TLR2 is a pattern-recognition receptor that is activated by a large variety of conserved microbial components, including lipoproteins, lipoteichoic acids, and peptidoglycan. Lipoglycans are TLR2 agonists found in some genera of the phylogenetic order Actinomycetales, including Mycobacterium. They are built from a mannosyl-phosphatidyl-myo-inositol anchor attached to a (α1→6)-linked d-mannopyranosyl chain whose units can be substituted by d-mannopyranosyl and/or d-arabinofuranosyl units. At this time, little is known about the molecular bases underlying their ability to induce signaling via this receptor. We have recently shown that the anchor must be at least triacylated, including a diacylglyceryl moiety, whereas the contribution of the glycosidic moiety is not yet clearly defined. We show herein that lipoglycan activity is directly determined by mannan chain length. Indeed, activity increases with the number of units constituting the (α1→6)-mannopyranosyl backbone but is also critically dependent on the substitution type of the 2-hydroxyl of these units. We thus provide evidence for the definition of a new pattern that includes the nonlipidic moiety of the molecules, most probably as a result of the (α1→6)-mannopyranosyl backbone being a highly conserved structural feature among lipoglycans. Moreover, we demonstrate that lipoglycans can bind cell surface-expressed TLR2 and that their ability to induce signaling might be, at least in part, dictated by their avidity for the receptor. Finally, our data suggest that lipoglycans and lipoproteins have a common binding site. The present results are thus discussed in the light of the recently published crystal structure of a TLR1-TLR2-lipopeptide complex.

The innate immune system is the first line of host defense against invading pathogens. It is mediated by phagocytes including macrophages and dendritic cells that recognize microorganisms via a limited number of germline-encoded pattern recognition receptors, among which are the TLRs. TLRs are a family of at least 12 membrane proteins that trigger innate immune responses through NF-κB-dependent and IFN regulatory factor-dependent signaling pathways (for recent reviews, see Refs. 1, 2, 3). They are type I transmembrane proteins that possess an N-terminal ectodomain of leucine-reach repeats, which are involved directly or through accessory molecules in ligand binding, a single transmembrane domain, and a C-terminal cytoplasmic Toll/IL-1 receptor domain that interacts with Toll/IL-1 receptor domain-containing adaptator molecules to induce intracellular signaling. TLRs are widely expressed in many cell types, and the immune sentinel cells, such as macrophages, neutrophils, and dendritic cells, express most of them. Some TLRs are expressed at the cell surface, whereas others are found almost exclusively in intracellular compartments such as endosomes (1, 2, 3).

TLRs recognize conserved microbial-associated molecular patterns that are essential for the survival of the microorganism and are therefore difficult for it to alter. However, they are unusual in that some can recognize several structurally unrelated ligands. TLR2, which plays a major role in detecting Gram-positive bacteria, is most probably the TLR that achieves the highest diversity in ligand recognition. Indeed, its ligands are as diverse as lipopeptides, lipoteichoic acid, peptidoglycan, porins, zymosan, or glycosyl-phosphatidyl-myo-inositol (GPI)3 anchors. However, TLR2 generally functions as a heterodimer with either TLR1 or TLR6, which appears to be involved in discrimination of the acylation state of lipoproteins (4, 5, 6) and GPI anchors (7). Indeed, triacylated lipoproteins and GPIs are preferentially recognized by the TLR2/TLR1 complex, whereas diacylated lipoproteins and GPIs are recognized by the TLR2/TLR6 complex. The later heterodimer is also required for activation by zymosan and peptidoglycan (4).

Recently, lipoglycans have been defined as a new class of TLR2 agonists (8, 9, 10, 11, 12, 13, 14, 15, 16). These ligands are recognized in the context of TLR2 heterodimerization with TLR1 (17, 18, 19, 20). Lipoglycans are found in some genera of the order Actinomycetales (21) but have been extensively studied in the genus Mycobacterium (22, 23). Interestingly, two polymorphisms in the exon part of TLR2 that attenuate receptor signaling also enhance the risk of developing tuberculosis and leprosy (24). Mycobacteria contain a family of lipoglycans whose archetypes are phosphatidyl-myo-inositol mannosides (PIM), lipomannan (LM), and lipoarabinomannan (LAM). They all share a conserved mannosyl-phosphatidyl-myo-inositol anchor (MPI), based on a sn-glycero-3-phospho-(1-d-myo-inositol) unit with one α-d-mannopyrannose (Manp) unit linked at O-2 of the myo-inositol. This MPI anchor contains four potential sites of acylation: positions 1 and 2 of the glycerol unit, position 6 of the Manp unit linked at O-2 of the myo-inositol, and position 3 of the myo-inositol. O-6 of myo-inositol can be glycosylated by one or five Manp units, yielding PIM2 and PIM6. LMs correspond to polymannosylated PIMs and are built from a conserved (α1→6)-Manp backbone in which some units are substituted, generally at O-2, by single α-Manp units. LAMs correspond to LMs with an attached d-arabinan domain (Fig. 1). In some mycobacterial species, the nonreducing termini of the arabinosyl side-chains can be modified by a cap motive consisting of either oligomannosyl or phospho-myo-inositol units. LM (13, 14) and phospho-myo-inositol-capped LAM (PILAM) (8, 10) have been described as strong TLR2 agonists, whereas PIMs (11, 12, 20) were found to be weak agonists. Lipoglycans identified in other genera such as Rhodococcus, Corynebacterium, Tsukamurella, Turicella, and Saccharothrix also consist of a MPI anchor glycosylated by a (α1→6)-Manp backbone. However, in most cases, they are simpler in structure. Most particularly, the arabinan moiety of the molecule can be reduced to single arabinosyl susbtituents (15, 16, 25, 26, 27) (Fig. 1). These different lipoglycan variants, either LAM or LM, show variable TLR2-dependent proinflammatory activities (15, 16).

FIGURE 1.

Structural models of lipoglycans tested in the present study. M, d-mannopyranoside units; A, d-arabinofuranoside units; Man, mannose; PI, phosphoinositol; MPI, mannosyl-phosphatidyl-myo-inositol. BCGLM, RvLM, SaeLM, TpaLAM and TpaLM, RruLAM, ReqLAM, and TotLAM are LM or LAM from M. bovis BCG, Mycobacterium tuberculosis H37Rv, Saccharothrix aerocolonigenes, Tsukamurella paurometabolum, Rhodococcus ruber, Rhodococcus equi, and Turicella otitidis, respectively.

FIGURE 1.

Structural models of lipoglycans tested in the present study. M, d-mannopyranoside units; A, d-arabinofuranoside units; Man, mannose; PI, phosphoinositol; MPI, mannosyl-phosphatidyl-myo-inositol. BCGLM, RvLM, SaeLM, TpaLAM and TpaLM, RruLAM, ReqLAM, and TotLAM are LM or LAM from M. bovis BCG, Mycobacterium tuberculosis H37Rv, Saccharothrix aerocolonigenes, Tsukamurella paurometabolum, Rhodococcus ruber, Rhodococcus equi, and Turicella otitidis, respectively.

Close modal

Thus far the molecular bases underlying the ability of lipoglycans to induce signaling via TLR2 are poorly understood. We have recently shown, using Mycobacterium bovis bacillus Calmette-Guérin (BCG) LM as a model, that the MPI anchor must be at least triacylated (20). In the present study, we have used highly purified lipoglycans with different carbohydrate domains to get further insights into the contribution of the glycosidic moiety.

Nomenclature and structural features of the lipoglycans studied are shown in Fig. 1 and Table I. Lipoglycans, mycobacterial LAMs (PILAM, ManLAM, AraLAM) (36), LMs (RvLM, BCGLM) (28), PIMs (PIM2, PIM6) (12), SaeLM (15), TpaLAM and TpaLM (16), RruLAM (25), ReqLAM (26), and TotLAM (27) were purified as previously described. Their purity was determined by a set of analyses, including SDS-PAGE, chemical degradations, mass spectrometry, and NMR. The 19-kDa lipoprotein was provided by John T. Belisle (Colorado State University, Fort Collins, CO). Pam3CSK4, and Pam2CSK4 lipopeptides were from Invivogen. Pam3CSK4 used for binding experiments was from Boehringer Mannheim. The fluoroprobe Alexa Fluor 488 was coupled to Pam3CSK4 (A-Pam3CSK4) using a labeling kit from Molecular Probes as described previously (37). Recombinant human soluble CD14 (sCD14) was purified from conditioned medium of Schneider-2 insect cells transfected with cDNA encoding human CD14 as previously described (38).

Table I.

Structural and functional features of lipoglycans

LipoglycanaStructural FeaturesActivities
Size (kDa)Mannan Chain Size/Lengthb% Substitution by ArafcStudiesdTLR2 Activation EC50 (ng/ml)e% Inhibition Pam3CSK4/TLR2f
BCGLM 5.5 25/14 (202879 
RvLM 5.5 25/14 UD ND 
SaeLM 36/14 (15ND 
TpaLM 11/11 (16ND 
RruLM 28/27 42 (2515 ND 
ReqLM 36/23 29 (2625 ND 
TpaLAM 13 11/11 Arabinan chain (1650 ND 
PIM6 2.6 5/5 (12100 74 
fPILAM ND ND Bulky arabinan UD 1,000 ND 
sPILAM 17 25/16 Bulky arabinan (2930312,000 ND 
PIM2 1.4 1/1 (122,000 65 
RvManLAM 17 30/18 Bulky arabinan (32332,000 ND 
BCGManLAM 17 30/18 Bulky arabinan (32342,000 60 
AraLAM 17 30/17 Bulky arabinan (354,000 ND 
TotLAM 28/28 100 (27>10,000 ND 
LipoglycanaStructural FeaturesActivities
Size (kDa)Mannan Chain Size/Lengthb% Substitution by ArafcStudiesdTLR2 Activation EC50 (ng/ml)e% Inhibition Pam3CSK4/TLR2f
BCGLM 5.5 25/14 (202879 
RvLM 5.5 25/14 UD ND 
SaeLM 36/14 (15ND 
TpaLM 11/11 (16ND 
RruLM 28/27 42 (2515 ND 
ReqLM 36/23 29 (2625 ND 
TpaLAM 13 11/11 Arabinan chain (1650 ND 
PIM6 2.6 5/5 (12100 74 
fPILAM ND ND Bulky arabinan UD 1,000 ND 
sPILAM 17 25/16 Bulky arabinan (2930312,000 ND 
PIM2 1.4 1/1 (122,000 65 
RvManLAM 17 30/18 Bulky arabinan (32332,000 ND 
BCGManLAM 17 30/18 Bulky arabinan (32342,000 60 
AraLAM 17 30/17 Bulky arabinan (354,000 ND 
TotLAM 28/28 100 (27>10,000 ND 
a

BCGLM, BCGManLAM, PIM6, and PIM2 were purified from M. bovis BCG; RvLM and RvManLAM were from M. tuberculosis H37Rv; AraLAM was from M. chelonae; fPILAM and sPILAM were from M. fortuitum and M. smegmatis, respectively; SaeLM, TpaLAM and TpaLM, RruLAM, ReqLAM, and TotLAM were from Saccharothrix aerocolonigenes, Tsukamurella paurometabolum, Rhodococcus ruber, Rhodococcus equi, and Turicella otitidis, respectively.

b

Mannan chain size corresponds to the total number of Manp units of the mannan domain, whereas mannan chain length refers to the number of units of the (α1→6)-Manp chain. Except for PIM2 and PIM6, these are average values from a Gaussian distribution.

c

Values correspond to the proportion of the (α1→6)-Manp units that are directly substituted by Araf units (see Fig. 1).

d

These references provide structural features of the various lipoglycans; UD indicates unpublished data from our group.

e

Determined by stimulation of HEK-TLR2 cells (see Fig. 3).

f

Inhibition of A-Pam3CSK4 binding to surface-expressed TLR2; values were obtained using a 20-fold molar excess of lipoglycan/sCD14 complexes compared to A-Pam3CSK4/sCD14 complexes (see Fig. 5).

Complexes between sCD14 and Pam3CSK4, A-Pam3CSK4, PIM6, PIM2, BCGLM, or BCGLAM were formed by incubating Pam3CSK4 (8 μg/ml), A-Pam3CSK4 (8 μg/ml), PIM2 (5.6 μg/ml), PIM6 (10 μg/ml), SaeLM (28 μg/ml), BCGLM (22 μg/ml), and BCGManLAM (68 μg/ml) with sCD14 (100 μg/ml) overnight at 37°C in Dulbecco’s PBS containing 0.05% pyrogen-free human serum albumin (HSA).

The THP-1 monocyte/macrophage human cell line was maintained in continuous culture with RPMI 1640 medium (Invitrogen), 10% FCS (Invitrogen) in an atmosphere of 5% CO2 at 37°C, as nonadherent cells. Lipoglycans were added in duplicate or triplicate, at concentrations of 10 or 20 μg/ml, to the THP-1 cells (5 × 105 cells/well) in 24-well culture plates and then incubated for 20 h at 37°C. Supernatants from THP-1 cells were assayed for TNF-α by sandwich ELISA using commercially available kits and according to manufacturer’s instructions (R&D Systems).

The HEK-Blue-2 cell line (Invivogen), a derivative of HEK293 cells that stably expresses the human TLR2 and CD14 genes along with a NF-κB-inducible reporter system (secreted alkaline phosphatase), was used according to the manufacturer’s instructions. Cells were plated at 5 × 104 cells per well in 96-wells plates and the different lipoglycans were added at concentrations ranging from 0.1 to 10,000 ng/ml in the HEK-Blue Detection medium (Invivogen) that contains a substrate for alkaline phosphatase. Alkaline phosphatase activity was measured after 18 to 40 h by reading OD at 630 nm. To investigate the CD14 and TLR dependence of lipoglycan activity, HEK-TLR2 cells were preincubated for 30 min at 37°C, before stimuli addition, with various Abs: 10 μg/ml monoclonal anti-CD14 (clone 134620, R&D Systems), 2 μg/ml monoclonal anti-TLR1 (Invivogen), 2 μg/ml monoclonal anti-TLR6 (Invivogen), or an IgG1 isotype control (eBioscience).

The HEK293 cells stably expressing a Flag-TLR2 protein (HEK-Flag-TLR2) (39) were cultured in complete culture medium on glass coverslips precoated with 0.5% gelatin for 24–48 h before experiments. The cells were washed twice with PBS containing 0.05% HSA and incubated for 15 min at 37°C in PBS-HSA with A-Pam3CSK4/sCD14 complexes (160 ng/ml Pam3CSK4, 2 μg/ml sCD14) in the presence or not of a 10- or 20-fold molar excess of unlabelled lipoglycan/sCD14 or Pam3CSK4/sCD14 complexes. At the end of incubation, coverslips were washed several times with PBS-HSA, fixed for 30 min in PBS containing 4% paraformaldehyde, stained for TLR2 by successive incubation for 30 min at 4°C with a mouse anti-Flag M2 mAb and a Cy3-labeled anti-mouse Ab from Sigma-Aldrich, and mounted. Images were obtained using a fluorescence microscope equipped with a CoolSNAP HQ charge-coupled device (CCD) camera (Ropper Scientific), and cell-surface fluorescence intensity ratios of A-Pam3CSK4 to Flag-TLR2 were measured from CCD images either manually using the Metamorph software (see Fig. 6,B) or automatically using a visual scripting interface for the ImageJ software (see Fig. 6 C) (that can be downloaded from http://www.mri.cnrs.fr) (40). Data are expressed as arbitrary units.

FIGURE 6.

Lipoglycans compete with A-Pam3CSK4 for binding to TLR2. HEK-Flag-TLR2 cells were incubated for 15 min at 37°C with A-Pam3CSK4/sCD14 complexes and a 10-fold (B) or 20-fold (A and C) molar excess of the indicated lipoglycan/sCD14 or Pam3CSK4/sCD14 complexes. A, CCD images of Cy3-Flag-TLR2 and A-Pam3CSK4 fluorescence in representative cells (center and right panels, respectively) and the merged images (left panels) are shown. B and C, Ratio of cell-surface fluorescence intensities of A-Pam3CSK4 to Flag-TLR2 measured from CCD images. Results are expressed as arbitrary units and are the mean of at least 20 independent measures (B) or of 1264–3584 measures (C). Statistical analysis of the data for C gives p values <10−24 as assessed by a Student’s t test comparison of the different conditions by pairs. Inhibition percentages are indicated on the graphs.

FIGURE 6.

Lipoglycans compete with A-Pam3CSK4 for binding to TLR2. HEK-Flag-TLR2 cells were incubated for 15 min at 37°C with A-Pam3CSK4/sCD14 complexes and a 10-fold (B) or 20-fold (A and C) molar excess of the indicated lipoglycan/sCD14 or Pam3CSK4/sCD14 complexes. A, CCD images of Cy3-Flag-TLR2 and A-Pam3CSK4 fluorescence in representative cells (center and right panels, respectively) and the merged images (left panels) are shown. B and C, Ratio of cell-surface fluorescence intensities of A-Pam3CSK4 to Flag-TLR2 measured from CCD images. Results are expressed as arbitrary units and are the mean of at least 20 independent measures (B) or of 1264–3584 measures (C). Statistical analysis of the data for C gives p values <10−24 as assessed by a Student’s t test comparison of the different conditions by pairs. Inhibition percentages are indicated on the graphs.

Close modal

We first compared the relative ability of different lipoglycans (Fig. 1) to stimulate the release of TNF-α by THP-1 cells. These cells express at their surface high levels of TLR2 but barely detectable levels of TLR4 (not shown) (41). We previously established, using blocking Abs, that the cytokine release induced by the various lipoglycans studied here was dependent on TLR2 but not TLR4 (15, 16). Each lipoglycan was tested at concentrations of 10 and 20 μg/ml (Fig. 2). LMs were found to be the most active lipoglycans. However, activity seemed to depend on the size of the mannan domain and/or the presence and length of lateral chains and additional domains (Fig. 2). Indeed, SaeLM, BCGLM, and TpaLM, the most active molecules, have mannan domains built from 36, 25, and 11 Manp units on average, respectively, and differ by the side chains substituting the (α1→6)-Manp backbone, which are dimannosyl units, monomannosyl units, or no units, respectively (15, 16, 20) (Fig. 1, Table I). In contrast, PIM6 and PIM2, which have a smaller mannan domain, with 5 and 1 Manp units, respectively, were virtually inactive in this model (data not shown). As previously established by us and others, LAM with a bulky arabinan domain (such as ManLAMs or TpaLAM) are inactive or weakly active (Fig. 2) (13, 16). This is likely due to a masking of the LM domain of the molecule because removal of the arabinan domain by chemical treatments restores the proinflammatory activity of the resulting LM moiety (13, 16), as shown here by the activity of TpaLM as compared with that of TpaLAM (Fig. 2). More interestingly, a direct substitution of the (α1→6)-Manp backbone by arabinosyl units also inhibits the TLR2-inducing activity. Indeed, partial substitution (as the in case of RruLAM and ReqLAM) resulted in lipoglycans with a lowered activity compared with that of LMs (Fig. 2) despite the presence in the mannan moiety of these molecules of 28 and 36 Manp units, respectively (25, 26) (Fig. 1, Table I). Moreover, complete substitution (as in the case of TotLAM) resulted in a fully abrogated activity. Because in these three lipoglycans the (α1→6)-Manp backbone is substituted at the 2-hydroxyl, these data indicate that this position of the Manp units plays a critical role in the ability of lipoglycans to induce signaling via TLR2. However, the presence of Manp, but not arabinofuranose (Araf), units glycosylating these positions can compensate for the loss of the free hydroxyl groups.

FIGURE 2.

TNF-α production by human THP-1 monocyte/macrophage cell line in response to various lipoglycans. Lipoglycans were tested at 10 (filled bars) and 20 (open bars) μg/ml. ManLAM was from M. bovis BCG (BCGManLAM).

FIGURE 2.

TNF-α production by human THP-1 monocyte/macrophage cell line in response to various lipoglycans. Lipoglycans were tested at 10 (filled bars) and 20 (open bars) μg/ml. ManLAM was from M. bovis BCG (BCGManLAM).

Close modal

Comparing the relative ability of lipoglycans to stimulate the release of TNF-α by THP-1 cells thus allows a better understanding of the structure/function relationship of these molecules. Lipoglycan activity appears to depend on the size of the mannan chain and is critically dependent on the substitution type of the 2-hydroxyl of the (α1→6)-Manp backbone that must be free or glycosylated by Manp, but not Araf, units. However further investigations, such as into the role of the mannan chain size, are limited by the fact that cytokine production by THP-1 cells shows a weak dynamic of response. Indeed, the dose-dependent activation of the cells is limited to a sharp range of lipoglycan concentration, and thus the results only give a “yes” or “no” response concerning lipoglycan activity.

To tentatively overcome this limitation, we used HEK293 cells stably transfected with human TLR2 and CD14 genes (HEK-TLR2 cells) and a NF-κB-inducible reporter system (secreted alkaline phosphatase). The 19-kDa lipoprotein (LP19), a well-known mycobacterial TLR2 agonist (42), and also BCGLM and BCGManLAM induced NF-κB activation in a dose-dependent manner in HEK-TLR2 cells (Fig. 3) but not in the parent HEK cells (data not shown), demonstrating that activation was specific for TLR2. LP19 and BCGLM were strong agonists of the receptor, and an EC50 of 1 ng/ml was determined for both molecules (Fig. 3). In sharp contrast, but in agreement with data obtained with THP-1 cells, BCGManLAM was a weak agonist with an EC50 of 2 μg/ml, thus exhibiting an activity reduced by more than three orders of magnitude as compared with BCGLM. In a similar way, we determined the EC50 value for each lipoglycan (Table I). Data were in agreement with those obtained with THP-1 cells (Fig. 2). The most active molecules were the various LMs as well as RruLAM and ReqLAM, that is, lipoglycans with the highest number of nonsubstituted mannosyl units. We noticed that SaeLM and BCGLM showed the same EC50 of 1 ng/ml. The mannan domain of SaeLM is larger when compared with that of BCGLM. However, this mainly results from longer side chains for the former (i.e., dimannosyl vs single mannosyl units) (15) (Fig. 1, Table I). Nevertheless, the length of the mannan chain (i.e., the number of mannosyl units building the (α1→6)-Manp backbone) is the same in both molecules: 14 mannosyl units on average (Fig. 1, Table I). We thus hypothesized that mannan chain length, rather than mannan chain size, might be a key parameter determining lipoglycan activity. We therefore plotted the activity of the lipoglycans (log of EC50 value) as a function of their mannan chain length (number of mannosyl units constituting the chain backbone) (Fig. 4). Intriguingly, a linear relationship was observed for lipoglycans containing solely mannosyl units, that is, LMs (BCGLM, RvLM, SaeLM, and TpaLM) and PIMs (PIM6 and PIM2), demonstrating that LM activity is directly determined by the mannan chain length. However, lipoglycans substituted with arabinosyl units, that is, LAMs, showed an activity weaker than that expected from their mannan chain length. For example, although ManLAMs have a mannan chain identical with that of LMs from the same species, they only show an activity equivalent to that of PIM2 (Fig. 4, Table I), suggesting that their bulky arabinan domain masks the mannan chain in such a way that they behave like molecules with a mannan restricted to a single mannosyl unit. Similarly, TpaLAM activity is equivalent to that of a LM with a mannan length of 7 mannosyl units, whereas its mannan is made from 11 mannosyl units (Figs. 1 and 4, Table I). The arabinan domain of TpaLAM is much simpler and smaller than that of ManLAMs (16) (Fig. 1), and thus its steric hindrance on the mannan domain is lowered as compared with that observed with ManLAMs. Concerning LAMs whose mannan chain is directly substituted by arabinosyl units, that is, RruLAM and ReqLAM, their activity corresponds to that of a mannan chain with 9 and 8 units, respectively. This is less than would be expected if we take into account only the number of Manp units that are not substituted (i.e., 16 in both cases; see Fig. 4, lipoglycan names labeled with asterisks), suggesting that Araf substitution has side effects. Finally, the EC50 of TotLAM, which has no free Manp units, was found to be higher than 10 μg/ml, a value >5-fold higher than that obtained for PIM2. The activity of all the lipoglycans was dependent on the expression of CD14 (Fig. 5,A) and on the heterodimerization of TLR2 with TLR1 but not TLR6 (Fig. 5 B) as determined by blocking Ab experiments and in agreement with previous data obtained on mycobacterial LAM and LM (13, 17, 18, 19, 20).

FIGURE 3.

NF-κB activation in HEK-TLR2 cells by various stimuli. Cells (5 × 104) were plated in 96-wells plates and stimulated at 37°C for 24 h by LP19 (○), BCGLM (▴), or BCGManLAM (▪). NF-κB activity was determined by reading OD at 630 nm.

FIGURE 3.

NF-κB activation in HEK-TLR2 cells by various stimuli. Cells (5 × 104) were plated in 96-wells plates and stimulated at 37°C for 24 h by LP19 (○), BCGLM (▴), or BCGManLAM (▪). NF-κB activity was determined by reading OD at 630 nm.

Close modal
FIGURE 4.

TLR2-dependent activity of lipoglycans as a function of their mannan chain length. The mannan chain length corresponds to the number of mannosyl units building the (α1→6)-Manp backbone. For lipoglycans labeled with asterisks, only the numbers of Manp units of the (α1→6)-Manp backbone that are not substituted by Araf units have been taken into account. Lipoglycan activity corresponds to decadic logarithm of the EC50 value determined by stimulation of HEK-TLR2 cells (Fig. 3, Table I).

FIGURE 4.

TLR2-dependent activity of lipoglycans as a function of their mannan chain length. The mannan chain length corresponds to the number of mannosyl units building the (α1→6)-Manp backbone. For lipoglycans labeled with asterisks, only the numbers of Manp units of the (α1→6)-Manp backbone that are not substituted by Araf units have been taken into account. Lipoglycan activity corresponds to decadic logarithm of the EC50 value determined by stimulation of HEK-TLR2 cells (Fig. 3, Table I).

Close modal
FIGURE 5.

Lipoglycan activity is dependent on CD14 expression (A) and TLR2 heterodimerization with TLR1 (B). HEK-TLR2 cells were preincubated for 30 min at 37°C, before lipoglycan addition, with various Abs: 10 μg/ml monoclonal anti-CD14, 2 μg/ml monoclonal anti-TLR1, 2 μg/ml monoclonal anti-TLR6, or an IgG1 isotype control (at the corresponding concentration). BCGManLAM, PIM2, and PIM6 were tested at a concentration of 500 ng/ml and TpaLM, RvLM, BCGLM, and SaeLM at a concentration of 10 ng/ml. Pam3CSK4 (5 ng/ml) and Pam2CSK4 (0.05 ng/ml) lipopeptides were used as positive controls of TLR2/TLR1 and TLR2/TLR6 agonists, respectively.

FIGURE 5.

Lipoglycan activity is dependent on CD14 expression (A) and TLR2 heterodimerization with TLR1 (B). HEK-TLR2 cells were preincubated for 30 min at 37°C, before lipoglycan addition, with various Abs: 10 μg/ml monoclonal anti-CD14, 2 μg/ml monoclonal anti-TLR1, 2 μg/ml monoclonal anti-TLR6, or an IgG1 isotype control (at the corresponding concentration). BCGManLAM, PIM2, and PIM6 were tested at a concentration of 500 ng/ml and TpaLM, RvLM, BCGLM, and SaeLM at a concentration of 10 ng/ml. Pam3CSK4 (5 ng/ml) and Pam2CSK4 (0.05 ng/ml) lipopeptides were used as positive controls of TLR2/TLR1 and TLR2/TLR6 agonists, respectively.

Close modal

We next asked whether the activity of lipoglycans was determined by their ability to bind TLR2. Vasselon et al. (37) have recently developed a binding assay of fluorescently labeled synthetic triacylated lipopeptide Pam3CSK4 (A-Pam3CSK4) to TLR2 expressed on the surface of TLR2-transfected HEK cells (HEK-Flag-TLR2). They have shown that TLR2 strongly and specifically binds A-Pam3CSK4 when presented as a complex with sCD14. We used this assay to determine whether lipoglycans were able to compete for Pam3CSK4 binding to TLR2. Lipoglycan/sCD14 complexes were prepared and we found that their addition in a 10- or 20-fold molar excess to A-Pam3CSK4/sCD14 complexes resulted in an inhibition of Pam3CSK4 binding to TLR2 (Fig. 6, Table I). Inhibition was shown to be dose dependent as evidenced with PIM2 and PIM6 complexes. These results demonstrate that lipoglycans directly bind TLR2 and that their binding site is at least partially common with that of bacterial lipopeptides. Interestingly, BCGLM was a better competitor than PIM6, which was better than PIM2 and ManLAM (Fig. 6,C; p < 10−24). The best agonists (Table I) were thus the best competitors (Fig. 6), suggesting that lipoglycan ability to induce signaling via TLR2 is at least partially dictated by their avidity for this receptor.

MPI-anchored lipoglycans, which are found in some genera of the Actinomycetales order, are agonists of TLR2. These are complex molecules bearing one to four fatty acids on their MPI anchor and showing a great degree of heterogeneity in their carbohydrate moiety (23). Although recent studies have shown that the lipidic part of the molecule is necessary for its activity, the contribution of the glycosidic moiety is still poorly understood. In this study, we have taken advantage of a wide collection of purified and structurally characterized lipoglycans with high diversity in the carbohydrate moiety (see Fig. 1) to investigate their structure/function relationship.

We report herein that the activity of lipoglycans is directly determined by their mannan chain length. Indeed, activity increases with the number of units constituting the (α1→6)-mannopyranosyl backbone and can reach that observed for triacylated lipoproteins (EC50 of 1 ng/ml in our cellular model). Moreover, for lipoglycans with a glycosidic moiety composed of Manp units only, a linear relationship can be observed between log(EC50) and the number of these units. However activity is critically dependent on the accessibility of this mannan chain as well as the substitution type of the 2-hydroxyl of these units, which must be free or glycosylated by Manp, but not Araf, units. Indeed lipoglycans containing a bulky arabinan domain such as mycobacterial LAM are very weakly active despite the presence of a LM moiety that has an intrinsic capacity to induce TLR2 signaling and that can be unmasked, as previously reported by removing arabinan domain (13, 16). Lipoglycans with a mannan chain whose units are directly substituted with Araf units have also an activity lower than that expected from the size of their mannan chain. PILAM (formerly known as AraLAM) has been reported in the literature to be proinflammatory (31, 43) via TLR2 signaling (10, 17). Herein we have obtained convincing evidence, using PILAM purified from Mycobacterium smegmatis and Mycobacterium fortuitum, that it is not more active than ManLAM or the noncapped AraLAM (Fig. 1). Thus, in contrast to what was previously suggested, the presence of phospho-myo-inositol caps is not likely to be a structural determinant conferring the capacity of LAM to become proinflammatory. The proinflammatory activity observed in some PILAM preparations is most probably due to contaminant lipopeptides (unpublished data). Therefore, concerning the glycosidic moiety, the activity mainly depends on the length of the mannan chain and on the nature of its substitution.

Lipoglycans exist in different acyl- forms that differ by the number of the fatty acids esterifying the MPI anchor (one to four) (12, 20, 23). Using BCGLM as a model, we have recently shown that acylation degree is critical for activity. Indeed, tri- and tetra-acylated molecules, bearing two fatty acids on the glycerol, were found to be active and required the heterodimerization of TLR2 with TLR1 while mono- and diacylated LMs did not activate TLR2 even in the presence of TLR6 (20). These results are in contrast to those observed with lipopeptides (4, 5, 6) or GPI anchors from Plasmodium falciparum (7) even though the latter have a structure close to that of lipoglycan MPI anchors. Indeed, in both cases the activity is independent of the number of fatty acids (two or three). However this activity differs considerably in the requirement of the auxiliary receptor (TLR6 vs TLR1, respectively). The different lipoglycans tested in the present study are predominantly triacylated, but they also consist in a mixture of different acyl- forms whose relative abundance can vary from one lipoglycan to another. Nevertheless, lipoglycan relative activity, which differs by several orders of magnitude according to the mannan chain length (Table I, Fig. 4), is not likely to be dramatically affected by <2-fold variation of the proportion of the active acyl-forms. Cumulatively, we provide herein evidence for the definition of a new pattern that includes both a lipidic and a glycosidic moiety. To our knowledge, this is the first report showing that the size of the nonlipidic moiety of a lipidic TLR2 ligand dramatically determines signaling (44).

Vasselon et al. (37) have shown that TLR2 recognizes Pam3CSK4 through direct binding and that the leucine-reach repeat region of TLR2 carries the specificity for binding the agonist and for initiating signaling. The crystal structure of the TLR1-TLR2 heterodimer in complex with Pam3CSK4 has been recently reported (45). The triacylated lipopeptide appears to form a bridge between TLR2 and TLR1 with the two ester-bound fatty acyl chains inserted deep into a pocket in the hydrophobic core of TLR2, the third amide-linked acyl chain occupying a hydrophobic channel at the surface of TLR1 and the conserved polar head located at the region of contact between the two receptors. In contrast, the four lysine residues have only limited interactions with TLR1 or TLR2, which is consistent with previous studies suggesting strongly that the highly variable polypeptide chain of lipoproteins is not included in the pattern that is recognized by TLR2 and TLR1 (44).

Herein we have found that lipoglycans PIM2, PIM6, BCGLM, and BCGManLAM can compete with fluorescently-labeled Pam3CSK4 for binding to cell surface-expressed TLR2 (Fig. 6). Our results thus show that the binding site for lipoglycans and triacylated lipopeptides is overlapping, presumably because the lipid anchors of both classes of molecules share common structural similarities, including a diacylglyceryl moiety shown to be critical for activity of both classes of ligands (20, 44). We also found that variation in the length and nature of the carbohydrate chain of lipoglycans does not affect the specificity of recognition by TLR2/TLR1 or TLR6 heterodimers but rather modulates the sensitivity of responses mediated by TLR2-TLR1 heterodimers (Fig. 5 B). The observation that the most potent agonists were the best competitors of Pam3CSK4 suggests that the most active molecules have a higher affinity for the receptor complex. Thus, in contrast to the peptidic chain of lipoproteins, the (α1→6)-mannopyranosyl backbone, which is a highly conserved structural feature of lipoglycans, is an integral part of the microbial-associated molecular pattern and probably establishes close contacts with the receptors. However, further studies will be required to test this hypothesis.

We gratefully acknowledge Dr. Daniel Drocourt, Elise Armau, and Sophie Gauthier (Cayla/Invivogen, Toulouse, France) for helpful discussions and technical assistance. We thank John T. Belisle (Colorado State University, Fort Collins, CO) for providing the 19-kDa lipoprotein (through National Institutes of Health, National Institute of Allergy and Infectious Diseases, contract NO1 AI-75320 titled, “Tuberculosis Research Materials and Vaccine Testing”).

The authors have no financial conflicts of interest.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1

This work was supported by grants from Centre National de la Recherche Scientifique. G.S.B. was supported by a Personal Research Chair from James Bardrick, a Royal Society Wolfson Research Merit Award, as a former Lister Institute-Jenner Research Fellow, the Medical Research Council (G9901077 and G0500590), and The Wellcome Trust (081569/2/06/2).

3

Abbreviations used in this paper: GPI, glycosyl-phosphatidyl-myo-inositol; A-Pam3CSK4, Alexa Fluor 488-labeled synthetic lipopeptide; AraLAM, uncapped LAM; BCG, bacillus Calmette-Guérin; HEK-Flag-TLR2, HEK293 cells stably expressing a Flag-TLR2 protein; HSA, human serum albumin; LAM, lipoarabinomannan; LM, lipomannan; ManLAM, mannose-capped LAM; Manp, mannopyranose; MPI, mannosyl-phosphatidyl-myo-inositol; PILAM, phospho-myo-inositol-capped LAM; PIM, phosphatidyl-myo-inositol mannosides; ReqLAM, LAM from Rhodococcus equi; RruLAM, LAM from Rhodococcus ruber; RvLM, LM from Mycobacterium tuberculosis H37Rv; SaeLM, LM from Saccharothrix aerocolonigenes; TotLAM, LAM from Turicella otitidis; TpaLAM and TpaLM, LAM and LM, respectively, from Tsukamurella paurometabolum; Araf, arabinofuranose; CCD, charge-coupled device.

1
Akira, S., S. Uematsu, O. Takeuchi.
2006
. Pathogen recognition and innate immunity.
Cell
124
:
783
-801.
2
Trinchieri, G., A. Sher.
2007
. Cooperation of Toll-like receptor signals in innate immune defence.
Nat. Rev. Immunol.
7
:
179
-190.
3
West, A. P., A. A. Koblansky, S. Ghosh.
2006
. Recognition and signaling by Toll-like receptors.
Annu. Rev. Cell. Dev. Biol.
22
:
409
-437.
4
Ozinsky, A., D. M. Underhill, J. D. Fontenot, A. M. Hajjar, K. D. Smith, C. B. Wilson, L. Schroeder, A. Aderem.
2000
. The repertoire for pattern recognition of pathogens by the innate immune system is defined by cooperation between Toll-like receptors.
Proc. Natl. Acad. Sci. USA
97
:
13766
-13771.
5
Takeuchi, O., T. Kawai, P. F. Muhlradt, M. Morr, J. D. Radolf, A. Zychlinsky, K. Takeda, S. Akira.
2001
. Discrimination of bacterial lipoproteins by Toll-like receptor 6.
Int. Immunol.
13
:
933
-940.
6
Takeuchi, O., S. Sato, T. Horiuchi, K. Hoshino, K. Takeda, Z. Dong, R. L. Modlin, S. Akira.
2002
. Cutting edge: Role of Toll-like receptor 1 in mediating immune response to microbial lipoproteins.
J. Immunol.
169
:
10
-14.
7
Krishnegowda, G., A. M. Hajjar, J. Zhu, E. J. Douglass, S. Uematsu, S. Akira, A. S. Woods, D. C. Gowda.
2005
. Induction of proinflammatory responses in macrophages by the glycosylphosphatidylinositols of Plasmodium falciparum: cell signaling receptors, glycosylphosphatidylinositol (GPI) structural requirement, and regulation of GPI activity.
J. Biol. Chem.
280
:
8606
-8616.
8
Means, T. K., S. Wang, E. Lien, A. Yoshimura, D. T. Golenbock, M. J. Fenton.
1999
. Human Toll-like receptors mediate cellular activation by Mycobacterium tuberculosis.
J. Immunol.
163
:
3920
-3927.
9
Means, T. K., E. Lien, A. Yoshimura, S. Wang, D. T. Golenbock, M. J. Fenton.
1999
. The CD14 ligands lipoarabinomannan and lipopolysaccharide differ in their requirement for Toll-like receptors.
J. Immunol.
163
:
6748
-6755.
10
Underhill, D. M., A. Ozinsky, K. D. Smith, A. Aderem.
1999
. Toll-like receptor-2 mediates mycobacteria-induced proinflammatory signaling in macrophages.
Proc. Natl. Acad. Sci. USA
96
:
14459
-14463.
11
Jones, B. W., T. K. Means, K. A. Heldwein, M. A. Keen, P. J. Hill, J. T. Belisle, M. J. Fenton.
2001
. Different Toll-like receptor agonists induce distinct macrophage responses.
J. Leukocyte Biol.
69
:
1036
-1044.
12
Gilleron, M., V. F. Quesniaux, G. Puzo.
2003
. Acylation state of the phosphatidylinositol hexamannosides from Mycobacterium bovis bacillus Calmette Guérin and mycobacterium tuberculosis H37Rv and its implication in Toll-like receptor response.
J. Biol. Chem.
278
:
29880
-29889.
13
Vignal, C., Y. Guerardel, L. Kremer, M. Masson, D. Legrand, J. Mazurier, E. Elass.
2003
. Lipomannans, but not lipoarabinomannans, purified from Mycobacterium chelonae and Mycobacterium kansasii induce TNF-α and IL-8 secretion by a CD14-Toll-like receptor 2-dependent mechanism.
J. Immunol.
171
:
2014
-2023.
14
Quesniaux, V. J., D. M. Nicolle, D. Torres, L. Kremer, Y. Guerardel, J. Nigou, G. Puzo, F. Erard, B. Ryffel.
2004
. Toll-like receptor 2 (TLR2)-dependent-positive and TLR2-independent-negative regulation of proinflammatory cytokines by mycobacterial lipomannans.
J. Immunol.
172
:
4425
-4434.
15
Gibson, K. J., M. Gilleron, P. Constant, B. Sichi, G. Puzo, G. S. Besra, J. Nigou.
2005
. A lipomannan variant with strong TLR-2-dependent pro-inflammatory activity in Saccharothrix aerocolonigenes.
J. Biol. Chem.
280
:
28347
-28356.
16
Gibson, K. J., M. Gilleron, P. Constant, T. Brando, G. Puzo, G. S. Besra, J. Nigou.
2004
. Tsukamurella paurometabola lipoglycan, a new lipoarabinomannan variant with pro-inflammatory activity.
J. Biol. Chem.
279
:
22973
-22982.
17
Tapping, R. I., P. S. Tobias.
2003
. Mycobacterial lipoarabinomannan mediates physical interactions between TLR1 and TLR2 to induce signaling.
J. Endotoxin Res.
9
:
264
-268.
18
Sandor, F., E. Latz, F. Re, L. Mandell, G. Repik, D. T. Golenbock, T. Espevik, E. A. Kurt-Jones, R. W. Finberg.
2003
. Importance of extra- and intracellular domains of TLR1 and TLR2 in NFκB signaling.
J. Cell Biol.
162
:
1099
-1110.
19
Elass, E., L. Aubry, M. Masson, A. Denys, Y. Guerardel, E. Maes, D. Legrand, J. Mazurier, L. Kremer.
2005
. Mycobacterial lipomannan induces matrix metalloproteinase-9 expression in human macrophagic cells through a Toll-like receptor 1 (TLR1)/TLR2- and CD14-dependent mechanism.
Infect. Immun.
73
:
7064
-7068.
20
Gilleron, M., J. Nigou, D. Nicolle, V. Quesniaux, G. Puzo.
2006
. The acylation state of mycobacterial lipomannans modulates innate immunity response through Toll-like receptor 2.
Chem. Biol.
13
:
39
-47.
21
Sutcliffe, I..
2005
. Lipoarabinomannans: structurally diverse and functionally enigmatic macroamphiphiles of mycobacteria and related actinomycetes.
Tuberculosis
85
:
205
-206.
22
Chatterjee, D., K. H. Khoo.
1998
. Mycobacterial lipoarabinomannan: an extraordinary lipoheteroglycan with profound physiological effects.
Glycobiology
8
:
113
-120.
23
Nigou, J., M. Gilleron, G. Puzo.
2003
. Lipoarabinomannans: from structure to biosynthesis.
Biochimie
85
:
153
-166.
24
Kang, T. J., G. T. Chae.
2001
. Detection of Toll-like receptor 2 (TLR2) mutation in the lepromatous leprosy patients.
FEMS Immunol. Med. Microbiol.
31
:
53
-58.
25
Gibson, K. J., M. Gilleron, P. Constant, G. Puzo, J. Nigou, G. S. Besra.
2003
. Structural and functional features of Rhodococcus ruber lipoarabinomannan.
Microbiology
149
:
1437
-1445.
26
Garton, N. J., M. Gilleron, T. Brando, H. H. Dan, S. Giguere, G. Puzo, J. F. Prescott, I. C. Sutcliffe.
2002
. A novel lipoarabinomannan from the equine pathogen Rhodococcus equi: structure and effect on macrophage cytokine production.
J. Biol. Chem.
277
:
31722
-31733.
27
Gilleron, M., N. J. Garton, J. Nigou, T. Brando, G. Puzo, I. C. Sutcliffe.
2005
. Characterization of a truncated lipoarabinomannan from the actinomycete Turicella otitidis.
J. Bacteriol.
187
:
854
-861.
28
Gilleron, M., J. Nigou, B. Cahuzac, G. Puzo.
1999
. Structural study of the lipomannans from Mycobacterium bovis BCG: characterisation of multiacylated forms of the phosphatidyl-myo-inositol anchor.
J. Mol. Biol.
285
:
2147
-2160.
29
Khoo, K. H., A. Dell, H. R. Morris, P. J. Brennan, D. Chatterjee.
1995
. Inositol phosphate capping of the nonreducing termini of lipoarabinomannan from rapidly growing strains of Mycobacterium.
J. Biol. Chem.
270
:
12380
-12389.
30
Khoo, K. H., E. Douglas, P. Azadi, J. M. Inamine, G. S. Besra, K. Mikusova, P. J. Brennan, D. Chatterjee.
1996
. Truncated structural variants of lipoarabinomannan in ethambutol drug-resistant strains of Mycobacterium smegmatis: inhibition of arabinan biosynthesis by ethambutol.
J. Biol. Chem.
271
:
28682
-28690.
31
Gilleron, M., N. Himoudi, O. Adam, P. Constant, A. Venisse, M. Riviere, G. Puzo.
1997
. Mycobacterium smegmatis phosphoinositols-glyceroarabinomannans: structure and localization of alkali-labile and alkali-stable phosphoinositides.
J. Biol. Chem.
272
:
117
-124.
32
Nigou, J., A. Vercellone, G. Puzo.
2000
. New structural insights into the molecular deciphering of mycobacterial lipoglycan binding to C-type lectins: lipoarabinomannan glycoform characterization and quantification by capillary electrophoresis at the subnanomole level.
J. Mol. Biol.
299
:
1353
-1362.
33
Gilleron, M., L. Bala, T. Brando, A. Vercellone, G. Puzo.
2000
. Mycobacterium tuberculosis H37Rv parietal and cellular lipoarabinomannans: characterization of the acyl- and glyco- forms.
J. Biol. Chem.
275
:
677
-684.
34
Venisse, A., J. M. Berjeaud, P. Chaurand, M. Gilleron, G. Puzo.
1993
. Structural features of lipoarabinomannan from Mycobacterium bovis BCG: determination of molecular mass by laser desorption mass spectrometry.
J. Biol. Chem.
268
:
12401
-12411.
35
Guerardel, Y., E. Maes, E. Elass, Y. Leroy, P. Timmerman, G. S. Besra, C. Locht, G. Strecker, L. Kremer.
2002
. Structural study of lipomannan and lipoarabinomannan from Mycobacterium chelonae: presence of unusual components with α1,3-mannopyranose side chains.
J. Biol. Chem.
277
:
30635
-30648.
36
Pitarque, S., J. L. Herrmann, J. L. Duteyrat, M. Jackson, G. R. Stewart, F. Lecointe, B. Payre, O. Schwartz, D. B. Young, G. Marchal, et al
2005
. Deciphering the molecular bases of Mycobacterium tuberculosis binding to the lectin DC-SIGN reveals an underestimated complexity.
Biochem. J.
392
:
615
-624.
37
Vasselon, T., P. A. Detmers, D. Charron, A. Haziot.
2004
. TLR2 recognizes a bacterial lipopeptide through direct binding.
J. Immunol.
173
:
7401
-7405.
38
Detmers, P. A., D. Zhou, E. Polizzi, R. Thieringer, W. A. Hanlon, S. Vaidya, V. Bansal.
1998
. Role of stress-activated mitogen-activated protein kinase (p38) in β2-integrin-dependent neutrophil adhesion and the adhesion-dependent oxidative burst.
J. Immunol.
161
:
1921
-1929.
39
Vasselon, T., W. A. Hanlon, S. D. Wright, P. A. Detmers.
2002
. Toll-like receptor 2 (TLR2) mediates activation of stress-activated MAP kinase p38.
J. Leukocyte Biol.
71
:
503
-510.
40
Baecker, P., P. Travo.
2006
. Cell Image Analyzer: a visual scripting interface for ImageJ and its usage at the microscopy facility Montpellier RIO. A. Jahnen, and C. Moll, eds.
Proceedings of the ImageJ User and Developer Conference, May 18–19
105
-110. Centre de Recherche Public Henri Tudor, Luxembourg.
41
Tapping, R. I., S. Akashi, K. Miyake, P. J. Godowski, P. S. Tobias.
2000
. Toll-like receptor 4, but not Toll-like receptor 2, is a signaling receptor for Escherichia and Salmonella lipopolysaccharides.
J. Immunol.
165
:
5780
-5787.
42
Brightbill, H. D., D. H. Libraty, S. R. Krutzik, R. B. Yang, J. T. Belisle, J. R. Bleharski, M. Maitland, M. V. Norgard, S. E. Plevy, S. T. Smale, et al
1999
. Host defense mechanisms triggered by microbial lipoproteins through Toll-like receptors.
Science
285
:
732
-736.
43
Chatterjee, D., A. D. Roberts, K. Lowell, P. J. Brennan, I. M. Orme.
1992
. Structural basis of capacity of lipoarabinomannan to induce secretion of tumor necrosis factor.
Infect. Immun.
60
:
1249
-1253.
44
Spohn, R., U. Buwitt-Beckmann, R. Brock, G. Jung, A. J. Ulmer, K. H. Wiesmuller.
2004
. Synthetic lipopeptide adjuvants and Toll-like receptor 2: structure-activity relationships.
Vaccine
22
:
2494
-2499.
45
Jin, M. S., S. E. Kim, J. Y. Heo, M. E. Lee, H. M. Kim, S. G. Paik, H. Lee, J. O. Lee.
2007
. Crystal structure of the TLR1-TLR2 heterodimer induced by binding of a tri-acylated lipopeptide.
Cell
130
:
1071
-1082.