .
Fig. 1. Construction and characterisation of the C. glutamicumΔaftAΔemb double mutant.(A) The C.glutamicum emb with its adjacent genes and the strategy to delete it using the deletion vector pK19mobsacBΔemb as previously described(Alderwick et al., 2005).The deletion vector carries 12 nucleotides of the 5′-end of emb and 12 nucleotides of its 3′-end thus enabling the in-frame deletion of almost the entire emb gene.The arrows marked as P1 and P2 locate primers used for the PCR analysis to confirm the absence of emb.The C. glutamicu-mΔemb was then employed to subsequently generate C. glutamicumΔaftAΔemb.Previously reported deletion vector pK19mobsacBΔaftA was used to delete aftA in C. glutamicumΔemb(Alderwick et al., 2006b).Distances are not drawn to scale.(B) The consequences of both aftA and emb double deletion on growth of wild type C. glutamicum (WT).

Fig. 8 .
Fig. 8. MALDI-TOF MS/MS analysis (A-C) of per-O-methylated Product A (B) and Product B (C).The added Araf residue is depicted in red, whereas cleavages and linkages that are diagnostic for determining the structure are depicted in blue.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 9 .
Fig. 9. MALDI-TOF MS/MS analysis of per-O-methylated Product C. The added Araf residue is depicted in red, whereas cleavages and linkages that are diagnostic for determining the structure are depicted in blue.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) . HRMS (ESI) m/z calcd for (M + Na) C 31 H 56 O 20 Na: 771.3257.Found: 771.3248.