Disruption of Mycobacterial AftB Results in Complete Loss of Terminal β(1 → 2) Arabinofuranose Residues of Lipoarabinomannan

Lipoarabinomannan (LAM) and arabinogalactan (AG) are the two major mycobacterial cell wall (lipo)polysaccharides, which contain a structurally similar arabinan domain that is highly branched and assembled in a stepwise fashion by variety of arabinofuranosyltransferases (ArafT). In addition to playing an essential role in mycobacterial physiology, LAM and its biochemical precursor lipomannan possess potent immunomodulatory activities that affect the host immune response. In the search of additional mycobacterial ArafTs that participate in the synthesis of the arabinan segment of LAM, we disrupted aftB (MSMEG_6400) in Mycobacterium smegmatis. The deletion of chromosomal aftB locus could only be achieved in the presence of a rescue plasmid carrying a functional copy of aftB, strongly suggesting that it is essential for the viability of M. smegmatis. Isolation and detailed structural characterization of a LAM molecule derived from the conditional mutant deficient in AftB revealed the absence of terminal β(1 → 2)-linked arabinofuranosyl residues. Furthermore, we demonstrated that truncated LAM displays proinflammatory activity, which is due to its ability to activate Toll-like receptor 2. All together, our results indicate that AftB is an essential mycobacterial ArafT that plays a role in the synthesis of the arabinan domain of LAM.

D espite the existence of treatments for tuberculosis (TB), TB continues to represent a major healthcare challenge, accounting for nearly 9 million new infections and over 1 million deaths each year. 1 Indeed, this global health threat is escalating, given the variable efficacy of the vaccine strain BCG, the increased susceptibility of HIV-infected individuals to TB, and the increased prevalence of multi-drug-resistant (MDR) and extensively drug-resistant (XDR) strains of Mycobacterium tuberculosis, the causative agent of the disease. 1 As current treatments lose their efficacy, it is vital that we address the increasing health burden of TB by replenishing the drug pipeline with new drugs, drug targets, and improved treatment regimes.
The cell envelope of M. tuberculosis is unique in that it contains a thick carbohydrate and lipid rich layer that accounts for its inherent resistance to numerous drugs and contributes to its persistence and virulence. This is due to the key cell envelope structures, notably the mycolyl-arabinogalactanpeptidoglycan complex (mAGP), lipomannan (LM), and lipoarabinomannan (LAM). The latter glycolipids, in addition to their structural roles, exhibit potent immunomodulatory activities and thus are probably important modulators of the host−pathogen interactions during the course of infection.
Targeting the enzymes responsible for the synthesis and assembly of the arabinan domains of AG and LAM continue to present opportunities for new chemotherapeutics. Indeed, the front-line anti-TB drug ethambutol targets the assembly of these vital cell wall structures by inhibiting the embCAB operon. 10 Individual genetic disruption of embA, embB, and embC in M. smegmatis revealed that both EmbA and EmB are α(1 → 5) Araf T that play a key role in the synthesis of the arabinan domain of AG, whereas EmbC is required exclusively for the elongation of the arabinan domain of LAM. 10,11 Recently identified α(1 → 6) Araf T, capable of utilizing synthetic (1 → 6)-Manp disaccharide and trisaccaride acceptors in vitro, was suggested to prime the mannan core. However, further analysis is required to identify the enzyme responsible. 6 Deletion studies in M. smegmatis identified Araf T with dual function −AftC− that is responsible for the α-1,3-branching of the linear arabinose polymers of AG and LAM. 12,13 In vitro assays using neoglycolipid acceptors and cell-free extracts from M. smegmatis demonstrated that AftD was able to add α(1 → 3) Araf residues to the linear α-1,5-linked acceptor, resulting in branching of the linear arabinan. It was concluded that AftD participates in the arabinan synthesis of both AG and LAM. 14 Finally, the branched arabinose motif of LAM is further modified to either a tetra-arabinoside or a hexa-arabinoside motif by an unknown β(1 → 2) Araf T.
As several Araf T's have a dual role and are involved in the arabinan synthesis of both AG and LAM polysaccharides, we investigated the potential role of AftB in LAM biosynthesis. Our previous studies in Corynebacterium glutamicum identified AftB as a nonessential Araf T responsible for the terminal β(1 → 2) linkages at the nonreducing end of the arabinan domain of AG. 15 However, the disruption and essentiality of af tB in Mycobacterium species or its essential role in LAM biosynthesis has not been reported. Herein, we demonstrate the essentiality of af tB in M. smegmatis using the conditional expressionspecialized transduction essentiality test (CESTET) and analyzed the structure of LAM from the conditional M. smegmatisΔaf tB::pMV306-af tB mutant. We have shown, for the first time, that AftB has a dual function and is responsible for β(1 → 2)-linked Araf incorporation into LAM. Moreover, we have analyzed the immunological properties of a LAM molecule isolated from the af tB conditional mutant in order to shed the light on the structure−function relationship between mycobacterial lipoglycans and host pattern recognition receptors (PRRs), notably Toll-like receptor 2 (TLR2).

■ RESULTS AND DISCUSSION
Construction and Growth of the M. smegmatis aftB Conditional Mutant. Our earlier studies demonstrated that the nonessential AftB from C. glutamicum (Cg-AftB) adds the terminal β(1 → 2)-linked Araf residues to the arabinan domain of AG before its decoration with mycolic acids. 15 In this study, we investigated the potential role of AftB in the biosynthesis of LAM in M. smegmatis. For this purpose, we constructed a knockout phage phMSMEG6400 designed to replace the chromosomal M. smegmatis af tB (Ms-af tB) with a hygromycin resistance cassette (Figure 2a). Several attempts to obtain a Msaf tB null mutant were unsuccessful, suggesting that Ms-af tB, unlike its homologue in C. glutamicum, is an essential gene. 15 We then employed CESTET and generated a Ms-af tB conditional mutant by transducing a merodiploid M. smegmatis::pMV306-af tB strain containing a second, inducible copy of af tB to yield M. smegmatisΔaf tB::pMV306-af tB. 16 The correct replacement of the native chromosomal copy of af tB with a hygromycin cassette in the transductants was confirmed by Southern blot analysis ( Figure 2b). In order to confirm af tB essentiality and to study the fate of the mycobacterial cells depleted of AftB, we monitored the growth of M. smegmatisΔaf tB::pMV306-af tB in liquid medium in the presence and absence of acetamide over 28 h (Figure 2c). Optical density (OD) measurements revealed normal growth kinetics for the mutant strain grown in media containing acetamide. However, cells cultured in the absence of inducer showed a decrease in OD after 16 h of growth and cell lysis after 28 h of incubation. This decrease in OD correlated with a decrease in the number of CFUs ( Figure 2c). Although, the af tB homologue in C. glutamicum has been reported to be nonessential, 15 the failure to obtain a af tB null mutant in the absence of second inducible copy of af tB, coupled with the inability of M. smegmatis to grow in the media without acetamide, an inducer that activates af tB expression, shows that AftB is crucial for the viability of M. smegmatis. In addition,

ACS Chemical Biology
Articles previous studies employing high density mutagenesis studies suggest that Rv3805c in M. tuberculosis is an essential gene. 17 It is reasonable to conclude that af tB is essential for M. tuberculosis viability and therefore represents an attractive drug target. Initial

ACS Chemical Biology
Articles CESTET is the restricted time scale where cells have to be cultured in the absence of acetamide long enough to deplete the intracellular protein of interest but short enough not to yield persister cells, which results in a modest yield of cell mass. In order to further analyze the different LAM preparations by NMR spectroscopy and improve the carbon signal intensity due to poor cell yield, we labeled M. smegmatisΔaf tB pMV306-af tB cultured in the presence and absence of acetamide with [ 13 C]glucose for 10 h, followed by lipoglycan extraction. Purified [ 13 C]-lipoglycans were recovered from the crude extracts by hydrophobic and gel exclusion chromatography before subjecting samples to 1 H− 13 C HSQC analysis. On the basis of our earlier studies with mycobacterial LAM 13 and previously published work, 20,21 we assigned the proton and carbon resonances of the different spin systems using 1 H− 13 C HSQC experiments. The 13 C resonances at δ101.1 ppm and δ102.3 ppm correlating to protons at δ5.11 ppm and δ4.91 ppm were assigned as 2,6-α-Manp and 6-α-Manp linkages, respectively (Figure 4a,b). The resonance at δ105.0 ppm, which correlated to an anomeric proton at δ5.04 ppm, was designated as t-α-Manp. As expected, the resonances associated with the mannan core (2,6-α-Manp, 6-α-Manp, and t-α-Manp) remained unaffected by the disruption of AftB activity and as a result were visible in both 1 H− 13 C HSQC spectra ( Figure  4a,b). The resonances associated with the arabinan domain, however, were notably more complex in the spectra of LAM isolated from the M. smegmatisΔaf tB::pMV306-af tB grown in the presence of acetamide (Figure 4a) than compared to the resonances of LAM extracted from M. smegmati-sΔaftB::pMV306-af tB grown in the absence of the inducer (Figure 4b). Several spin systems with 13 C resonances at δ110.0 ppm and δ110.4 ppm were assigned to anomeric protons at δ5.16 ppm and δ5.08 ppm, respectively, and designated as 5-α-Araf in different chemical environments. The 3,5-α-Araf residues corresponded to δ110.4 ppm and were assigned to protons at δ5.09 ppm. The 2,3-α-Araf and 2,5-α-Araf linkages corresponded to δ108.4 ppm and δ108.6 ppm with protons at δ5.24 ppm and δ5.17 ppm, respectively (Figure 4a). Finally, the 13 C resonance at δ103.5 ppm was designated to δ 5.14 ppm as the t-β-Araf linkage (Figure 4a). Most importantly, only 5-α-Araf and 3,5-α-Araf residues were conserved in the LAM of M. smegmatisΔaf tB::pMV306-af tB depleted of AftB, but 2,3-α-Araf, 2,5-α-Araf, and t-β-Araf linkages were absent due to the loss of β(1 → 2) linkages (Figure 4a,b). The 1 H− 13 C HSQC experiments confirmed that AftB acts as a β(1 → 2) Araf T in the biosynthesis of LAM in M. smegmatis. The full HSQC spectra are available in the Supporting Information (Supporting Figure 1). Further chemical characterization, such as gas chromatography analysis, could provide accurate glycosyl composition of the truncated LAM; however due to the nature of the conditional mutant and a modest yield of cell mass generated, such analysis has proved to be extremely difficult to perform.
Analysis of Toll-like Receptor 2 Activation and Cytokine Production by Truncated LAM. Both LM and LAM display immunoregulatory and anti-inflammatory properties that affect the host immune response. 22,23 A family of pattern recognition receptors, named TLR2 in association with TLR1 and TLR6, detect a wide range of ligands including mycobacterial LAM. Studies have demonstrated that the lipidic part of the molecule is required for its activity, whereas the glycosidic moiety was shown to contribute by directly controlling the extent of this activity. 24 Thus, we explored the capacity of truncated LAM extracted from the conditional M. smegmatisΔaf tB::pMV306-af tB mutant depleted of AftB to induce TLR2-dependent innate immune responses. Previous studies demonstrated that a positive relation exists between the length of the mannan chain and the ability of the lipoglycan to activate TLR2. 24 Lipoglycans with accessible long mannan domains, such as LM, were shown to be potent inducers of TLR2, whereas LAM molecules were poor activators of the TLR2 signaling due to their large arabinan domain somehow obstructing the availability of the mannan chain. 13,25,26 Crude lipoglycan preparations extracted from M. smegmati-sΔaftB::pMV306-af tB, cultured in the presence and absence of acetamide, were subsequently subjected to hydrophobic and size exclusion chromatography to yield pure LAM fractions (Figure 5a,b). 27,28 Earlier studies have reported that some of the LAM preparation from M. smegmatis may have been contaminated with lipopeptides and as a result had an immune stimulatory activity. 13,24 Therefore, we have assessed both purified truncated and full-length LAM fractions for their activity on HEK293 cells that stably express human TLR4, which is the receptor for bacterial lipopolysaccharide (LPS) and lipid A. 29 Cells were incubated with LPS, which served as a positive control, at concentrations ranging from 0.01 ng mL −1 to 1 μg mL −1 and both LAM fractions at concentrations ranging from 0.5 ng mL −1 to 10 μg mL −1 . In the presence of

ACS Chemical Biology
Articles LPS, a dose-dependent effect was observed, whereas HEK293-TLR4 cells stimulated with the purified full-length and truncated LAM resulted in similar results to those of the noninduced cells (Figure 6a). Therefore, we have concluded that the isolated LAM fractions were devoid of lipopeptide contamination. Subsequently, a derivative of HEK293 cells that stably expresses human TLR2 along with a NF-κB-inducible reporter system were incubated with full-length and truncated LAM extracted from M. smegmatisΔaf tB::pMV306-af tB grown in the presence and absence of acetamide, respectively, at concentrations ranging from 0.5 ng mL −1 to 10 μg mL −1 . The synthetic Pam 3 CSK 4 lipopeptide was used as a positive control of TLR2 activation at concentrations ranging from 0.01 ng mL −1 to 1 μg mL −1 . As shown in Figure 6b, both purified LAM molecules induced NF-κB activation in a dose-dependent fashion with the truncated LAM exhibiting a stronger TLR2 activation than compared to the full length LAM. A similar dose-dependent effect was observed when human monocyte/ macrophage THP-1 cells, which also stably express a NF-κBinducible reporter system, were incubated with full length LAM, truncated LAM, and Pam 3 CSK 4 ligands ( Figure 6c). As activation of TLR2 results in production of various cytokines and ultimately regulates the adaptive immunity, TLR2-dependent IL-8 production was measured in the supernatants of stimulated THP-1 cells (Figure 6d). Cells were incubated with full length and truncated LAM at concentrations of 2 μg mL −1 and 10 μg mL −1 and a positive reference ligand Pam 3 CSK 4 at concentrations of 100 ng mL −1 and 1 μg mL −1 . The TLR2 dependent activation of NF-κB and subsequently the production of IL-8 by truncated LAM was significantly stronger than compared to the one induced by full-length LAM ( Figure   6d). The potency of truncated LAM to induce TLR2 response in both HEK293 and THP-1 cells suggests that a more exposed mannan chain may allow TLR2 to more readily recognize the lipoglycan. This is also consistent with the previous reports. 13 In addition, we assessed if the induction of NF-κB and production of IL-8 is a direct result of TLR2 activation by using an anti-TLR2 antibody in THP-1 cells (Figures 6e, 6f). Cells were preincubated with the anti-TLR2 or isotype control antibodies (10 μg mL −1 ) for 30 min at 37°C followed by incubation with full length LAM (5 μg mL −1 ), truncated LAM (5 μg mL −1 ), or Pam 3 CSK 4 lipopeptide (1 μg mL −1 ). Both full length and truncated LAM induced activation of NF-κB ( Figure  6e) and as a result production of IL-8 (Figure 6f) that was almost completely abolished by an anti-TLR2 antibody, thus demonstrating signaling through the TLR2.
A substantial number of biological activities have been associated with phosphatidyl-myo-inositol-based glycolipids, notably LM and LAM. 22,23,30 LM, a precursor of LAM, is composed of the conserved mannosyl-phosphate inositol anchor and mannan core. It was demonstrated to carry a dual modulatory function where it acts as a strong agonist for TLR2dependent stimulation as well as exhibits TLR2 independent inhibition of cell activation and proinflammatory cytokine synthesis in murine primary macrophages. 25 Further studies revealed that the degree of LM acylation plays an important role in pro-and anti-inflammatory properties 31,32 whereas its mannan chain directly determines the lipoglycan activity. 24 Specifically, the activity of lipoglycans seemed to increase with the number of Manp units composing the mannan core. Interestingly, mycobacterial LAMs have been shown to be poor agonists of TLR2 despite containing the LM glycosidic moiety. It is unclear how, but the large arabinan domain in LAM perturbs the biological activity of the mannan core. This was further confirmed by the chemical degradation of the arabinan domain of LAM and the regain of proinflammatory properties. 26 The data presented here are in agreement with these findings as the truncated LAM isolated from the conditional af tB mutant demonstrated an increase in TLR2-dependent proinflammatory activity compared to the full-length LAM. It was suggested that the arabinan domain may exert its inhibitory effect by steric hindrance, which prevents interaction between TLR2 and the underlying mannan domain. 26 The arrangement of the arabinofuranose residues in LAM is largely dependent on the flexibility of the glycosidic linkages and puckering of the rings. It is known that ring puckering along with the more rigid β(1 → 2) linkages promote water mediated hydrogen bonding between different arabinofuranose residues. A structural study using fragments of the nonreducing termini of LAM and a murine antibody CS-35Fab has suggested that such hydrogen bonding between the arabinofuranose rings might contribute to the stabilization and orientation of the arabinan component of LAM. 33 Therefore, we hypothesize that removal of the terminal β(1 → 2) linkages can lead to a decrease in the propensity of hydrogen bonding and subsequently to a less organized and structurally compact arabinan domain as well as offer additional flexibility. This in turn can result in a more exposed "bioactive" mannan core. Further investigation is required to fully understand arabinan's role of LAM in relation to TLR2 activation. 26 ■ METHODS Plasmid Construction. The 1938 bp coding region af tB (MSMEG_6400) was amplified by PCR from M. smegmatis mc 2 155

ACS Chemical Biology
Articles genomic DNA using the primer pair (restriction sites underlined): 5′-GAT CGA TCG GAT CCG TGC GCA TCA GCC TGT GGC-3′ and 5′-GAT CGA TCA TCG ATC TAC GGT CCC GTT GCC GGC-3′. A single copy integrating plasmid pMV306-af tB was generated by ligating a 2.6 kb inducible acetamide promoter digested with XbaI-BamHI and a 1.9 kb af tB digested with BamHI-ClaI into pMV306 digested with XbaI-ClaI resulting in a pMV306 construct containing af tB gene cloned downstream of the acetamide promoter. The merodiploid strain was generated by electroporation of M. smegmatis with pMV306-af tB and selecting for kanamycin resistant colonies.
Generation of Δaf tB Conditional Mutant in M. smegmatis. Approximately 1 kb upstream and downstream, flanking sequences of MSMEG_6400 (af tB) were PCR amplified from M. smegmatis mc 2 155 genomic DNA using the primer pairs: MS6400LL (5′-TTT TTT TTC CAT AAA TTG GGA GTT ACA CCA GCA GCT ACC-3′) and MS6400LR (5′-TTT TTT TTC CAT TTC TTG GAC CAG CAC ACC ATC ATC C-3′) as well as MS6400RL (5′-TTT TTT TTC  CAT AGA TTG GAT GGG CAT GCT GGG CAT GAA CG-3′) and  MSMEG6400RR (5′-TTT TTT TTC CAT CTT TTG GCA CCG  AGA TGC CCG AGT TGT AG-3′). Following restriction digestion of the primer incorporated Van91I sites, the PCR fragments were cloned into Van91I-digested p0004S to yield the plasmid pΔMSMEG6400. The resultant plasmid was then packaged into the temperature sensitive phage phAE159 as described previously 34 to yield the knockout phage phΔMSMEG6400. Specialized transduction was performed as described earlier, 34 except that the host strain was the merodiploid strain and half of the transduction mix was spread on TSB agar plates containing kanamycin, hygromycin, and 0.2% (w/v) acetamide while the other half was spread on plates lacking acetamide. Deletion of the af tB gene was confirmed using Southern blot analysis.
Conditional Depletion of AftB. Strains to be tested were grown in either TSB or minimal medium supplemented with 0.05% (v/v) Tween-80 and 0.2% (w/v) acetamide to an OD 600 nm of 0.5. Cells were washed twice with media to remove traces of acetamide and resuspended in the original volume of appropriate media. Culture was used as a 20% inoculum in minimal media and grown for 12 h to deplete intracellular AftB. The depleted subculture served as inoculum (5%) for cultures with or without 0.2% (w/v) acetamide. Aliquots were taken and labeled with either [1, 2-14 C]-glucose 1 μCi mL −1 or [ 13 C 6 ]-glucose for 10 h. Cells were harvested, washed with phosphate buffered saline, and dried.
Cell Stimulation Assays. For stimulation assays, HEK293 and THP-1 cells were plated at 50 000 cells per well. Cells were stimulated with LAM at 0.5 ng mL −1 to 10 μg mL −1 in a final volume of 200 μL. Nonstimulated cells and cells stimulated with Pam3CSK4 (InvivoGen) or LPS (InvivoGen) served as negative and positive controls, respectively. Alkaline phosphatase activity was measured after 16−18 h by mixing 20 μL of the culture supernatant and 180 μL of Quanti-BlueTM, reading absorbance at OD 630 nm. To investigate TLR2 dependence, THP-1 cells were preincubated for 30 min at 37°C with 5 or 10 μg mL −1 of anti-TLR2 monoclonal antibody (clone T2.5, Figure 6. Immune response profiles elicited by LAM extracted from conditional af tB mutant cultured in the presence and absence of acetamide. (a) TLR4-dependent NF-κB activity by human HEK293-TLR4 cell line in response to the lipoglycans. (b) TLR2-dependent NF-κB activity by human HEK293-TLR2 cell line in response to the lipoglycans. TLR2-dependent NF-κB (c) activation and IL-8 (d) production by human monocyte/ macrophage THP1 cell line in response to lipoglycans. TLR2-dependent NF-κB (e) activation and IL-8 (f) production by human THP-1 cells in response to lipoglycans were tested using blocking antibody anti-TLR-2. Results are mean ± SD from n = 2 (a−c, e) or n = 3 (d, f) and are from one representative of two (d, f) or three (a−c, e) independent experiments. The ×10 means that the concentration of cytokine detected is amplified 10fold. *< 0.05; n.i., not induced.
Cytokine ELISA. Culture supernatants from THP-1 cells were harvested and assayed for IL-8 cytokine production using commercially available ELISA kit and according to the manufacturer's instructions (Ready Set-Go ELISA kits, eBioscience).

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acschembio.6b00898.
Materials and Methods section (PDF)