A truncated lipoglycan from mycobacteria with altered immunological properties

Results

Effects of aftC Inactivation on LM/LAM Biosynthesis.

M. smegmatis wild type (WT) (Fig. 1, lane 1) and M. smegmatis ∆aftC lipoglycans (Fig. 1, lanes 2–4) were purified using conventional methods (20) resulting in the recovery of a highly purified lipoglycan with an intermediate size between M. smegmatis LAM and LM, now termed AftC-LAM (Fig. 1, lane 3). Complementation of M. smegmatis ∆aftC with Ms-aftC restored the lipoglycan profile to WT M. smegmatis (Fig. 1, lane 5). Plasmid-borne Mt-aftC also resulted in complementation of the mutant (Fig. 1, lane 6). The molecular weight of AftC-LAM was investigated by negative-ion matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MS). The mass of WT-LAM and AftC-LAM exhibited broad unresolved peaks centered at m/z 15,000 and 8,000, respectively, indicating a weight decrease of ∼7 kDa for the mutant LAM (Fig. S1 A and B).

Fig. 1.

SDS–PAGE analysis of lipoglycans extracted from M. smegmatis and M. smegmatis ∆aftC. Lane 1, lipoglycans extracted from M. smegmatis; lane 2, lipoglycans extracted from M. smegmatis ∆aftC; lane 3, purified AftC-LAM from M. smegmatis ∆aftC; lane 4, purified LM from M. smegmatis ∆aftC; lane 5, lipoglycans extracted from M. smegmatis ∆aftC pMV261-Ms-aftC; and lane 6, lipoglycans extracted from M. smegmatis ∆aftC pMV261-Mt-aftC.

Structural Characterization of AftC-LAM.

The ratio of Ara to Man in WT-LAM and AftC-LAM was determined using gas chromatography (GC) of alditol acetate derivatives (11) (Fig. S2 A and C). WT-LAM had a molar ratio of Ara:Man of 2.7:1, which is consistent with previously reported data (22), whereas, for AftC-LAM, the Ara:Man ratio was 0.59:1 (Fig. S2 A and C). Complementation of M. smegmatis ∆aftC with either Ms-aftC or Mt-aftC restored the Ara:Man ratio to that of the wild-type LAM (Fig. S2 E and G). In M. smegmatis, LAM consists of approximately 71 Ara, 27 Man, and 1 Ins units (22, 23). The loss of 7 kDa for AftC-LAM equates to 45 Ara residues, suggesting that AftC-LAM contains one or more short arabinan domains of up to ∼16 Ara residues. The ^{H-NMR spectrum of WT-LAM (}Fig. 2A) was much more complex than the anomeric region of AftC-LAM (Fig. 2B). Indeed, the AftC-LAM 1D ^{H spectrum exhibits three major, well-defined resonances characterized by several overlapping resonances arising from six different classes of glycosidic residues. On the basis of our data for WT-LAM and previously published work (}19), the ^{C resonance at δ101 ppm that correlated to an anomeric proton at δ5.15 ppm with a} J_H1,C1 coupling constant of ≈170 Hz was assigned as 2,6-Manp. The resonances at δ105 and δ102.3, correlating to protons at δ5.07 and δ4.90, were assigned as t-Manp and 6-Manp, respectively. The t-β-Araf residues corresponded to δ103.4 with ^{H at δ5.16. The well-separated spin systems for 2-α-Ara}f attached to the 3-position (2-α-Araf→3, δ108.2, δ5.27 ppm) and the 5-position (2-α-Araf→5, δ108.5, δ5.20) of the 3,5-Araf were also visible in the spectra of WT-LAM. Several spin systems were observed (δ110.3 and δ5.19 ppm, δ110.3 and δ5.14 ppm, and δ110.3 and δ5.11 ppm) assigned to 5-α-Araf in different chemical environments with one overlapping set of 3,5-α-Araf (δ110.2-δ5.12) for WT-LAM. As reported, two distinct chemical shifts of t-α-Araf could occur for the respective arms of a branched 3,5-α-Araf, as observed for 2-α-Araf→3 and 2-α-Araf→5 (19). In AftC-LAM (Fig. 2B), the resonances associated with the mannan core (t-Manp, 6-Manp, and 2,6-Manp) are conserved, but resonances associated with Araf residues are notably less abundant. In particular, 2-α-Araf→3 is absent and the complex set of signals for the different 5-α-Araf are much more simplified due to the loss of 3,5-α-Araf (Fig. 2B). GC/MS of per-O-methylated alditol acetate derivatives confirmed this result in that it showed an unaltered glycosidic linkage profile for the mannan backbone of both WT- and AftC-LAM and a complete loss of 3,5-Araf and a significant reduction in t-Araf, 2-Araf, and 5-Araf-linkages for the latter LAM (Fig. S2 B and D). Complementation of M. smegmatis ∆aftC with either Ms-aftC or Mt-aftC restored the glycosyl linkage profile to that of WT (Fig. S2 F and H), demonstrating that MSMEG2785 (Ms-aftC) and Rv2673 (Mt-aftC) are functional orthologs. Overall, the compositional analysis suggests that, as compared to WT-LAM, AftC-LAM has an unaltered mannan domain composed of an α(1→6)-Manp backbone substituted by t-Manp units at O-2 positions, to which is attached one or more short α(1→5)-linked linear arabinan chains (in total representing ≈12–15 Araf units) terminating in a single β(1→2) Ara residue (Fig. S3).

Fig. 2.

Two-dimensional NMR spectra of WT-LAM and AftC-LAM purified from M. smegmatis and M. smegmatis ∆aftC. Structural characterization of WT-LAM (A) and AftC-LAM (B). ^{H, C HSQC NMR spectra were acquired in D}₂O at 313K. Expanded regions (δ ^{H: 5.0–5.30; δ C: 101–111) are shown.}

Effect of Ethambutol on AftC-LAM Formation.

EMB blocks cell-wall arabinan biosynthesis by targeting the Emb proteins (7, 8). Although EmbC is known to be involved in LAM arabinan biosynthesis (22), its precise role is as yet inconclusive. We exploited the phenotype of M. smegmatis ∆aftC and the structural properties of AftC-LAM to investigate this role. For this, we treated [^{C]-glucose-labeled cultures of} M. smegmatis with subinhibitory concentrations of EMB, extracted the lipoglycans, and performed a radiochemical quantification of [^{C]-incorporated sugars. SDS–PAGE analysis of the [C]-LM/LAM pool of WT} M. smegmatis revealed two broad bands corresponding to LAM and LM migrating to their expected sizes (Fig. 3A, lane 1) and an Ara:Man ratio of 1.2:1 (Fig. 3B, lane 1). Subsequent analysis of the [^{C]-LM/LAM lipoglycan pool from} M. smegmatis ∆aftC resulted in labeling of AftC-LAM and LM (Fig. 3A, lane 2) and an Ara:Man ratio of 0.5:1 (Fig. 3B, lane 2). Addition of EMB at 0.5 μg·ml ^{resulted in a decrease in the size of the AftC-[C]-LAM, an accumulation of [C]-LM (}Fig. 3A, lane 3), and a corresponding change in the Ara:Man ratio of 0.3:1 (Fig. 3B, lane 3). Because AftC-LAM contains an intact mannan core, the only possible effect of EMB is the direct inhibition of EmbC activity [addition of the terminal β(1→2) Araf residues by AftB is insensitive to EMB (9)], thus showing its involvement in the very early stages of LAM arabinan biosynthesis and inhibition of an α(1→5) AraT. Notably, we observed that residual [^{C]-Ara labeling always remained upon EMB treatment (}Fig. 3B, lane 3), suggesting that another AraT possibly adds the first Araf residue to LM, akin to AftA in AG biosynthesis (9).

Fig. 3.

SDS–PAGE (A) and total sugar analysis (B) of [^{C]-labeled lipoglycans extracted from} M. smegmatis and M. smegmatis ∆aftC treated with EMB. Growing cultures of M. smegmatis and M. smegmatis ∆aftC were labeled with [^{C]-glucose, and their lipoglycans were extracted and analyzed by SDS–PAGE (}A). Lane 1, M. smegmatis; lane 2, M. smegmatis ∆aftC; and lane 3, M. smegmatis ∆aftC + 0.5 μg·ml ^{EMB. The above lipoglycans were hydrolyzed in 2 M TFA and analyzed by TLC to determine the total sugar composition (}B). Lane 1, M. smegmatis; lane 2, M. smegmatis ∆aftC; and lane 3, M. smegmatis ∆aftC + 0.5 μg·ml ^EMB.

AftC-LAM Displays Proinflammatory Properties.

LMs, in contrast to LAMs, are potent proinflammatory lipoglycans (24–27). This difference has been attributed to the presence of the arabinan domain in LAM, which masks the “bioactive” mannan core (24, 28). To investigate the consequence of the aftC mutation on the proinflammatory activity of LAM, we compared WT- and AftC-LAM for release of TNF-α by human THP1 cells. Consistent with earlier studies (24), truncated AftC-LAM, as compared to WT-LAM, exhibited an increased proinflammatory activity (Fig. 4A). LM is a known agonist for TLR2 (29). Therefore, we investigated the ability of WT- and AftC-LAM to activate this receptor. HEK293 cells expressing TLR2 were stimulated with increasing amounts of LAM after which TLR2-dependent IL-8 production was determined. As shown in Fig. 4B, AftC-LAM induced a much stronger TLR2 activation than did WT-LAM (∼10-fold). This result demonstrates that the presence of the full-length arabinan domain somehow hampered TLR2 activation. Previously, it has been reported that immune stimulatory activity of some M. smegmatis LAM preparations may have been caused by lipopeptide contamination (26). To investigate this issue, both WT and AftC-LAM were pretreated with H₂O₂ [a procedure that inactivates lipopeptides (30, 31)], after which they were retested for their activity on HEK293 TLR2 cells. As shown in Fig. 4C, pretreatment with H₂O₂ substantially reduced the activity of both WT- and AftC-LAM. However, in contrast to WT-LAM, which lost all of its activity, a substantial part of the activity arising from AftC-LAM was sustained for up to 168 h of treatment (Fig. 4C). This was not due to the inactivation of the H₂O₂ itself because addition of fresh H₂O₂ after 96 h did not further reduce the activity (Fig. 4C). Therefore, we conclude that both LAM preparations were probably contaminated with lipopeptides to a certain extent. Nevertheless, inactivation of these molecules with H₂O₂ clearly demonstrated that AftC-LAM was able to activate TLR2, whereas WT-LAM became completely inactive for concentrations up to 100 μg·ml ⁽Fig. 4D).

Fig. 4.

TNF-α production by human THP-1 cells and IL-8 production by HEK293 TLR-2 cells in response to WT-LAM and AftC-LAM. (A) Human THP-1 cells were incubated with 1 μg·ml ^{(solid bars) or 10 μg·ml (open bars) WT- or AftC-LAM. TNF-α was quantified after 24 h by ELISA. (}B–D) HEK293 cells transfected with TLR2 were stimulated with (B) increasing amounts of WT- or AftC-LAM with LPS (50 ng·ml^{), with Pam}₃CSK₄ (50 ng·ml^{), or with H}₂O as a negative control; (C) 50 μg·ml ^{of WT- or AftC-LAM [with and without 1% hydrogen peroxide (±H}₂O₂) for the indicated time period] with LPS (50 ng·ml^{), with Pam}₃CSK₄ (50 ng·ml^{) (±H}₂O₂ for 48 h), or with H₂O as a negative control; and (D) with increasing amounts of WT- or AftC-LAM (treated with H₂O₂ for 168 h). In all cases, cells were stimulated for 24 h at 37 °C, after which the supernatants were harvested and analyzed for IL-8 by ELISA. (E) Binding of DC-SIGN–Fc to WT- (▴) or AftC-LAM (■) as determined by ELISA. Absorption was measured at 490 nm. In all panels, data are expressed as the mean ± SD from one representative of three independent experiments.

To determine whether the increased proinflammatory activity of AftC-LAM indeed coincided with a more exposed mannan domain, the ability of WT- and AftC-LAM to interact with a human immunoglobulin G Fc construct harboring the extracellular domain of the C-type lectin dendritic-cell-specific ICAM-3-grabbing nonintegrin (DC-SIGN) was investigated. DC-SIGN is highly expressed on dendritic cells and recognizes high-mannose structures (32, 33) including LM (34) but not PILAM (35). As shown in Fig. 4E, the reactivity of DC-SIGN–Fc toward AftC-LAM was stronger as compared to the reactivity against WT-LAM, demonstrating that the mannan core of AftC-LAM was more accessible.

Discussion

The data described in this article unequivocally demonstrate that AftC is responsible for introducing 3,5-Araf branching in the LAM arabinan domain of M. smegmatis and proposes a previously undescribed model for LAM biosynthesis (Fig. 5). Furthermore, we show that EmbC is involved in the very early stages of LAM arabinan biosynthesis and that, with the truncation of the arabinan domain, LAM gains the ability to activate TLR2. Although the structure of the arabinan domain of LAM is well understood, less is known of the enzymes involved in its biosynthesis. One important reason is that Corynebacterium glutamicum, a preferred model organism to study mycobacterial AG and LM biosynthesis (9–11), does not produce a convoluted LAM, as present in mycobacteria. During our investigation of various putative GT-C glycosyltransferases, we deleted msmeg2785 (AftC) from M. smegmatis, which resulted in a phenotype that displayed a severely truncated AG (11). In the present study, we revisited this mutant in an attempt to investigate the potential role of AftC in LAM biosynthesis. Chemical analysis revealed that LAM isolated from the aftC mutant contained an unaltered mannan core, but with one or more simple arabinan moieties of ≈12–16 Araf units composed of α(1→5) linkages terminating in a single β(1→2) fashion. It is reasonable to conclude from our data that AftC is not involved in early arabinan LAM biosynthesis because a more pronounced truncation of LAM would then be expected. AftC has dual functionality in terms of its involvement in the biosynthesis of both AG and LAM. Attempts to generate an aftC-deficient strain of M. tuberculosis proved unsuccessful (36), highlighting the essentiality of aftC in M. tuberculosis and the species’ intolerance to cell-wall changes. Hence, AftC represents an excellent drug target.

Fig. 5.

Mycobacterial LAM biosynthesis and the role of AftC. The GT-A/B family of glycosyltransferases performs sequential glycosidic transfer of mannose residues using the high-energy nucleotide GDP-Man to a PI-based anchor in the cytoplasm. The GT-C family of glycosyltransferases then continue LM and LAM biosynthesis by elaborating Ac₁PIM₄ with mannose and arabinose residues. MptA, MptB, and Rv2181 are the only known ManTs involved in core mannan biosynthesis, for which they use the lipid-linked substrate polyprenylphosphomannose (PPM) as the sugar donor. Decaprenyl-1-monophosphoarabinose (DPA) is used by an unknown AraT to prime LM at an as-yet-unidentified position. EmbC and AftC are the other two identified AraTs, which mature the final LAM molecule before species-specific capping occurs.

EmbC has long been implicated in the biosynthesis of mycobacterial LAM and has been shown to be a target of the front-line drug EMB (21, 22). However, speculation around its precise enzymatic function has remained controversial. By using the unique phenotype of M. smegmatis ΔaftC, we provide unequivocal evidence that EmbC is an α(1→5) AraT, which is inhibited by EMB (Fig. 3). We now propose a working model, whereby an as-yet-unidentified GT-C AraT primes the mannan backbone with singular Araf residues in much the same way as AftA primes the galactan backbone of AG (9). After this priming, EmbC extends the arabinan chain in an α(1→5) fashion before AftC branches the polysaccharide with an α(1→3) Araf residue, and the domain is further matured by a combination of AraTs, including AftC, AftB, and perhaps other as-yet-unidentified enzymes (Fig. 5).

Several mycobacterial cell-wall lipoglycans are proinflammatory stimuli of the host immune response. Both LM and LAM are members of a family of mycobacterial lipoglycans that all contain a conserved MPI anchor (29). Interestingly, LM, in addition to being a potent inducer of proinflammatory cytokines, also displays strong anti-inflammatory properties and represses lipopolysaccharide (LPS)-induced cytokine responses in macrophages (25). Recently, by making use of LM preparations, separated by their degree of acylation, it was demonstrated that these pro- and anti-inflammatory properties are, at least in part, dependent on the degree of acylation (27, 37). Whereas, tri- and tetra-acylated LM forms displayed strong proinflammatory properties, mono- and di-acylated molecules did not (37). In contrast, di-acylated LM was found to be the most potent inhibitor of LPS-induced cytokine responses, a feature that was independent of TLR2, mannose receptor, or SIGNR1 (27). In addition to the degree of acylation, additional structural features of MPI-anchored lipoglycans determine their biological activity. Recently, it was shown that the chain length of the mannan core of MPI-anchored lipoglycans directly correlated with their ability to activate TLR2 (26). Whereas lipoglycans with short oligomannopyranosyl backbones, such as the PIMs, were marginally active, the TLR2-activating potency of lipoglycans with longer backbones, i.e., LM, was strongly enhanced (26). In addition to mannan chain length, the type of substitution of the mannan core was found to be critical, with activity being retained in the case of Manp substitutions but absent when the core was substituted with Araf residues (26). This was evidently clear for mycobacterial LAMs, which carry a bulky arabinan domain. Although their LM core would in principle allow for TLR2 activation, the LAMs were found to be inactive (26). The assumption that the arabinan domain is directly involved in “silencing activity” is further sustained by the observation that chemical degradation of the arabinan domain restores the proinflammatory properties of LAM (24). Our results are in agreement with these findings as AftC-LAM, which expresses a severely truncated arabinan domain, exhibited an increased proinflammatory activity (Fig. 4). Because WT and AftC-LAM did not differ in their pattern of acylation or core mannan composition, the increased proinflammatory activity must be directly related to the shortening of the arabinan domain. Exactly how the arabinan domain prevents TLR activation is currently not fully understood. One thought is that the arabinan domain exerts its inhibitory effect by steric hindrance (24). Although several lines of evidence point to this direction, the exact mechanism and the level at which this occurs remains unclear. Furthermore, Nigou and coworkers (26) demonstrated that M. bovis bacillus Calmette–Guérin LAM complexed with soluble CD14 efficiently competes with Pam₃CSK₄ for binding to TLR2, suggesting that binding to, and signaling through TLR2, may not be equal. In several mycobacterial species, the nonreducing termini of the branched arabinan domain are modified by capping motifs consisting of either oligomannosyl, i.e., ManLAMs, or PILAM units such as in M. smegmatis (16). Whereas reports on the immune stimulatory properties of ManLAMs are consistent and unambiguously demonstrate that these types of LAMs are inactive (24, 26), some controversy exists on the activity of PILAMs. Early reports demonstrate that PILAM is proinflammatory and signals via TLR2 (17, 24, 38–40). However, the mechanism by which a low degree of inositol–phosphate capping would confer this proinflammatory activity has never been explained. One hypothesis is that the myo–inositol–phosphate motif in the MPI-anchored lipoglycans may mimic the inositol–phosphate caps on PILAMs, thereby conferring activity. However, the notion that the unsubstituted phospho-inositol (PI) caps are unlikely to mimic the acylated MPI anchor, plus the observation that PIMs are only marginally active, clearly demonstrates that the PI motif alone does not confer activity (24). Interestingly, in a more recent study, Nigou et al. (26) demonstrated that highly purified PILAMs from M. smegmatis and M. fortuitum were equivalent in activity to ManLAMs. Therefore, in contrast to what was previously suggested, the authors concluded that the presence of PI caps did not make LAM proinflammatory, and they suggested that the activity in earlier experiments was probably due to lipopeptide contamination. Even though contaminant lipopeptides may not be detected by the current analytical methods, trace amounts of these molecules (<0.1%) are enough to influence the outcome of much more sensitive biological assays. We therefore determined the effect of H₂O₂ treatment on the TLR2-stimulating activity of both WT and AftC-LAM. As shown in Fig. 4C, treatment with H₂O₂ severely reduced the activity of WT-LAM (PILAM), suggesting that the earlier observed activity was indeed due to lipopeptide contamination. In contrast, although the activity of AftC-LAM was also substantially reduced, a stable part of the activity was insensitive to H₂O₂ treatment (Fig. 4C). Although we cannot exclude that H₂O₂ also acts on LAM and may thereby abrogate its activity, the observation that the residual activity of AftC-LAM stabilized after 4 days of treatment, instead of gradually declining further, argues against this notion. Furthermore, H₂O₂ treatment fully inactivated the lipopeptide Pam₃CSK₄, demonstrating that inactivated lipopeptides do not display any residual activity (Fig. 4C). Overall, these data strongly suggest that the proinflammatory activity observed in some PILAM preparations was indeed the result of contaminating lipopeptides. Nevertheless, on the basis of our experiments, we conclude that truncation of the arabinan domain by the aftC mutation resulted in an increased proinflammatory activity of AftC-LAM.

Materials and Methods