Acylation State of the Phosphatidylinositol Hexamannosides from Mycobacterium bovis Bacillus Calmette Gue´rin and Mycobacterium tuberculosis H37Rv and Its Implication in Toll-like Receptor Response*

The dimannoside (PIM 2 ) and hexamannoside (PIM 6 ) phosphatidyl- myo -inositol mannosides are the two most abundant classes of PIM found in Mycobacterium bovis bacillus Calmette Gue´rin, Mycobacterium tuberculosis H37Rv, and Mycobacterium smegmatis 607. Recently, these long known molecules received a renewed interest due to the fact that PIM 2 constitute the anchor motif of an important constituent of the mycobacterial cell wall, the lipoarabinomannans (LAM), and that both LAM (phos-phoinositol-capped LAM) and PIM are agonists of Toll-like receptor 2 (TLR2), a pattern recognition receptor involved in innate immunity. Due to the biological importance of these molecules, the chemical structure of PIM was revisited. The structure of PIM 2 was recently pub- lished (Gilleron, M., Ronet, C., Mempel, M., Monsarrat, B., Gachelin, G., and Puzo, G. (2001) J. Biol. Chem. 276, 34896– 34904). Here we report the purification and molecular characterization of PIM 6 in their native form.

A variety of phosphatidyl-myo-inositol mannosides (PIM) 1based compounds are known to be part of the mycobacterial cell wall. Among these compounds are the lipoarabinomannans (LAM) and the lipomannans (LM). The importance of these lipoglycans in the immunopathogenesis of tuberculosis is now established, and a rising number of studies in the literature are devoted to the delineation of their biological activities (2,3). PIM, LM, and LAM derive from the same biosynthetic pathway as demonstrated using biochemical (1,2,4,5) and genetic approaches (6 -8). PIM appear to be the common anchor of LM and LAM, as LM correspond to polymannosylated PIM and then give rise to LAM by further glycosylation with arabinosyl units. This anchor plays a fundamental role in the biological functions of LAM. Indeed, it is now clearly established that most of the LAM immunoregulatory effects are abolished by alkaline hydrolysis, highlighting the importance of the lipidic part of the anchor (2).
PIM are known from the 1940s and have been structurally investigated by Ballou and co-workers in the 1960s (9). By 1965, studies of deacylated PIM from M. tuberculosis and Mycobacterium phlei revealed the structure of the saccharidic part. PIM 6 were the highest PIM which were fully characterized from M. phlei (10) and were shown to contain a pentamannoside of sequence Manp␣132Manp␣132Manp␣13 6Manp␣136Manp␣13 attached to position 6 of the myo-inositol, whereas a Manp unit is linked to the position 2 of the myoinositol. Recently, the complete structure of native PIM 2 has been achieved (1,2). These last studies focused on the characterization of their lipidic part and unambiguously established the existence of a tetra-acylated form that was thus far suggested (4).
Several biological functions have been recently attributed to PIM. PIM 2 were shown to recruit natural killer T cells, which have a primary role in the local granulomatous response (1,11). Moreover, a role for surface-exposed PIM as M. tuberculosis adhesins that mediate attachment to non-phagocytic cells has also been established (12,13). Analysis of infected macrophages revealed that PIM, among other mycobacterial lipids, are actively trafficked out of the mycobacterial phagosome (14). This could be of particular importance relating to the potential role played by these constituents in extending the influence of the bacterium over its surroundings. An unfractionated preparation of PIM, as well as phosphoinositol capped LAM (PI-LAM), was recently shown to activate cells via Toll-like receptor-2 (TLR-2) (15). Activation of TLR-dependent signaling pathways leads to the activation of genes that participate in innate immune responses, such as expression of cytokines, coactivation molecules, and nitric oxide (16,17). Finally, PIM 6 as well as ManLAM from Mycobacterium leprae and M. tuberculosis are presented by antigen-presenting cells in the context of CD1b (18). The high affinity interaction of CD1b molecules with the PIM 2 acyl side chains was then established (19). The phosphatidylinositol moiety plays a central role in the process of PIM and ManLAM binding to CD1b proteins.
Here we investigated the structure of the most polar PIM isolated from M. bovis BCG, PIM 6 . The earlier NMR studies conducted on PIM 2 and PIM 6 by Severn et al. (20) focused on the deacylated molecules, thus excluding the study of the lipidic moieties. In this study, we investigated the chemical structure of native PIM 6 , focusing on the characterization of the different "acyl forms," using sophisticated analytical tools such as MALDI-MS and two-dimensional NMR. Then we demonstrated the capacity of the PIM 2 and PIM 6 acyl forms to stimulate macrophages to produce cytokines, and we investigated the implication of the different TLR in this process.

EXPERIMENTAL PROCEDURES
PIM Extraction-The PIM-containing lipidic extract was obtained through purification of the phenolic glycolipids from M. bovis BCG 1173P2 (the Pasteur strain) (21) and was briefly summarized in Ref.
1. An acetone-insoluble phospholipids-containing lipid extract was prepared (1) and applied to a QMA-Spherosil M (BioSepra SA, Villeneuvela-Garenne, France) column that was first irrigated with chloroform, chloroform/methanol (1:1, v/v), methanol in order to elute neutral compounds. Phospholipids were eluted using ammonium acetate containing organic solvents. Indeed, 0.1 M ammonium acetate in chloroform/methanol (1:2, v/v) (fraction A) allowed elution of 750 mg of phospholipids (enriched in phosphatidyl-myo-inositol di-mannosides (PIM 2 )),and 0.2 M ammonium acetate in chloroform/methanol (1:2, v/v) (fraction B) was subdivided into two fractions. The first volumes allowed elution of 440 mg of phospholipids (cardiolipids essentially), and the next ones allowed elution of 160 mg of phospholipids (mixture of phosphatidyl-myoinositol dimannosides (PIM 2 ) and hexamannosides (PIM 6 )), and finally, 0.2 M ammonium acetate in methanol (fraction C) allowed elution of 55 mg of phospholipids (enriched in PIM 6 ). Repeated lyophilizations were necessary to eliminate ammonium acetate salts. PIM from M. smegmatis (ATCC 607) and M. tuberculosis H37Rv (ATCC 27294) were also analyzed. 2 Purification of the PIM Acyl Forms-Fraction C (20 mg) was loaded in 0.1 M ammonium acetate solution containing 15% (v/v) propanol-1 to octyl-Sepharose CL-4B (Amersham Biosciences) column (20 ϫ 1.5 cm) pre-equilibrated with the same buffer. The column was first eluted with 50 ml of equilibration buffer and then with a linear propanol-1 gradient from 15 to 65% (v/v) (250 ml each) in 0.1 M ammonium acetate solution at a flow rate of 5 ml/h. The fractions were collected every 30 min. 20 l of each fraction was dried and submitted to acidic hydrolysis (100 l of trifluoroacetic acid, 2 M, 2 h at 110°C). The hydrolysates were dried, reconstituted in water, and then analyzed by high pH anion exchange chromatography for mannose content giving the presented chromatogram ( Fig. 2A). Fractions were pooled according to the purification profile, and repeated lyophilizations were performed to eliminate ammonium acetate salts. An acetone precipitation step was done on each fraction in order to eliminate contaminants issued from the propanol-1. Finally, 1.2 (fraction I), 1 (fraction II), 7.5 (fraction III), and 3 mg (fraction IV) were obtained.
Acetolysis Procedure-200 g of PIM were treated with 200 l of anhydrous acetic acid-d 4 /acetic anhydride-d 6 , 1:1 (v/v), at 110°C for 12 h. The reaction mixture was dried under stream of nitrogen and was submitted to acetylation. The mixture was dissolved in acetic anhydride/anhydrous pyridine, 1:1 (v/v), at 80°C for 2 h. The reaction mixture was dried under stream of nitrogen. 20 l of chloroform/methanol, 9:1 (v/v), was added and analyzed in MALDI-Tof-MS in positive and negative modes.
Matrix-assisted Laser Desorption/Ionization-Mass Spectrometry (MALDI-Tof-MS)-Analysis by MALDI-Tof-MS was carried out on a Voyager DE-STR (PerSeptive Biosystems, Framingham, MA) using the reflectron mode. Ionization was effected by irradiation with pulsed UV light (337 nm) from an N 2 laser. PIM were analyzed by the instrument operating at 20 kV in the negative ion mode using an extraction delay time set at 200 ns. Typically, spectra from 100 to 250 laser shots were summed to obtain the final spectrum. All of the samples were prepared for MALDI analysis using the on-probe sample cleanup procedure with cation-exchange resin (22). The HABA matrix (from Sigma) was used at a concentration of ϳ10 mg/ml in ethanol/water (1:1, v/v). Typically, 0.5 l of PIM sample (10 g) in a CHCl 3 /CH 3 OH/H 2 O solution and 0.5 l of the matrix solution, containing ϳ5-10 cation exchange beads, were deposited on the target, mixed with a micropipet, and dried under a gentle stream of warm air. The measurements were externally calibrated at two points with PIM.
NMR Analysis-NMR spectra were recorded with an Avance DMX500 spectrometer (Bruker GmbH, Karlsruhe, Germany) equipped with an Origin 200 SGI using Xwinnmr 2.6. Samples were dissolved in CDCl 3 /CD 3 OD/D 2 O, 60:35:8 (v/v/v), and analyzed in 200 ϫ 5-mm 535-PP NMR tubes at 308 K. Proton chemical shifts are expressed in ppm downfield from the signal of the chloroform (␦ H /TMS 7.27 and ␦ C /TMS 77.7). The one-dimensional phosphorus ( 31 P) spectra were measured at 202 MHz with phosphoric acid (85%) as external standard (␦p 0.0). All the details concerning NMR sequences used and experimental procedures were detailed in previous study on PIM 2 (1).
Alternatively the macrophages were infected with M. bovis BCG (Pasteur strain 1173P2; kind gift from G. Marchal, Pasteur Institute, Paris, France; at a multiplicity of infection of 2 bacteria per cell). After 18 h of stimulation, the supernatants were harvested and analyzed for cytokine content using commercially available enzyme-linked immunosorbent assay reagents for TNF-␣ and IL-12p40 (Duoset R & D Systems, Abingdon, UK). 2

Purification of the PIM 6 Acyl Forms from M. bovis BCG-An
enriched fraction of phosphatidyl-myo-inositol hexamannosides (PIM 6 ) was purified from M. bovis BCG as described previously (fraction C (1)). Briefly, PIM are known to be found in the acetone-insoluble fraction of mycobacterial lipidic extract (28). The contaminating neutral compounds were eliminated by QMA anion exchange chromatography, irrigated with neutral eluents. The phospholipids were then eluted with ammonium acetate-containing organic solvents resulting in three fractions, A-C. These fractions were analyzed by ESI-MS in negative mode (1) and revealed that fraction A mainly includes PIM 2 containing a total of three and four fatty acids in addition to phosphatidyl-myo-inositol; fraction B contains PIM 2 and PIM 6 containing a total of three and four fatty acids, and fraction C mainly contains the different acyl forms of PIM 6 . Fig. 1 presents the negative MALDI spectrum of M. bovis BCG fraction C that is dominated by peaks assigned to deprotonated molecular ions (M Ϫ H) Ϫ revealing the different acyl forms of PIM 6 present in this fraction. From the predominant fatty acids deduced from gas chromatography-mass spectrometry analysis (not shown), i.e. palmitic (C 16 ), tuberculostearic (C 19 ), and stearic (C 18 ) acids and in minor amounts heptadecanoic acids (C 17 ), the fatty acid composition of each acyl form was determined. We then confirmed mono-acylated forms (1Ac) with C 16 (m/z 1543.6) or C 19 (m/z 1585.7), di-acylated forms (2Ac) with 2C 16 (m/z 1781.8) or 1C 16 ,1C 19 (m/z 1823.9), triacylated forms (3Ac) mainly with 2C 16 ,1C 19 (m/z 2062.1), and tetra-acylated forms (4Ac) mainly constituted by 3C 16 ,1C 19 (m/z 2300.3) or 2C 16 ,2C 19 (m/z 2342.4). M. tuberculosis H37Rv was found to produce as major PIM families PIM 2 and PIM 6 in the same acyl forms as the ones described for M. bovis BCG (not shown).
To proceed in the separation of the acyl forms, 20 mg of fraction C were applied on an octyl-Sepharose column, using propanol-1 as eluent ( Fig. 2A). The different acyl forms were eluted at concentrations of propanol-1 ranging from 25 to 50% and separated into 4 sub-fractions according to the profile elution determined by the mannose content. Each sub-fraction (I to IV) was collected and analyzed by negative MALDI-Tof-MS.
MALDI-Tof-MS Characterization of the Octyl-Sepharose Column Fractions-The mass spectra of the fraction I (Fig. 2, B-C) revealed that it contained mono-acylated forms of the mole- , and for a lesser extent stearate (C 18 ) and heptadecanoate (C 17 ). B and C, fraction I corresponded to mono-acylated forms, acylated by 1C 16 (m/z 1543.6) or 1C 19 (m/z 1585.7) (a, fraction Ia corresponded to fraction 52 of the octyl-Sepharose, and b, fraction Ib corresponded to fraction 62 of the octyl-Sepharose). D, fraction II corresponded to di-acylated forms, acylated by 2C 16 (m/z 1781.8) and C 16 ,C 19 (m/z 1823.9) and in a minor part by C 16 ,C 17 (m/z 1795.8). E, fraction III corresponded to tri-acylated forms, acylated predominantly by 2C 16 ,1C 19 (m/z 2062.1), and in a minor part by C 16 ,C 18 ,C 19 (m/z 2090.1). F, fraction IV corresponded to tetra-acylated forms, acylated predominantly by 3C 16 19 , respectively. Species esterified by unsaturated fatty acids were also present as each peak consisted of a Ϫ2 analog. cules (1Ac). Indeed, the analysis of fraction Ia revealed deprotonated molecular ions at m/z 1543.6 characterizing monoacylated forms with C 16 (Fig. 2B), whereas mono-acylated forms with C 19 appeared more abundant in fraction Ib (m/z 1585.7) (Fig. 2C). The mass spectrum corresponding to fraction II showed three peaks with an intensity above 10% and were assigned to (M Ϫ H) Ϫ ions of di-acylated forms: two major peaks at m/z 1781.8 and 1823.9 and one minor one at m/z 1795.8 (Fig. 2D). The two major peaks were characterized as (M Ϫ H) Ϫ ions of di-acylated forms containing 2C 16 and 1C 16 ,1C 19 , respectively, and the minor peak corresponded to (M Ϫ H) Ϫ ions of the molecules acylated by 1C 16 ,1C 17 . The mass spectrum of fraction III (Fig. 2E) revealed two peaks of intensity superior to 10% assigned to tri-acylated forms of the molecules. The major peak at m/z 2062.1 was attributed to (M Ϫ H) Ϫ ions corresponding to tri-acylated forms containing 2C 16 ,1C 19 , whereas the minor one at m/z 2090.1 contains 1C 16 ,1C 18 , and 1C 19 . The mass spectrum of fraction IV appeared more complex, constituted by a series of peaks (Fig. 2F 16 and 1C 17 ,2C 19 , respectively. In addition, species esterified by unsaturated fatty acids were also present as each peak consisted of a Ϫ2 analog. Therefore, we have developed a powerful preparative method of fractionation, leading to the purification of four purified PIM 6 , corresponding to mono-, di-, tri-, and tetra-acyl forms. The structural analysis is further detailed below for the most complex entities, tri-and tetra-acyl forms. Glycosidic Analysis of Native Tetra-acyl Forms-The sequence of glycosyl residues in PIM 6 was established using a range of high resolution NMR techniques applied to the tetraacylated molecules. The 1 H NMR spectrum of the native molecules in CDCl 3 /CD 3 OD/D 2 O, 60:35:8 (v/v/v), at 500 MHz showed a complex anomeric proton region (between 4.4 and 5.1 ppm) (Fig. 3). Anomeric signals were investigated thanks to the 1 H-13 C HMQC spectrum (Fig. 3). Indeed, from the protons resonating between 4.4 and 5.1 ppm, two of them correlated with carbons out of the anomeric zone: proton at ␦ 5.02 correlated with a carbon at 70.6 ppm and proton at ␦ 4.53 correlated with a carbon at 71.7 ppm. From previous studies (1, 2), they were respectively assigned to H2 of Gro and H3 of myo-Ins. These protons are deshielded due to the presence of gem acyl group. Moreover, the coupling constants measured on the proton at ␦ 4.53 (J 2,3 2.4 Hz and J 3,4 10.6 Hz) confirmed its attribution to myo-Ins H3. The other proton resonances correlating with carbons resonating around 100 ppm accounted for the six mannose units labeled from I to VI in decreasing order of their chemical shifts. Indeed, integration of the anomeric signals (H1) proved that the signal resonating at 4.88 ppm corresponded to two protons (II 1 and III 1 ). The ␣-anomeric configuration of the mannoses was deduced from the values of the one bound coupling constant ( 1 J C1,H1 ) above 170 Hz, measured on non-decoupled 1 H-13 C HMQC spectrum (not shown).
The H2 protons of each ␣-Manp unit was determined using the COSY spectrum (Fig. 4A) although H2 of system I was partly hidden by the H3/H3Ј of glycerol (Gro). The entire spin system of Gro can be analyzed from the COSY spectrum (Fig.   FIG. 3. 1 H- 4A) and those of the myo-Ins from the HOHAHA spectrum (Fig.  4B). The six mannose spin systems could be completely assigned (Table I) using 1 H-1 H HOHAHA (Fig. 4B) and 1 H-13 C HMQC (Fig. 3).  (Table I). Thus, these data demonstrated that this unit is acylated in position 6.
The glycosidic sequence was next deduced from the interresidual nuclear Overhauser effect contacts observed in the 1 H-1 H ROESY spectrum (Fig. 4C). This sequence was used here to observe short through space connectivities between the anomeric proton of each mannose and the proton of the adjacent glycosidically linked residue. The anomeric proton of spin system I (␦ 5.02) correlated with H2 of system IV, indicating that ␣-Manp unit I is linked to the C2 position of ␣-Manp unit IV. In the same way, the occurrence of cross-peaks relating H1 The H1 of system III showed a nuclear Overhauser effect contact with proton 2 of myo-Ins, revealing that this unit is linked in position 2 of the myo-Ins. The H1 of system II showed two cross-peaks with protons 1 and 6 of myo-Ins. As explained below, two-dimensional 1 H-31 P HMQC analysis allowed us to define one of the positions of substitution of the phosphorus as position 1 of the myo-Ins. Thus, system II was deduced to be linked in position 6 of the myo-Ins.
Anchor Structure of Native Tetra-acyl Forms-The NMR strategy developed to study PIM 2 containing a total of 4 fatty acids (1) was used here. From the 1 H-31 P HMQC experiment, the prochiral H-3 and H-3Ј Gro protons and the myo-Ins H-1 were easily assigned (not shown). The remaining myo-Ins and Gro protons were then observed on the two-dimensional 1 H-31 P HMQC-HOHAHA spectrum (Fig. 6B) and were assigned from their multiplicity and chemical shifts (29) and with the help of the 1 H-1 H HOHAHA spectrum (Fig. 6, C-D). The different chemical shifts typified the presence or absence of an acyl appendage. The chemical shifts of the H-1 (4.19/3.95 ppm) and H-2 (5.02 ppm) revealed a di-acyl-Gro (29). The downfield shift of the myo-Ins H-3 resonating at 4.53 ppm then signed the C3 acyl appendage. A fourth position of acylation, the C6 of the Manp, could be assigned by analysis of the two-dimensional 1 H-13 C HMQC spectrum. Indeed, the chemical shifts of H-6/ H-6Ј of ␣-Manp unit III (4.02/4.15 ppm) proved that this position is also acylated. The slight deshielding of C6 resonance (from 61.5 to 63.9 ppm) confirmed this assignment.
Therefore, the analytical approach applied to this acyl form of PIM 6 (containing four fatty acids in total) revealed that the four positions of acylation were the same as the ones described in case of the corresponding acyl form of PIM 2 (1): the positions C1 and C2 of the Gro, position C3 of the myo-Ins unit, and position C6 of the Manp unit linked to C2 of the myo-Ins.
Acyl Distribution of Native Tetra-acyl Forms-The nature of the fatty acids esterifying the different sites was investigated by mass spectrometry using MALDI-Tof-MS analysis of the acetolysis reaction products of the native tetra-acyl forms of the molecules (not shown). Acetolysis cleaves the phosphoglycerol linkage without altering the fatty acid esters, leading to two entities: the hexamannosyl-inositol phosphate moiety (Man 6 -Ins-P) and the acyl-glycerol residue, as already described in Ref. 1. The hexamannosyl-inositol phosphate moiety was observed in negative mode as [M Ϫ H] Ϫ ions, whereas the acylglycerol part was analyzed in positive mode as [M ϩ Na] ϩ ions. As described previously, the acetolysis reaction produces two populations of Man 6 -Ins-P moieties that differ by the presence or absence of an acetyl group on the phosphate (1). But as this acetate present on the phosphate is very labile (mixed anhydride), it is partially hydrolyzed when the sample is mixed with the matrix (HABA diluted in EtOH/H 2 O). To discriminate between acetate groups and changes of C19 by C16, the reaction was made with perdeuterated acetic acid and acetic anhydride.
The positive MALDI-Tof-MS spectrum of the perdeuteroacetolyzed of the tetra-acylated molecules showed an intense peak at m/z 678.6 assigned to sodium adduct of the di-acylated C 16 /C 19 Gro (not shown). The negative mass spectrum mainly showed two peaks at m/z 2653.7 and 2695.8 corresponding to the Man 6 -Ins-P moiety acylated with 2C 16 and C 16 ,C 19 , respectively. Less intense peaks were also observed corresponding to the Man 6 -Ins-P moiety acylated with other combinations of fatty acids. Taken together, these results indicate that the glycerol is always acylated by C 16 ,C 19 , and we can postulate that the mannose is acylated by a C 16 , whereas the nature of the fatty acid present on the inositol is variable, being predominantly C 16 or C 19 .
Analysis of Native Tri-acyl Forms-1 H-13 C HMQC spectrum of the tri-acylated molecules exhibited the same cross-peaks as for the tetra-acylated one, except for the inositol   age on C3. The three fatty acids were found on both positions of Gro and on the ␣-Manp unit on the C2 of myo-Ins. The positive MALDI-Tof-MS spectrum of deutero-acetolyzed molecules showed a similar intense peak at m/z 678.6 assigned to the sodium adduct of the di-acylated C 16 /C 19 Gro as the one observed in the case of the tetra-acylated molecules (not shown). The negative mass spectrum mainly showed one peak at m/z 2460.5 corresponding to the Man 6 -Ins-P moiety acylated with one 1C 16 . Thus, the results indicate that the tri-acylated molecules predominantly exist with 2C 16 and 1C 19 , in agreement with the negative MALDI-Tof analysis of the native molecules (Fig. 2E), the glycerol being di-acylated by C 16 /C 19 and the mannose bearing a C 16 .
Macrophage Stimulation by PIM 2 and PIM 6 -Unfractionated PIM have been shown to stimulate the murine RAW 264.7 monocytic cell line to produce TNF-␣ (15), and we asked whether purified PIM 2 and PIM 6 were equally pro-inflammatory. First, unfractionated PIM 2 preparation was tested and shown to stimulate TNF-␣ and IL-12 p40 secretion by primary murine macrophages (not shown). Then two different well defined acyl forms of PIM 2 were assayed as follows: a fraction (F6) containing lyso-PI and lyso-PIM 2 (both containing C 16 ) and a fraction (F7) containing more acylated forms of PIM 2 (tri-acylated and tetra-acylated molecules). Both PIM 2 fractions induced TNF-␣ secretion irrespective of their acyl forms (Fig. 7). The concentration of TNF-␣ secreted was clearly sub-maximal (ϳ1 ng/ml) as compared with those achieved by stimulation with strong stimuli such as LPS or BLP or after infection with live BCG (ϳ15 ng/ml; Fig. 7A). We next asked whether PIM 6 exhibited a similar function. The unfractionated PIM 6 preparation (F1) also stimulated TNF-␣ secretion by macrophages (Fig. 7B). To determine whether the number of fatty acids played a role in the stimulation of TNF-␣ secretion, the purified acyl forms of PIM 6 (F2 to F5) were tested for their capacity to stimulate macrophages to produce TNF-␣. No clear enrichment in stimulatory activity was observed with the purified PIM 6 acylated forms (mono-to tetra-acylated molecules) as compared with the unfractionated PIM 6 (Fig. 7B). Only marginal levels of IL-12p40 were detected after incubation of macrophages with PIM 2 or PIM 6 (data not shown).
Thus, both PIM 2 and PIM 6 stimulated TNF-␣ secretion by primary macrophages, and this activity seemed independent from the number and the nature of the PIM 2 and PIM 6 acyl moieties.
TLR Dependence of PIM Activity-The unfractionated PIM preparation was shown to be a TLR2 agonist, based on reporter assay with Chinese hamster ovary cell lines transfected with the tlr2 gene (15). Here bone marrow-derived macrophages prepared from mice rendered deficient for TLR2 and/or TLR4, for TLR6, or for MyD88, the adaptor common to the different TLR, were used to investigate the TLR dependence of the PIM 2 and PIM 6 responses. The macrophages were stimulated with the fraction (F6) containing lyso-PI and lyso-PIM 2 (both containing C 16 ), and the fraction (F7) containing more acylated forms of PIM 2 (tri-acylated and tetra-acylated molecules) or with the PIM 6 fraction (F1) and TNF-␣ (Fig. 8) and IL-12 p40 (not shown) secretions were assessed. The production of TNF-␣ by primary macrophages in response to the PIM 2 and PIM 6 fractions is dependent on TLR2, as no TNF-␣ could be detected in the supernatant of macrophages deficient for TLR2, although cells deficient for TLR4 were efficiently stimulated by these fractions (Fig. 8, B and C). As expected, no cytokine production was detected in the supernatants of macrophages isolated from the double knock-out mice (TLR2Ϫ/Ϫ and TLR4Ϫ/Ϫ) (Fig. 8, B and C). Thus, PIM 6 are capable of inducing TNF-␣ secretion via a TLR2-dependent pathway. TLR2 has been shown to heterodimerize with TLR1 or TLR6 (30), and we next looked at the implication of TLR6 in the PIM-TLR2 interaction. TLR6 deficiency did not impair the ability of the macrophages to respond to PIM 2 or PIM 6 (Fig. 8, D-F). MyD88 is involved in most of the TLR-mediated signals (31), including TLR2 as shown here for the TLR2 agonist BLP and for the best of the response to the TLR4 agonist LPS (Fig. 8D), and we showed that macrophages deficient for MyD88 did not respond to the PIM 2 or PIM 6 stimulation (Fig. 8, E and F). Thus, both PIM 2 and PIM 6 stimulate TNF-␣ secretion by macrophages through TLR2 and not TLR4. TLR6 is not essential in this response which involves signaling through the adaptor MyD88. DISCUSSION The present study constitutes the first complete structural analysis of the native PIM 6 from M. bovis BCG allowing the definition of the structure of the tetra-acylated form of the molecule as depicted in Fig. 5. The saccharidic part is in complete agreement with that proposed in the 1960s (for a review, see Ref. 32) and determined for the deacylated PIM 6 by Severn et al. (20). Beyond this, the analysis presented here allows us to discriminate the different structures hiding beneath the term "PIM 6 ." Indeed, the "PIM 6 family" corresponds to a mixture of 10 or 12 acyl forms. Purification of these different species proved to be crucial for the complete structural study of the native molecules. We developed a rapid and powerful method for allowing us in a first step to separate PIM 2 from PIM 6 using a QMA column and in a second step to separate acyl forms using an octyl-Sepharose column, similar to that used to purify LAM acyl forms, as described by Leopold and Fischer (33) and applied in our laboratory by Nigou et al. (34). The purification of the PIM family from M. bovis BCG as well as M. smegmatis 607 and M. tuberculosis H37Rv revealed that very small quantities of intermediate forms (PIM 1 , PIM 3 , PIM 4 , and PIM 5 ) were found and that the more polar PIM corresponded to PIM 6 and not PIM 5 , in contrast to early conclusions (35) and still reported in the literature (5,7,36). The purification to homogeneity of the different PIM using isocratic silicic flash chromatography was unsuccessful, except for the PIM 2 containing four fatty acyl appendages in total (2). Indeed, as shown on Fig.  9, the more polar PIM 2 co-migrate with the different acyl forms of PIM 6 .
The structural strategy used to characterize PIM 6 acyl forms was similar to the one successfully used to define PIM 2 acyl forms (1) and combines the potency of complementary analytical techniques, NMR and mass spectrometry. NMR was used to characterize the acylation positions, whereas mass spectrometry gave information concerning the number of fatty acids and the nature of the fatty acids present at each position. MALDI or ESI modes were chosen rather than fast atom bombardment, because the molecules could be analyzed without derivatization preventing the loss of any labile substituents. In this study, MALDI-MS was chosen rather than ESI-MS as a same deposit could be analyzed in positive or negative mode. Indeed, concerning the analysis of the acetolysis products, the hexamannosyl-inositol phosphate moiety (Man 6 -Ins-P) was observed in negative mode, whereas the acyl-glycerol residue was analyzed in positive mode. The complete NMR study required some milligrams of purified product. This was detailed for the more complex acyl forms, the tri-acylated and the tetra-acylated ones.
The results obtained are summarized in Table II. The monoacylated forms are identified for lyso-PIM 6 with C 16 or C 19 in position 1 of the glycerol. The di-acylated forms appear as two populations almost equally represented as one with 2C 16 and one with 1C 16 and 1C 19 , both fatty acids being on the glycerol and structurally corresponding to "true" PIM 6 . In contrast, concerning tri-acylated forms, a major acyl form was observed, containing 2C 16 and 1C 19 , the glycerol being di-acylated by C 16 ,C 19 and the mannose bearing a C 16 . They should then correspond to mono-acylated-PIM 6 (or Ac 1 PIM 6 ). Interestingly, the acyl form containing 3C 16 was not observed, indicating that  F2, F3, F4, and F5 correspond to mono-(Ac1), di-(Ac2), tri-(Ac3), and tetra-acylated (Ac4) forms, respectively; F6, a fraction containing lyso-PI and lyso-PIM 2 (both containing C 16 ); F7, a fraction containing more acylated forms of PIM 2 (tri-acylated and tetra-acylated molecules). TNF-␣ was measured in the supernatants. Results are mean Ϯ S.D. from n ϭ 2 mice and are from one representative experiment out of two independent experiments. the tri-acylated forms arose from the di-acylated forms containing C 16 ,C 19 as fatty acyl appendages. The tetra-acylated forms were present as essentially two populations, 3C 16 ,1C 19 or 2C 16 ,2C 19 . The acylation positions were here clearly elucidated as being both positions of Gro, the C3 of the myo-Ins unit and the C6 of the Manp linked to the C2 of the myo-Ins unit. Taken together, the same conclusions as for the acyl forms of PIM 2 could be made. Indeed, the results indicate that the glycerol is preferentially acylated by C 16 ,C 19 . However, the nature of the fatty acid present on the myo-inositol appears highly variable, being essentially C 16 or C 19 .
As mentioned previously (1), the biosynthetic linkage between PIM 2 , LM, and LAM appears now to be more and more evident, as the same anchor structures were found for all these lipoglycans. However, even if the acyl forms found for PIM 6 were exactly the same as the ones found for PIM 2 , LM, and LAM, PIM 6 are not part of the LM/LAM biosynthetic pathway.
Indeed, the linear mannan backbone found in LM and LAM is constituted by a linear ␣-(136)-linked Manp, whereas PIM 6 exhibit at its extremity ␣-(132) links. PIM 6 appears to be an end product and may have a specific role.
Mammalian TLR proteins are pattern recognition receptors for a wide array of bacterial and viral products (17). Gramnegative bacterial lipopolysaccharide (LPS) activates cells through TLR4, whereas the mycobacterial cell wall lipoglycans, AraLAM, activate cells through TLR2. Recently, Jones et al. (15) identified a secreted TLR2 agonist activity in short term culture filtrates of M. tuberculosis, which they called STF, for "soluble tuberculosis factor." To determine the identity of the TLR2 agonist present in soluble tuberculosis factor, they used preparative SDS-PAGE. The TLR2 agonist activity was present in one fraction, with an apparent molecular size of 6 kDa, raising the possibility that the TLR2 agonist was PIM 2 . By using TLC, the authors reported the presence of two major species, PIM 1 and PIM 2 . In that case, the precise structure of the TLR2 agonist was not determined. We confirm here that structurally defined acyl forms of PIM 2 stimulated TNF-␣ secretion by primary macrophages in a TLR2-dependent fashion. We then demonstrated that the major polar PIM from M. bovis BCG and M. tuberculosis, PIM 6 , also stimulated TNF-␣ secretion and that this secretion was mediated by TLR2 but not TLR4. TLR2 has been shown to heterodimerize with TLR1 or TLR6 (30). We show here that the TLR2-mediated stimulation of macrophages by PIM 2 and PIM 6 was independent of TLR6 but that both PIM 2 and PIM 6 signaled via the adaptor molecule MyD88.
It has been demonstrated in the case of LPS that saturated fatty acids acylated in lipid A moiety are essential for its biological activities. Saturated fatty acids, but not unsaturated fatty acids, induce NF-B activation and expression of inflammatory markers in macrophages (37). In addition, it has been proposed that the shape of lipid A, influenced by the length and number of acyl chains, asymmetry of acyl groups, and distribution of negative charges, determine the interaction of LPS with different TLR (38). This would explain how E. coli LPS, with a strong conical shape lipid A, interact with TLR4, whereas Porphyromonas gingivalis or Rhodobacter sphaeroides LPS, with a more cylindrical shape lipid A, are TLR2 agonists (38). Here we asked whether different PIM 2 and PIM 6 acyl forms, bearing 1-4 acyl residues, were equally potent in stimulating macrophages to produce TNF-␣. We show that the TNF-␣ stimulating activity of PIM 6 is independent from the number and the nature of the acyl moieties present on the molecules. This seemed to be also the case for PIM 2 as both mono-(F6) and tri-and tetra-acylated (F7) forms studied here had similar activity.
In infected macrophages, PIM were shown to traffic out of the mycobacterial phagosome among other mycobacterial lipids, such as LAM, and are released to the medium and bystander uninfected cells (14). We show here that both major PIM species from M. bovis BCG, M. smegmatis 607, and M. tuberculosis H37Rv, PIM 2 and PIM 6 , are agonists of TLR2, irrespective of their acylation state. ManLAM from slow-growing mycobacteria such as M. bovis or M. tuberculosis do not show such pro-inflammatory effects or TLR2-dependent activity (39). 2 TLR2 may be present intracellularly, recruited to macrophage phagosomes, where they sample the content of the vacuole, and contribute to elaborate an inflammatory response appropriate for defense against a specific pathogen (40). The trafficking and export of PIM 2 and the more polar PIM 6 from M. bovis or M. tuberculosis could thus contribute to maintaining the smoldering activation state of the infected macrophage and of the neighboring cells in the tuberculous granuloma through TLR2 activation, and contribute to the innate immunity necessary to contain latent infection.  6 acyl forms evidenced in M. bovis BCG The relative abundance of the different species for each acyl form was determined from the integration of the corresponding monoisotopic signals in the negative MALDI spectrum of the PIM 6 -enriched fraction C from M. bovis BCG (Fig. 1).