The Cell Surface Receptor DC-SIGN Discriminates between Mycobacterium Species through Selective Recognition of the Mannose Caps on Lipoarabinomannan*

Interactions between dendritic cells (DCs) and Mycobacterium tuberculosis , the etiological agent of tuberculosis, most likely play a key role in anti-mycobac-terial immunity. We have recently shown that M. tuberculosis binds to and infects DCs through ligation of the DC-specific intercellular adhesion molecule-3-grabbing nonintegrin (DC-SIGN) and that M. tuberculosis man-nose-capped lipoarabinomannan (ManLAM) inhibits binding of the bacilli to the lectin, suggesting that ManLAM might be a key DC-SIGN ligand. In the present study, we investigated the molecular basis of DC-SIGN ligation by LAM. Contrary to what was found for slow growing mycobacteria, such as M. tuberculosis and the vaccine strain Mycobacterium bovis


SIGN binds poorly to the PILAM and uncapped AraLAMcontaining species Mycobacterium fortuitum and
Mycobacterium chelonae, respectively. Interestingly, smooth colony-forming Mycobacterium avium, in which ManLAM is capped with single mannose residues, was also poorly recognized by the lectin. Altogether, our results provide molecular insight into the mechanisms of mycobacteria-DC-SIGN interaction, and suggest that DC-SIGN may act as a pattern recognition receptor and discriminate between Mycobacterium species through selective recognition of the mannose caps on LAM molecules.
The interaction between Mycobacterium tuberculosis and host dendritic cells (DCs) 1 is thought to be critical for mounting a protective anti-mycobacterial immune response and for determining the outcome of infection (1)(2)(3)(4). However, the molecular bases of DC infection by mycobacteria remain poorly understood. We have recently shown that M. tuberculosis, as well as the vaccine strain Mycobacterium bovis bacillus Calmette-Guérin (BCG), use the DC-specific intercellular adhesion molecule-3 (ICAM-3)-grabbing nonintegrin (DC-SIGN) to bind to and enter human DCs (5), a feature that may allow the bacillus to persist within a unique immature compartment of the cells (6). DC-SIGN/CD209 is a calcium-dependent (C-type) transmembrane lectin that contains a single carbohydrate recognition domain at its extracellular C-terminal end. It is expressed on DCs as well as on some macrophage (M) subsets, including alveolar Ms (7,8). DC-SIGN has been described initially as a receptor for ICAM-3 and ICAM-2, as well as for human and simian immunodeficiency viruses, enabling the trans infection of susceptible CD4 ϩ T lymphocytes by these viruses (9 -12). Thereafter, it was shown to bind to other microbes, namely Ebola virus and Leishmania pifanoi (13,14).
The DC-SIGN carbohydrate recognition domain binds to mannose-rich glycoconjugates (15), a feature that is consistent with our finding that M. tuberculosis lipoarabinomannan (termed ManLAM; see below), a highly mannosylated surface lipoglycan, might be a key mycobacterial ligand for DC-SIGN (5). Indeed, purified M. tuberculosis-derived ManLAM was found to inhibit the binding of M. tuberculosis to human monocyte-derived DCs, as well as to recombinant HeLa-derived cells expressing DC-SIGN. LAM is a major component of the mycobacterial cell wall. It contains a carbohydrate backbone composed of D-mannan and D-arabinan (Fig. 1). The mannan core is attached to an acylated mannosylphosphatidylinositol (MPI) anchor at its reducing end; the arabinan domain is capped with either mannose residues in so-called ManLAMs or with phosphoinositide motifs in so-called PILAMs (16,17), or uncapped in so-called AraLAM (40). The caps of ManLAMs consist of mono-, ␣(132)-di-, and ␣(132)-tri-mannopyranosides, with di-mannopyranosides being the most abundant motif (18 -20). So far, ManLAMs have been detected in slow growing mycobacteria only. These include various strains of M. tuberculosis, M. bovis BCG, the leprosy agent Mycobacterium leprae, and the opportunistic species Mycobacterium avium (16,17,21,22). By contrast, PILAM or AraLAM expression seems to be fairly limited to fast growing mycobacteria, including nonpathogenic Mycobacterium smegmatis (16,17,21,23), Mycobacterium chelonae (40), and Mycobacterium fortuitum. 2 In addition to their structural role in organizing of the cell wall, LAMs are known to be potent inducers of various cytokines when in contact with mammalian phagocytic cells (24).
We have shown that the slow growing mycobacteria M. tuberculosis and M. bovis BCG, which express ManLAM, interact with human DCs through DC-SIGN and that purified M. tuberculosis-derived ManLAM inhibits M. tuberculosis binding to recombinant HeLa-derived cells expressing DC-SIGN and to monocyte-derived DCs. The goal of the present report was to obtain a better understanding of the molecular determinants of the LAM molecule involved in binding to DC-SIGN. Using a DC-SIGN-expressing recombinant cell line as a readout, we first report that DC-SIGN binds poorly to the fast growing species M. smegmatis and that M. smegmatis-derived PILAM, which lacks mannose caps, exhibits a very limited ability to inhibit M. tuberculosis binding to the lectin. Using various chemically or enzymatically generated variants of the ManLAM molecule, we further demonstrate that both the acyl chains on the MPI anchor and the mannose-capping residues play a key role in the ManLAM-DC-SIGN ligation process. Moreover, we show that DC-SIGN does not bind to the PILAMand AraLAM-containing species M. fortuitum and M. chelonae, respectively. Altogether, our findings provide evidences that DC-SIGN may discriminate between ManLAM-containing slow growers, such as M. tuberculosis, and nonpathogenic PILAMcontaining fast growers, such as M. smegmatis, through a high affinity for mannose-capping residues on ManLAM.
Mycobacterial Glycoconjugate Purification and Chemical Degradation-ManLAM from M. bovis BCG Pasteur and PILAM from M. smegmatis were purified as described previously (19,23,26). M. tuberculosis H37Rv-purified ManLAM was a kind gift from the Colorado State University. dManLAM was prepared by incubating ManLAM (200 g) in 200 l of NaOH 0.1 M for 2 h at 37°C. After neutralization with 200 l of HCl, 0.1 M, the reaction products were dialyzed against water and freeze-dried. ␣ManLAM was prepared by incubating ManLAM (200 g) for 6 h at 37°C in 30 l of a jack beans ␣-mannosidase (Sigma) solution (2 mg/ml, 0.1 M sodium acetate buffer, pH 4.5, 1 mM ZnSO 4 ). After a second addition of 50 l of enzyme solution, the reaction was continued overnight. The reaction products were then dialyzed against 50 mM NH 4 CO 3 , pH 7.6. Elimination of ␣-mannosidase was achieved by denaturation (2 min at 110°C) followed by overnight tryptic digestion (37°C, 3.2 g of trypsin). ␣ManLAM was recovered after dialysis against water, freeze-dried, and analyzed for cap contents by CE-LIF as previously described (27). Briefly, ManLAM or ␣ManLAM (1 g) was submitted to mild acidic hydrolysis (15 l of HCl, 0.1 M, for 20 min at 110°C) in the presence of mannoheptose (100 pmol) as the internal standard. The reaction products were then submitted to 1-aminopyrene-3,6,8-trisulfonate (APTS) tagging and subjected to CE-LIF analysis. Separations were performed using an uncoated, fused silica capillary column (50 m internal diameter; 40 cm effective length; 47 cm total length; Sigma), and analyses were carried out at a temperature of 25°C with an applied voltage of 20 kV using acetic acid 1% (w/v), triethylamine 30 mM in water, pH 3.5, as a running electrolyte. The amount of each cap motif was determined relative to the internal standard.
Binding and Inhibition Assay-Cells were infected at a multiplicity of infection of 1 bacterium/cell for 4 h at 4°C in RPMI 1640, washed extensively in RPMI 1640, and analyzed by scoring colony-forming units after plating on agar and incubation at 37°C. In binding inhibition experiments, cells were preincubated for 1 h at 4°C with 10 g/ml of the indicated components. These components were left in the culture medium during the infection process.

RESULTS AND DISCUSSION
Mycobacterial species can be divided into slow and fast growers. To gain a better understanding of the molecular basis of their ligation to DC-SIGN, we first compared the relative ability of the slow growing pathogenic species M. tuberculosis versus the fast growing saprophytic species M. smegmatis to bind to the lectin. A binding assay was performed on HeLa-derived cells expressing or not DC-SIGN (HeLa-DC and HeLa, respectively). As we reported previously, M. tuberculosis was found to bind to HeLa-DC ϳ15 times more than to HeLa cells ( Fig. 2A), and in a multiplicity of infection-dependent manner (data not shown). Conversely, M. smegmatis binding to HeLa-DC was found to be only ϳ2 times more than to HeLa cells ( Fig. 2A). Our previous finding that M. tuberculosis ManLAM can inhibit M. tuberculosis binding to DC-SIGN suggests that the reduced ability of M. smegmatis to bind to HeLa-DC cells may be due to the inability of M. smegmatis PILAM to bind to the lectin. To test this hypothesis, we performed a M. tuberculosis binding assay on HeLa-DC cells that had been preincubated or not with LAM molecules from various mycobacterial species. As reported previously, yeast mannan and M. tuberculosis-as well as M. bovis BCG-derived ManLAMs were found to inhibit mycobacterial binding to HeLa-DC cells by as much as ϳ90% (Fig.  2B) (18).
Because DC-SIGN is a mannose-binding lectin and because PILAMs are devoid of mannose caps (Fig. 1), we next reasoned that the results described above might indicate that ManLAM capping residues may be the ManLAM subdomains preferentially recognized by the lectin. To test this hypothesis, we treated M. bovis BCG-derived ManLAM with ␣-exomannosidase to obtain ManLAM devoid of mannose caps (␣ManLAM). The reaction was assessed by CE-LIF analysis as previously described (18,27). A typical electropherogram obtained for 2 L. Bala, M. Gilleron, M. Rivière, and G. Puzo, unpublished data.

Mycobacterial Binding to DC-SIGN 5514
␣ManLAM is presented in Fig. 3A. Peaks corresponding to mannooligosaccharide caps, i.e. mono-, ␣(132)-di-, and ␣(132)-tri-mannopyranosides, were almost undetectable. Quantification indicated that more than 95% of cap demannosylation was achieved. The ability of ␣ManLAM to inhibit M. tuberculosis binding to DC-SIGN was evaluated in binding experiments. In contrast to native ManLAM, ␣ManLAM failed to inhibit mycobacterial binding to the lectin (ϳ10% binding inhibition, Fig. 3B). Similar results were obtained when cells were treated with M. tuberculosis-derived ␣ManLAM prior to the binding assay (data not shown). These results indicate that mannooligosaccharide caps are critical structural features for ManLAM-mediated inhibition of M. tuberculosis binding to DC-SIGN.
Because MPI anchor has been shown previously to be involved in some of the biological activities of ManLAM, particularly their binding to C-type lectins (28,29), we then evaluated the role of the acyl part of the MPI anchor in ManLAM-DC-SIGN interaction. To this end, M. bovis BCG-derived dManLAM was prepared by alkali treatment. As shown on Fig.  3B, dManLAM failed to inhibit M. tuberculosis binding to HeLa-DC cells, revealing that a native acylated MPI anchor is required for ManLAM-mediated inhibition of mycobacterial binding to DC-SIGN.
Finally, we wished to know whether our finding was still valid in other Mycobacterium species for which the LAM struc-ture was known. In agreement with what we reported above, the PILAM-and AraLAM-containing species M. fortuitum and M. chelonae were poorly recognized by DC-SIGN. Indeed, in a representative binding assay done in triplicate, M. fortuitum and M. chelonae bound to HeLa-DC cells 2.3 and 1.5 times more than to HeLa cells, respectively (data not shown). Interestingly, the ManLAM-containing slow grower M. avium was also found to bind poorly to DC-SIGN-expressing HeLa cells (ϳ1.7 times more than to HeLa cells; data not shown). This is not surprising because ManLAM from smooth colony-forming M. avium, which is the one used in our assay, has been reported to be capped mainly with single mannose residues instead of the di-and tri-mannopyranoside motifs found in M. tuberculosis-and M. bovis BCG-derived ManLAM (22). Because DC-SIGN does not bind to single mannose molecules but to more complex mannosylated structures (15), it is likely that such mono-mannosylated ManLAM is not recognized by the lectin. These results raise the possibility, currently under investigation, that DC-SIGN may recognize mycobacteria from the tuberculosis complex only.
Altogether, our results demonstrate that the DC-SIGN-Man-LAM interaction involves both the MPI anchor acyl chains and the mannose residues from caps of the ManLAM molecule. As established recently for the binding of ManLAM to the human surfactant pulmonary protein A C-type lectin (29,30), the MPI fatty acids are most likely involved in the supermolecular organization of the ManLAM molecules in aggregates, allowing macromolecular clustering in aqueous solution. Micelle formation probably results in a huge increase in ManLAM valence and increases ManLAM avidity to DC-SIGN. This is likely to explain the poor ability of dManLAM to inhibit M. tuberculosis Mycobacterial Binding to  binding to HeLa-DC cells but does not indicate whether LAM acyl chains, which are likely to be buried within the bacterial cell wall, can interact with the lectin in vivo. However, the latter definitely should be investigated, as our result is reminiscent of the involvement of the acyl chains of the M. tuberculosis 19-kDa lipoprotein antigen in binding to tolllike receptor-2 (TLR2) on phagocytic cells (31). Selective recognition of the ManLAM mannose-capping residues by DC-SIGN on the surface of DCs is likely to have important consequences for both the pathogenesis and immunology of tuberculosis and other mycobacterial diseases. LAMs have various effects on phagocytic cells, including Ms and DCs (24). PILAMs induce the secretion of proinflammatory cytokines, such as tumor necrosis factor-␣, interleukin-1 (IL-1), IL-12, and IL-6, and the production of microbicidal radicals, such as NO 2 Ϫ , in a much more potent way than do ManLAMs. In addition, ManLAMs, but not PILAMs, inhibit the M activation effect of interferon-␥ produced by effector T lymphocytes. PILAMs are thus now considered as proinflammatory molecules, whereas ManLAMs are viewed rather as anti-inflammatory components (17,24), which is consistent with the known ability of ManLAM-containing slow growing mycobacteria to resist immune defense mechanisms of their susceptible host (32). In particular, our recent results demonstrate that ManLAM inhibits the secretion of IL-12 by human DCs, a process that, like DC-SIGN ligation, requires both the MPI anchor acyl chains and the mannose caps (27). In this previous study (27), based on experiments using monoclonal antibodies, we suggested that ManLAM was acting through ligation of the mannose receptor (MR). Although MR is also involved in LAM mannose caps recognition (33), one cannot rule out the possibility that ManLAM could act also through the ligation of DC-SIGN, which is currently under investigation. Indeed, DC-SIGN ligation by ManLAM, either attached to the bacilli or released in the milieu through exocytosis (34), is likely to induce major signaling events, possibly including cell deactivation and/or secretion of anti-inflammatory cytokines such as transforming growth factor-␤ or IL-10 (35). Interestingly, PIL-AMs but not ManLAMs can activate cells in a TLR2-dependent manner (36). It will be of interest to study the cross-talk between phagocytic cell surface lectins, such as MR and DC-SIGN, and TLRs in response to mycobacterial ligands, including LAMs from various mycobacterial species.
From an evolutionary perspective, it is interesting that DC-SIGN can discriminate between fast growing saprophytic and slow growing virulent or potentially virulent mycobacteria. Until recently, it was proposed that mannose capping of the LAM molecules was a unique feature of virulent mycobacteria. This is unlikely to be the case because the attenuated species M. bovis BCG also contains mannose-capped LAM (21). Even if not virulent stricto sensu, M. bovis BCG can be pathogenic under certain conditions, especially in children and immunocompromised patients, in whom it may cause a variety of effects ranging from local adenitis to disseminated disease (37). Moreover, M. bovis BCG is derived from virulent M. bovis, which shares a common ancestor with M. tuberculosis (38). One cannot exclude that mannose capping of the ManLAM molecule is a feature of virulent mycobacteria that has been conserved during the recent evolution of M. bovis BCG from M. bovis. In this context, DC-SIGN could be viewed as a pattern recognition receptor (39) that has evolved to recognize potentially harmful mycobacteria through their specific surface glycosylated moieties. In parallel, mycobacteria could have evolved mechanisms (LAM capping) to resist host immunity (deactivation of the inflammatory response) in contrast with their fast growing, soil-living, harmless ancestors.