Acylation state of the phosphatidylinositol mannosides from Mycobacterium bovis bacillus Calmette Guérin and ability to induce granuloma and recruit natural killer T cells.

Previous studies have found that, when injected into mice, glycolipidic fractions of mycobacterial cell walls containing phosphatidylinositol mannosides (PIM) induced a granuloma and recruitment of Natural Killer T cells in the lesions. The dimannoside (PIM(2)) and the hexamannoside (PIM(6)) PIM were isolated from Mycobacterium bovis bacillus Calmette Guérin and shown to act alike, but the activity was found to be dependent on the presence of the lipidic part. The chemical structure of PIM was then re-evaluated, focusing on the characterization of their lipidic part, defining mono- to tetra-acylated PIM(2). The structure of these acyl forms was elucidated using a sophisticated combination of chemical degradations and analytical tools including electrospray ionization/mass spectrometry, electrospray ionization/mass spectrometry/mass spectrometry, and two-dimensional NMR. Finally, the acyl forms were purified by hydrophobic interaction chromatography and tested for their capacity to induce the granuloma and Natural Killer T cell recruitment. We found that there is an absolute requirement for the molecules to possess at least one fatty acyl chain, but the number, location, and size of the acyl chains was without effect. Moreover, increasing the complexity of the carbohydrate moiety did not lead to significant differences in the biological responses.

group of phospholipids found in the cell wall and the cytoplasmic membrane of mycobacteria along with cardiolipid, phosphatidylethanolamine, and phosphatidylinositol. Known from the 1940s and fully described by Ballou and colleagues (1) in the 1960s, they were shown to consist of phosphatidyl-myoinositol di-, tri-, tetra-, and pentamannosides (PIM 2 to PIM 5 ). These authors showed unequivocally that in PIM 2 the mannosyl units were glycosidically attached at positions 2 and 6 of the myo-Ins ring (2) and that for the more glycosylated forms, chain elongation occurred on the mannose present at position 6 (3). Recently, PIM from Mycobacterium smegmatis were reanalyzed in their deacylated form (4). They were shown to have a structure based on that defined by Ballou et al. (1) but containing six mannosyl residues (PIM 6 ). The structure of the glycosidic part of PIM 6 is now unambiguously established and corresponds to Manp␣-13(2-Manp␣-1) 2 3(6-Manp␣-1) 2 3 linked to position 6 of the myo-Ins ring beside the mannosyl residue present at position 2. This structure was confirmed by an NMR spectroscopy strategy applied to deacylated PIM 6 (5).
Concerning the lipidic part of the PIM, the situation is still unclear, although the presence of multiacylated forms of PIM was already reported in 1960s by several authors (6 -9) working on different mycobacterial strains. More recently, in a chromatographic and fatty acid quantitation study of mycobacterial lipoglycans, Leopold and Fisher (10) also inferred the existence of multiacylated forms of PIM. However, the sites of attachment of the acyl groups were not established. Re-evaluating the multiacylated nature of Mycobacterium tuberculosis and Mycobacterium leprae PIM using FAB-MS analysis, Brennan's group (11) presented evidence for tri-acylated PIM, assuming an extra fatty acyl group on the C-6 of one of the Manp residue beside the fatty acids already localized on the C-1 and C-2 of the glycerol. Our data derived from an NMR study of a tetra-acylated form of PIM 2 (Ac 4 PIM 2 ) purified from Mycobacterium bovis BCG clearly inferred the previously described positions and demonstrated the fourth acylation site as the C-3 of the myo-Ins (12).
Recently, PIM received a renewed interest due to the fact that PIM 2 constitute the anchor motif of two important constituents of the mycobacterial cell walls, namely lipomannans (LM) and lipoarabinomannans (LAM). LAM, ubiquitous in the Mycobacterium genus, exhibit a wide spectrum of immunoregulatory effects (for reviews see Refs. [13][14][15]. It is now clearly established that most of these effects are abolished by alkaline hydrolysis, highlighting the importance of the lipidic part of the anchor. Moreover, PIM 6 as well as ManLAM from M. leprae and M. tuberculosis Erdman strain are recognized by human CD4 Ϫ CD8 Ϫ ␣/␤ T cells in the context of a presentation by CD1b-expressing antigen-presenting cells (16). The high affin-* 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. ity interaction of CD1b molecules with the PIM 2 acyl side chains was then established (17). The phosphatidylinositol moiety plays a central role in the process of PIM and ManLAM binding to CD1b proteins. In the mouse, CD1a, -b, and -c genes are missing. The murine CD1d gene is homologous to the CD1d gene in man (18). This CD1d molecule restricts CD4 ϩ or DN T cells (NK T cells) using a distinctive TCR characterized by an invariant ␣ chain (19 -21). The crystal structure of the mouse CD1d1 molecule showed a hydrophobic binding site (22), and cellular glycosylphosphatidylinositol was proposed as a major natural ligand (23). Due to the homology between CD1a, -b, -c, and -d molecules, it has been hypothesized that mouse CD1d may bind mycobacterial glycolipids. In this respect, a recent study reported that granuloma formation in response to a crude fraction containing PIM from H37Rv M. tuberculosis was dependent upon CD1d-restricted T cells (20). However, no reports have mentioned PIM presentation by CD1d molecules, and the respective roles of the lipidic and carbohydrate moieties in the PIM-induced response was not established.
In this study, the chemical structure of PIM isolated from M. bovis BCG was evaluated, focusing on the characterization of the lipidic part of the PIM, thus defining the notion of "acyl forms." The structure of these acyl forms was elucidated using sophisticated analytical tools such as ESI-MS, ESI-MS/MS, and two-dimensional NMR. Finally, these acyl forms were purified and used in a bioassay aimed at determining the extent of inflammation they caused and the recruitment of NK T cells into the granulomatous lesions they induced.

EXPERIMENTAL PROCEDURES
Mice-C57BL/6 mice were 6 -8-week-old and obtained from IFFA-Credo (L'Arbresle, France). Mice were injected subcutaneously with 50 g of each molecule and adsorbed onto 12.5 l of a neutral carrier (Alu-Gel-S, Serva) in a total volume of 0.1 ml. The cellular infiltrate at the site of the injection was excised 7 days after the injection and either fixed for histological analysis or used for mRNA characterization.
RNA Extraction and cDNA Synthesis-Total RNA was isolated by using the Trizol technique (Life Technologies, Inc.). cDNA was synthesized from 10 g of total RNA, using a (dT) 17 primer, 25 units of RNasin (Promega Corp., Madison, WI), and 10 units of avian myeloblastosis virus-reverse transcriptase (Roche Molecular Biochemicals) in the provided buffer.
mRNA Quantification, PCR Procedure, and Immunoscope Analysis-HGPRT-standardized amounts of cDNA solution were PCR-amplified using V␣14-and C␣-specific primers (20). PCR products were subjected to primer extension using an internal, fluorescently labeled C␣-specific primer (to approach the total usage of V␣14 ϩ segments) or a J␣281specific primer (to generate only signals due to the NK T-specific rearranged V␣14-J␣281 TCR ␣ chain) (20). Run-off products were analyzed on an automatic sequencer. The size and the intensity of each band were recorded and analyzed using the Immunoscope software (Applied Biosystems, Inc.). Typically, a random usage of all rearranged V␣14 ␣ chains is depicted as a series of peaks corresponding to in-frame mRNA, each corresponding to a define length of the CDR3 region. A selected usage of a rearranged chain appears as a peak that distorts the abovementioned gaussian-like distribution; the NK T-specific V␣14-J␣281 TCR ␣ chain appears as a peak with a CDR3 length of 10 amino acids (20).
PIM Extraction-The PIM-containing lipidic extract was obtained during the M. bovis BCG phenolic glycolipids purification (24). Briefly, 1300 g of M. bovis BCG cells were suspended in chloroform/methanol (1:1, v/v) and filtered. The cells were suspended four times in the same way. The whole lipid extract was concentrated and partitioned between water and chloroform. The chloroform phase was evaporated, and the residue was suspended in hot acetone and then allowed to deposit overnight at 4°C. The suspension was then centrifuged at 4°C for 20 min (3000 ϫ g) yielding 40 g of acetone-insoluble phospholipids-containing lipid extract. 7 g of this extract was applied to a QMA-Spherosil M (BioSepra SA, Villeneuve-la-Garenne, France) column (3.0 ϫ 20 cm) that was first irrigated with 300 ml of 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) (frac-tion A) allowed elution of 750 mg of phospholipids (phosphatidyl-myoinositol dimannosides (PIM 2 ) essentially); 0.2 M ammonium acetate in chloroform/methanol (1:2, v/v) (fraction B) was subdivided into two fractions as follows: the 1st volumes allowed elution of 440 mg of phospholipids (cardiolipids essentially) and the next ones allowed elution 160 mg of phospholipids (mixture of phosphatidyl-myo-inositol di-(PIM 2 ) and hexa-(PIM 6 ) mannosides), and finally, 0.2 M ammonium acetate in methanol (fraction C) allowed elution of 55 mg of phospholipids (PIM 6 essentially). Repeated lyophilizations were necessary to eliminate ammonium acetate salts.
Purification was checked by TLC, on aluminum backed plates of silica gel (Alugram Sil G, Macherey-Nagel, Duren, Germany) using as migration solvent chloroform/methanol/water, 60:35:6 (v/v/v). A sulfuric anthron spray and a Dittmer-Lester spray were used to detect carbohydrates containing lipids and phosphorus-containing lipids, respectively.
Acetolysis Procedure-200 g of PIM were treated with 200 l of anhydrous acetic acid/acetic anhydride, 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. 200 l of chloroform/methanol, 9:1, v/v was added, and negative mass spectrum was recorded using about 100 l of mixture. The remaining mixture was dried and then dissolved in 100 l of 0.1 M ammonium acetate in chloroform/methanol, 9:1, v/v, and positive mass spectrum was recorded. The same protocol was applied using perdeutero-acetylated reactives in order to deduce the number of acetates present after peracetylation.
Electrospray Mass Spectrometry (ESI-MS)-The phospholipids were dissolved in chloroform/methanol/water (60:35:8, v/v/v) in order to be analyzed in ESI-MS. ESI-MS analyses were carried out on a Finnigan Mat TSQ 700 triple quadrupole mass spectrometer (Finnigan Mat, San Diego, CA) system. The negative ESI mode (spray potential 3.8 kV, temperature of heated capillary 140°C) and positive ESI mode (spray potential 4.8 kV) were used. The mass spectra were obtained by direct infusion using a syringe pump (Harvard Apparatus) at a flow rate of 4 l min Ϫ1 . Full scan spectra were acquired in the ion peak centroid or profile modes over the mass/charge range of 200 -3000 at 2.0 -3.0 s. Tandem mass spectrometry (MS/MS) experiments were based by conducting collision-induced dissociation occurring in the radio-frequencyonly collision cell of the triple quadrupole at a collision energy of 40 -80 eV. Argon was used as collision gas in the range of 1.5-2.5 mtorr. At least 15 scans were accumulated and averaged. Data analysis was controlled by a DEC ␣ 500 data system.
All two-dimensional NMR data sets were recorded without sample spinning, and data were acquired in the phase-sensitive mode using the time-proportional phase increment (25). Two-dimensional homonuclear Hartmann-Hahn (HOHAHA) spectrum was recorded using MLEV-17 mixing sequence of 30 ms (26). 1 H-13 C correlation spectra were recorded in the proton-detected mode with a Bruker 5-mm 1 H-broadband tunable probe with reversal geometry. The 1 H-13 C and 1 H-31 P single-bond cor-relation spectra (HMQC) were obtained according Bax's pulse sequence (27). The GARP sequence (28) at the carbon or phosphorus frequency was used as a composite pulse decoupling during acquisition. Multiplebond 1 H-13 C correlation spectra (HMBC) (29) were processed in the magnitude mode. The pulse sequence used for 1 H-detected heteronuclear relayed spectra (HMQC-HOHAHA) was that of Lerner and Bax (30).

Isolation of PIM Acyl Forms Inducing Granuloma-From
previous data (31), phosphatidyl-myo-inositol mannosides (PIM) are known to be found in the acetone-insoluble fraction of the chloroform/methanol (1:1, v/v) mycobacterial extract. The contaminating neutral compounds were eliminated by QMA anion exchange chromatography, irrigated with neutral eluents (chloroform, followed by chloroform/methanol, 1:1, v/v and then by methanol). Then, phospholipids were eluted with ammonium acetate-containing organic solvents resulting in three fractions, namely A-C. Each fraction was adsorbed onto alum and injected subcutaneously to C57BL/6 mice. Seven days after the injection, the cell infiltrates were excised, sectioned, and stained. The three fractions generated a highly organized and large granuloma with a central core of neutrophils surrounded by a densely organized rim of macrophages, lymphocytes, and fibroblasts ( Fig. 1, E and F), mimicking the granuloma observed during murine tuberculosis. By contrast, a minor inflammatory reaction was evidenced in PBS-injected (control mice) (Fig. 1, A and B) or deacylated PIM-injected mice ( Fig. 1, C and D). In this case, the lesions were found to predominantly contain a disordered array of macrophages and neutrophils with few lymphocytes. This points to the absolute requirement for the lipidic moiety of PIM to cause a granulomatous lesion.
In order to characterize the PIM composition of the different fractions, the fractions A-C were analyzed by ESI-MS in negative mode. The attribution of the deprotonated molecular ions (M Ϫ H) Ϫ (Fig. 2) was based on the predominant fatty acids as deduced from GC/MS analysis (not shown), i.e. palmitic (C 16 ), tuberculostearic (C 19 ), and stearic (C 18 ) acids. ESI-MS mass spectrum of fraction A predominantly showed ions corresponding to Ac 3 -(m/z 1415.0) and Ac 4 -(m/z 1653.5, 1681.5 and 1695.5) PIM 2 in addition to the diacyl form of phosphatidylinositol (Ac 2 PI at m/z 852.3) (Fig. 2a). Fraction B mass spectrum ( Fig. 2b) appeared more complex showing predominantly the previously described Ac 3 -and Ac 4 PIM 2 and also Ac 3 -(m/z 2063.7) and Ac 4 -(m/z 2302.1, 2330.2, and 2344.2) PIM 6 . Finally, the mass spectrum of fraction C (Fig. 2c) was dominated by one peak at m/z 2063.7 assigned to (M Ϫ H) Ϫ of Ac 3 PIM 6 , beside the other acyl forms of PIM 6 . In summary, tri-and tetra-acylated forms of PIM 2 and PIM 6 appeared to be the most abundant PIM acyl forms of M. bovis BCG, and they may be responsible for the granuloma formation.
Purification of the PIM Acyl Forms-The first attempt to go further in the separation of the different PIM acyl forms consisted of using silicic phase flash-chromatography with chloroform/methanol/water in different proportions as eluents. By using this strategy, Ac 4 PIM 2 was purified from fraction A with a chloroform/methanol/water, 70:30:2, v/v/v eluent (12). The fractionation of the more polar PIM 2 and PIM 6 was unsuccessful using successive isocratic flash chromatographies. An alter- native approach was developed, using hydrophobic interaction chromatography on an octyl-Sepharose column, with propanol-1 as eluent. Fig. 3 shows the purification of PIM 2 acyl forms issued from fraction A. The different acyl forms were eluted at propanol-1 concentrations ranging from 25 to 50% and separated into 5 sub-fractions according to the mannose content profile. Each sub-fraction (I-V) was collected and analyzed by ESI-MS and NMR.
Strategy for the Structural Elucidation of the Acyl Forms, the Example of Ac 3 PIM 2 -The negative ESI mass spectrum of fraction IV is dominated by the signal at m/z 1414.2 (Fig. 4c), beside peaks at m/z 1442.2 and m/z 1456.2 assigned to deprotonated molecular ions of tri-acylated forms of PIM 2 (Ac 3 PIM 2 ). Indeed, the most abundant (M Ϫ H) Ϫ ions at m/z 1414.2 (77%) characterized Ac 3 PIM 2 esterified by 2C 16  The acylation sites on PIM were determined previously to be at both positions of the Gro, the C-3 of the myo-Ins unit and the C-6 of the Manp unit linked to C-2 of the myo-Ins thanks to an NMR strategy based on two-dimensional 1 H-31 P and 1 H-13 C experiments applied on the tetra-acylated form of PIM 2 (Ac 4 PIM 2 ) (12). The NMR strategy we applied on each acyl form is presented briefly below. 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. Then, the remaining myo-Ins and Gro protons were observed on the two-dimensional 1 H-31 P HMQC-HOHAHA spectrum (Fig. 5) and were assigned from their multiplicity and chemical shifts (32) and with the help of the 1 H-1 H HOHAHA spectrum. The different chemical shifts typified the presence or absence of an acyl appendage. For example, a di-acyl-Gro could be evidenced by an H-2 around 5.10 ppm (and H-1 around 4.00/4.25 ppm) (32) and a lyso-Gro should then be assigned by an H-2 around 3.75 ppm and H-1 around 3.95 ppm. These attributions were then supported by the downfield resonance of the lyso-Gro 31 P resonance compared with that of di-acyl-Gro 31 P. Likewise, the downfield shift of the myo-Ins H-3 of about 1 ppm signed a C-3 acyl appendage. The fourth position of acylation, the C-6 of the Manp, was checked with the 1 H-13 C HMBC spectrum, which implied the complete determination of 1 H and 13 C chemical shifts achieved by 1 H-1 H HOHAHA and 1 H-13 C HMQC analysis. When this position is acylated, the H6/H6Ј was deshielded from 3.47/3.64 to 3.95/4.07 ppm.
This analytical approach applied to Ac 3 PIM 2 (Fig. 5c) revealed that the major acyl form corresponded to PIM with two fatty acids on the Gro and one fatty acid on the Manp unit. This was definitively demonstrated from the 1 H-13 C HMQC spectrum, by the chemical shifts of the H-2 (4.98 ppm) and C-2 (70.6 ppm) of Gro and of the H-6/H-6Ј (4.07/3.95 ppm) and C-6 (63.6 ppm) of Manp. Beside this species, another one was observed in less quantity. Indeed, 1 H-13 C HMQC and 1 H-1 H HOHAHA highlighted the presence of an acylated (␦ H-3 4.55) myo-Ins and of a mono-acyl Gro with the acyl appendage present in position 2 as evidenced by the H-2 at ␦ 4.80.
The nature of the fatty acids esterifying the different sites was investigated by mass spectrometry using two different approaches. The first consisted in obtaining fragment ions produced by ESI-MS/MS from each pseudo-molecular ion of the native molecule (Fig. 6a). Concerning Ac 3 PIM 2 , Fig. 6b shows the MS/MS spectrum obtained from the precursor ions at m/z 1414.2 (2C 16 /C 19 PIM 2 ). This spectrum is dominated by peaks at m/z 297.7 and 255.6 assigned to C 19 and C 16 carboxylate fatty acids, respectively. In addition, in the high mass range loss of C 16 gives the fragment ions at m/z 1158.5 which in turn led to the fragment ions m/z 860.0 by loss of C 19 . More interesting are the fragment ions C at m/z 803.9 highlighting a C 16 acyl appendage on the C-6 of the Manp residue. Moreover, this spectrum shows fragment ions at m/z 433.4 corresponding to mono-C 19 Gro and at m/z 153.2 (not shown) corresponding to deacylated Gro. So, ESI-MS/MS fragmentations did not appear suitable in order to establish accurately the structure of the fatty acid located on the glycerol moiety. Moreover, in the case of Ac 4 forms, C ions were unable to discriminate between acyl appendages located on the Manp and the myo-Ins.
A second approach was then developed, consisting in ESI-MS analysis of the acetolysis reaction products of each PIM 2 species. Acetolysis cleaves the phosphoglycerol linkage without altering the fatty acid esters, leading to two entities as follows: the dimannosyl-inositol phosphate moiety (Man 2 -Ins-P) (I, II) and the acyl-glycerol residue (III) (Fig. 7a). The dimannosylinositol phosphate moiety was observed in negative mode as [M Ϫ H] Ϫ ions I and II, and the acyl-glycerol residue was analyzed in positive mode in the presence of ammonium acetate as [M ϩ NH 4 ] ϩ ions III. The acetolysis reaction produces two populations of dimannosyl-inositol phosphate moieties, one with one acetate on the phosphate group (I) and the other without (II). So to simplify the MS spectra, the acetolyzed products were subsequently peracetylated to obtain predominantly the di-ester species (ions I). The number of acetate groups was verified by deutero-acetolysis followed by perdeutero-acetylation. Ac 3 PIM 2 was treated by acetolysis, and the reaction products were analyzed by ESI-MS in positive and negative modes. As expected, the positive ESI-MS spectrum (Fig. 7c) showed an intense peak at m/z 670.9 assigned to ammonium adduct of the di-acylated C 16 /C 19 Gro. The negative mass spectrum (Fig. 7b) mainly showed one peak at m/z 1284.2 corresponding to the Man 2 -Ins-P moiety acylated with one C 18 arising from the C 16 /C 18 /C 19 -PIM 2 . Besides this peak, other peaks in lower intensity were observed at m/z 1312.3 and 1326.4 corresponding to Man 2 -Ins-P moiety acylated with one C 16 and one C 19 , arising from 2C 16 /C 19 -and C 16 /2C 19 -PIM 2 , respectively. Heterogeneity in terms of the nature of the fatty acyl appendage was then observed on the Manp unit, but the major population corresponded to C 16 -Man, as deduced from the intensity of the m/z 1414.2 ions in the ESI source spectrum.
Surprisingly, the parent ion at m/z 1456.2 (C 16 /2C 19 -PIM 2 ) gave rise by MS/MS to m/z 745.7 fragment ion identifying the 2C 19 -Gro and to the corresponding m/z 803.9 fragment ion identifying the C 16 -Man 2 -Ins-P. This is proof that a Gro bearing 2C 19 exists in minor amounts. Moreover, another ion at m/z 1042.0 was observed corresponding to 2C 16  and confirming the presence of the lyso-Gro population evidenced by NMR.
Taken together, these results indicate that Ac 3 PIM 2 predominantly exists with 2 C 16 and 1 C 19 , the glycerol being diacylated by C 16 /C 19 and the mannose bearing a C 16 .
Ac 4 PIM 2 Acyl Forms-The negative ESI mass spectrum of fraction V (Fig. 4d) showed more peaks than that of the fraction IV (Fig. 4c) revealing a higher number of acyl forms. However, all these peaks were in agreement with tetra-acyl forms of PIM 2 (Ac 4 PIM 2 ) and arose from the presence of three types of fatty acids, C 16 19 . Besides these major compounds, another rather abundant species (5%) was observed at m/z 1666.4 corresponding to PIM 2 acylated with 2C 16 and 2C 18 .
The acylation positions were determined previously in an NMR study of Ac 4 PIM 2 (12). Three of the fatty acyl residues were localized on both positions of Gro and on the C-3 of the myo-Ins unit by two-dimensional 1 H-31 P HMQC and HMQC-HOHAHA analysis (Fig. 5d). The localization of the fourth fatty acid, on the C-6 of the Manp, was deduced from the 1 H-13 C HMBC spectrum. As for the Ac 3 PIM 2 , the acetolysis experi- , and for a lesser extent stearate (C 18 ). Species Ͻ2% were not considered. a, fraction I corresponds to C 16 PIM 2 (895.5). b, fraction III corresponds to Ac 2 PIM 2 acylated by C 16

FIG. 5. 1 H-31 P HMQC-HOHAHA spectra of Ac 1 PIM 2 (a), Ac 2 PIM 2 (b), Ac 3 PIM 2 (c), and Ac 4 PIM 2 (d) dissolved in CDCl 3 / CD 3 OD/D 2 O 60:35:8 at 303 (a and b), 305 (c), and 293 K (d).
Numerals correspond to the proton number of the myo-Ins unit, and numerals with letter G correspond to the proton number of the glycerol unit. FIG. 6. Fragmentation scheme (a)   In order to investigate the relative positions of these fatty acids on the Man 2 -Ins-P moiety, peracetylated Ac 4 PIM 2 was analyzed by pyrolysis EI/MS. This method results in the cleavage of the interglycosidic linkage leading to oxonium ions (33). Besides the peracetylated mannose oxonium ions at m/z 331.2, only the oxonium fragment at m/z 527.2 (triacetylated mannose with C 16 ) could be detected but not that at m/z 555.2 (triacetylated mannose with C 18 ) nor that at m/z 569.2 (triacetylated mannose with C 19 ). Taken together, these results indicate that the glycerol is always acylated by C 16 /C 19 and the mannose by a C 16 , whereas the nature of the fatty acid present on the inositol is highly variable, being C 16 , C 18 , or C 19 .
The Minor PIM 2 Acyl Forms (Fractions I-III)-When analyzed in ESI-MS, the fraction I spectrum revealed deprotonated molecular ions at m/z 895.5 (Fig. 4a), whereas fraction II showed deprotonated molecular ions at m/z 937.5. These ions were assigned to mono-acyl forms of PIM 2 (Ac 1 PIM 2 ), containing C 16 and C 19 for fractions I and II, respectively. The 1 H-31 P HMQC-HOHAHA (Fig. 5a) and HOHAHA (not shown) spectra realized on C 19  The ESI mass spectrum corresponding to fraction III showed two peaks at m/z 1133.8 and 1175.9 (Fig. 4b) assigned to deprotonated molecular ions of di-acyl forms of PIM 2 (Ac 2 PIM 2 ). They characterized (M Ϫ H) Ϫ ions of PIM 2 acylated by 2C 16 (m/z 1133.8 (38%)), 1C 16 , and 1C 19 (m/z 1175.9 (62%)). The acylation positions were investigated by NMR analysis. The 1 H-31 P HMQC-HOHAHA spectra (Fig. 5b) clearly showed two lines of correlations, highlighting two distinct populations of Ac 2 PIM 2 . The most representative one (80%, determined from the 31 P signal height) corresponded to a population bearing fatty acids on both positions of Gro since the H-2 was deshielded at 5.06 ppm (Fig. 5b). The Gro proton chemical shifts of the other population were similar to those described for Ac 1 PIM 2 since all the Gro protons resonated between 3.7 and 4.0 ppm. These chemical shifts were in agreement with acylation of Gro at position 1. Moreover, as the myo-Ins proton chemical shifts were also similar to those described in the case of Ac 1 PIM 2 , we assumed that myo-Ins was not acylated and, by deduction, that Manp was acylated in position 6. The existence of these two populations is perhaps more obvious from the 1 H-13 C HMQC analysis, where the two sets of correlations, i.e. those corresponding to the di-acyl Gro and those corresponding to the lyso-Gro, were clearly observed as two different phosphorus resonances (not shown). The two populations of Ac 2 PIM 2 , evidenced by NMR, were confirmed by MS/MS data. Indeed, beside the ions corresponding to a non-acylated Man 2 -Ins-P moiety (m/z 565.4), ions arising from C 16 Ac). c, positive mode in presence of ammonium acetate: 432.5, III as (M ϩ NH 4 ) ϩ with (R 1 , R 2 ) ϭ (C 16 , Ac); 474.6, III as (M ϩ NH 4 ) ϩ with (R 1 , R 2 ) ϭ (C 19 , Ac); 670.9, III as (M ϩ NH 4 ) ϩ with (R 1 , R 2 ) ϭ (C 16 , C 19 ); 1303.1, I as (M ϩ NH 4 ) ϩ with (R 3 , R 4 ) ϭ (C 16 , Ac); 1331.2, I as (M ϩ NH 4 ) ϩ with (R 3 , R 4 ) ϭ (C 18 , Ac). acetolysis experiment evidenced a weak signal at m/z 1284.2 corresponding to C 16 -Man 2 -Ins-P.
The question is now to localize the fatty acids on both positions of the glycerol. Working on phosphatidylinositol, it was noted by Hsu and Turk (34) that loss of fatty acid at the sn-2 position occurs more favorably than that at sn-1 by MS-MS of deprotonated molecular ions. The ESI-MS/MS spectrum of C 16 / C 19 -PIM 2 (at m/z 1175.9) showed fragment ions at m/z 877.4 (-C 19 ) which appeared more intense than the ones at 919.5 (-C 16 ), leading to the conclusion that C 19 is present at the sn-2 position and C 16 at sn-1 ( Table I).
Recruitment of CD1d-restricted NK T Cells in the Granulomatous Lesions Generated by PIM and PI Acyl Forms-As mentioned previously, fractions containing PIM 2 and PIM 6 generated similar lesions with no obvious difference in size revealing that the degree of mannosylation plays no role. To examine the specificity of PIM 2 to recruit NK T cells in the granulomatous lesions, the purified PIM 2 acyl forms described previously were tested. In addition, the cell infiltrates or granuloma were assayed for the presence of CD1d1-restricted T cells identified by the presence of the invariant V␣14-J␣281 TCR ␣ chain. PBS-Alum, used as a negative control and which induced only a minor infiltration (Fig. 8) was found negative for both the infiltration of total V␣14 ϩ T cells (Fig. 8, left panel) and that of V␣14-J281 ϩ (i.e. CD1d1-restricted T cells) (Fig. 8,  right panel). The fully deacylated PIM 2 generated identical signals thus extending to the recruitment of NK T cells, the absolute dependence of the response upon the presence of fatty acid chains. By contrast, PI with two acid chains as well as PIM 2 with one to four acyl chains, in addition to generating a granulomatous response, also induced the local infiltration of the lesions by NK T cells (Fig. 8, right panels). Moreover, the infiltration by NK T cells was nearly exclusive since no other V␣14 ϩ T cells were recovered in the lesions, excluding random infiltration of lymphocytes that should have mimicked the V␣14 profile of total spleen or liver T cells (Fig. 8, bottom  panels). The data obtained by using PCR at saturation were only semi-quantitative but, for an identical HPRT mRNA content, suggested an equivalent number of V␣14-J␣281 mRNA chains in all positive samples. Thus, the number of fatty acid chains does not interfere with the recruitment of NK T cells, and the recruitment of the latter cells by PI of mycobacterial origin strengthened the hypothesis according to which the carbohydrate moiety played only a minor role in the biological response.

DISCUSSION
The results presented herein solve several questions left in abeyance for many years concerning the structural definition of the phosphatidylinositol-mannosides. Although their mannooligosaccharidic part is perfectly defined (5), a blur existed concerning the acylated state of these molecules. This point is plainly of great importance as the metabolic control of the acylation state may be involved in the fine-tuning of their biological activities. For a structural elucidation of the different populations existing, purification is a crucial step. We developed a rapid and effective method allowing the retrieval of purified acyl forms in three steps as follows: (i) lipidic extraction and phospholipidic precipitation, (ii) QMA column to discriminate neutral from charged molecules and PIM 2 from PIM 6 , and (iii) octyl-Sepharose column on PIM 2 or PIM 6 mixtures to separate acyl forms. PIM were predominantly found in the chloroform/methanol extract. Despite our efforts to purify them from the ethanol/water extract, very little was retrieved.
The heterogeneous nature of these closely related molecules defies their purification to homogeneity using isocratic silicic flash chromatography. Ac 4 PIM 2 could be easily purified in this way (12) as it is very apolar, but this strategy was unsuccessfully applied to purify the more polar PIM 2 , as they co-migrate with the more acylated forms of PIM 6 . The strategy described here allowed us first to separate PIM 2 from PIM 6 and second to  fractionate them according to their acylation degree.
M. bovis BCG and M. smegmatis strain 607 3 were found to mainly contain two PIM families, the dimannoside (PIM 2 ) and the hexamannoside (PIM 6 ). PIM 1 , PIM 3 , PIM 4 , and PIM 5 were observed in very small quantities, suggesting that they are biosynthetic intermediates. For each glyco form, tri-and tetraacylated molecules were the most abundant.
The structural strategy used to characterize acylation sites combines the potential of two complementary analytical techniques, NMR and mass spectrometry. NMR has the advantage of taking into account the globality of the fraction but the sensitivity is less than that of mass spectrometry. Electrospray ionization mode was chosen rather than FAB as the molecules could be analyzed without derivatization preventing the loss of any labile substituents. Classical chemical procedures as protection of free hydroxyls by base-stable O-(1-methoxy)-ethylated groups after reaction with methyl vinyl ether, where only acylation sites are revealed as O-methylated sugars by GC/MS analysis (35) or Prehm methylation followed by FAB-MS analysis as described by Khoo et al. (11), could help answering this question. They were discarded as they imply several chemical steps and as they require higher amounts of sample than ESI mass spectrometry. Moreover, in our strategy, NMR was used to characterize the acylation positions, and mass spectrometry gave information concerning the nature of the fatty acids present at each position. The results obtained are summarized in Table I. Ac 1 PIM 2 is identified with C 16 or C 19 in position 1 of the Gro. Ac 2 PIM 2 appears as two populations equally represented, one with 2 C 16 and one with 1 C 16 and 1 C 19 , both fatty acids being on Gro. These data contrast with those of Brennan and co-workers (11) who demonstrated that the Ac 2 PIM 2 exists predominantly as lyso-PIM 2 . Concerning Ac 3 PIM 2 , a major acyl form was observed, corresponding to PIM 2 with 2 C 16 and 1 C 19 , the glycerol being di-acylated by C 16 /C 19 and the mannose bearing a C 16 . Interestingly, the acyl form corresponding to 3C 16 -PIM 2 was not observed, indicating that Ac 3 PIM 2 arose from C 16 /C 19 -PIM 2 . Ac 4 PIM 2 was present as three populations: 3C 16 /C 19 -, 2C 16 /2C 19 -, and 2C 16 /C 18 /C 19 -PIM 2 . The acylation positions were clearly elucidated in an earlier NMR study (12) as being both positions of Gro, the C-3 of the myo-Ins unit and the C-6 of the Manp linked to the C-2 of the myo-Ins unit, excluding the presence of a fatty acid to the C-6 of the Manp linked on the C-6 of the myo-Ins unit envisaged by Khoo et al. (11). Taken together, the results indicate that the glycerol is preferentially acylated by C 16 /C 19 , the mannose by one C 16 , even if a weak variation exits. However, the nature of the fatty acid present on the inositol appears highly variable being C 16 , C 18 , or C 19 .
These results have led to advances in answering the question whether PIM are biosynthetic precursors of LM and LAM or not. The definition of the acyl forms of LAM and LM was approached by an NMR strategy applied on the native molecules (12,36,37). Indeed, the one-dimensional 31 P NMR spectra of LAM and LM from M. bovis BCG typify the anchor heterogeneity and lead to the characterization of at least five populations of differently acylated molecules. Then two-dimensional 1 H-31 P NMR spectroscopy allowed the definition of mono-to at least tri-acylated molecules. In the case of LAM and LM, NMR was unable to confirm the highly suspected presence of fatty acid on the C-6 of the Manp due to the larger number of mannose units. Nevertheless, the purification of each population followed by matrix-assisted laser desorption ionization/ time of flight-mass spectrometric analyses allowed the unambiguous characterization of the tetra-acyl form, in addition to the mono-to tri-acyl forms. 3 Taken together, the results highlight the same major fatty acids and acylation sites for all these molecules strongly suggesting a biosynthetic filiation between them.
PIM, independently of their glyco forms and acyl forms, have the capacity to recruit NK T cells in the granulomatous region. From our results, two main conclusions can be reached concerning the structural requirement of such activity. First, to generate a granulomatous response and to recruit NK T cells, there is an absolute requirement for the molecules to possess at least one fatty acyl chain, with no evidence for differences related to their number, location, or size. Second, the increasing complexity of the carbohydrate moiety is not reflected in significant differences in the biological responses. In this respect, it is worth noting that mycobacterial PI behaved primarily like PIM 2 and PIM 6 , a finding indicative of the minor role of the carbohydrates in NK T cell recruitment as well as a lack of specificity of the TCR of NK T cells toward sugars as suggested independently by analysis of the TCR repertoire (21,38) and the use of CD1d1 tetramers (39).
The V␣14 ϩ invariant TCR ␣ chain typifies CD1d-restricted NK T cells that are activated by exogenous antigens such as ␣-GalCer. This TCR of CD1d-restricted NK T cells is highly discriminating for ␣-GalCer analogs. Indeed, discrete structural modification of the ␣-GalCer carbohydrate part either by adding an extra ␣-Gal or by changing the ␣-anomer into a ␤-anomer or by epimerization of the hydroxyl group at position 2 of ␣-GalCer led to non-activation of CD1d-restricted NK T cells (40 -42). In contrast to these results, our data reveal that PI, which does not contain carbohydrate, presents the same ability as PIM 2 or PIM 6 to recruit V␣14-invariant NK T cells preferentially to other rearranged V␣14 ϩ T cells, which are undetectable in the granuloma. This suggested the absence of a TCR-specific recognition by the TCR of NK T cells and consequently the absence of role of the TCR-CD1 axis in NK T cell recruitment.
In addition, Ac 3 -and Ac 4 PIM 2 show the same activity as Ac 1 PIM 2 and Ac 2 PIM 2 . To date, the glycolipids loaded by CD1b molecules have two aliphatic chains (43). This is also the case for the exogenous antigen ␣-GalCer presented by CD1d. This is consistent with the fact that CD1d is composed of a hydrophobic antigen-presenting groove with two large pockets that probably accommodate two acyl appendages (44). In the case of PIM, the extra acyl appendages located on Manp and/or on myo-Ins could prevent tri-and tetra-acylated PIM recognition by the TCR of NK T cells again supporting the idea that recruitment of V␣14 invariant NK T cells is not controlled by the TCR-CD1d axis. This hypothesis is in agreement with adoptive transfer experiments of wild type NK T cells into CD1d1Ϫ/Ϫ mice showing that CD1d1 expression is not required for the early recruitment of NK T cells to the injection site. 4 Taken together, these data strongly suggest that the TCR-CD1d axis is not important for the NK T cell activation and recruitment to the injection site. These data rather fit with NK T cells being attracted to the site of inflammation caused by glycolipids, irrespective of their carbohydrate and fatty acyl moieties.
It should be kept in mind that the present bioassay does not reflect only early events in the recognition of glycolipids but rather the result of complex and sequential cell-to-cell interactions due to the occurrence of foreign glycolipids. The rationale for the use of the present bioassay is based on the inability to activate NK T cells in vitro using reagents other than ␣-galac-tosylceramide (41) or anti-CD3 monoclonal antibodies (45). This precludes the monitoring of the early release of interferon-␥ and interleukin-4 (46) that could occur following encounters with foreign ligands. The need for such an assay aimed at deciphering early events following glycolipid recognition is stressed by the finding that PIM recruit NK T cells without inducing their apoptosis (data not shown), whereas NK T cells are involved in a variety of immune processes ranging from the keeping of pregnancy to anti-cancer responses (47).