A Novel Lipoarabinomannan from the Equine PathogenRhodococcus equi

Rhodococcus equi is a major cause of foal morbidity and mortality. We have investigated the presence of lipoglycan in this organism as closely related bacteria, notablyMycobacterium tuberculosis, produce lipoarabinomannans (LAM) that may play multiple roles as virulence determinants. The lipoglycan was structurally characterized by gas chromatography-mass spectrometry following permethylation, capillary electrophoresis after chemical degradation, and 1H and 31P and two-dimensional heteronuclear nuclear magnetic resonance studies. Key structural features of the lipoglycan are a linear α-1,6-mannan with side chains containing one 2-linked α-d-Manpresidue. This polysaccharidic backbone is linked to a phosphatidylinositol mannosyl anchor. In contrast to mycobacterial LAM, there are no extensive arabinan domains but single terminal α-d-Araf residue capping the 2-linked α-d-Manp. The lipoglycan binds concanavalin A and mannose-binding protein consistent with the presence of t-α-d-Manp residues. We studied the ability of the lipoglycans to induce cytokines from equine macrophages, in comparison to whole cells of R. equi. These data revealed patterns of cytokine mRNA induction that suggest that the lipoglycan is involved in much of the early macrophage cytokine response to R. equi infection. These studies identify a novel LAM variant that may contribute to the pathogenesis of disease caused by R. equi.

Rhodococcus equi is a significant cause of disease in foals between the age of 1 and 5 months and is responsible for ϳ3% of global foal mortality (1). This organism has also emerged as an opportunistic human pathogen, notably of people with compromised immunity (2). R. equi is an intracellular pathogen of alveolar macrophages, and infection is characterized by bronchopneumonia. The bacterium is known to enter macrophages primarily (but not exclusively) via the complement receptor type 3 (Mac-1) following complement component C3 deposition (2,3). Once within the macrophage the bacteria resist host-killing mechanisms and multiply, eventually killing the macrophage. Specific bacterial factors that facilitate entry of the organisms into the macrophages or that aid intra-macrophage persistence have yet to be identified.
R. equi is a member of the mycolata, a supra-generic taxon including the extensively studied facultative intracellular pathogen Mycobacterium tuberculosis (4). Members of the mycolata have a characteristic cell envelope architecture, dominated by lipids, notably the high molecular weight branchedchain mycolic acids. The cell envelope profoundly affects the properties of these bacteria, and its composition and organization have been a major focus of mycobacterial research (5,6). Lipoarabinomannan (LAM) 1 is a complex mycobacterial cell envelope component that has been identified as a putative virulence factor of M. tuberculosis (7,8). The structure of this macroamphiphile has been studied in detail and consists of a glycosylphosphatidylinositol anchor unit that bears a branched D-mannan and D-arabinan heteropolysaccharide. However, many elaborations of this core structure have been described, including variations in the pattern of the lipid anchor acylation, the presence of succinate residues, and capping motifs on the non-reducing termini of arabinan branches (7, 9 -14). Some of these structural variations, notably the presence of particular capping motifs, may vary between the LAM of different mycobacterial species in a species-specific manner (7). To date LAM have been classified into ManLAM (15) and PILAM (16), according to their small mannooligosaccharide or phosphoinositide cap structures, respectively. The former were found in slow growing mycobacterial species (as Mycobacterium bovis BCG, M. tuberculosis), whereas the latter were found in fast growing mycobacterial species (as Mycobacterium smegmatis).
ManLAM has been shown to have many properties that potentially influence the pathogenicity of M. tuberculosis. In the early stages of infection, ManLAM may facilitate the adherence of bacteria to alveolar macrophages, particularly to mannose receptors (8,(17)(18)(19). It has been demonstrated that mycobacterial internalization via these receptors evades macrophage bactericidal mechanisms (20). LAM has also been reported to have powerful immunomodulatory properties, promoting distinctive patterns of macrophage cytokine induction that subsequently directs host immune responses (8,21). Small differences in LAM structure can strongly influence these biological activities, demonstrating the value of detailed structural studies.
Lipoglycans apparently structurally related to LAM have been identified in representative organisms of other genera within the mycolata, including Corynebacterium matruchotii (22), Dietzia maris (23), Gordonia rubropertincta (24,25), and Rhodococcus rhodnii (26). Although these lipoglycans display components typical of LAM, considerable variation in monosaccharide composition has been found. Thus, further detailed study of the lipoglycan components of these taxa is necessary. By analogy with mycobacterial ManLAM, we hypothesized that a lipoglycan may be involved in the pathogenic success of R. equi. This study describes the isolation, purification, and structural characterization of a R. equi lipoglycan. The potential role of this lipoglycan in R. equi virulence was assessed by comparing the early cytokine responses of equine macrophages to the lipoglycan and to infection with virulent R. equi.

EXPERIMENTAL PROCEDURES
Growth and Maintenance of Organisms-Three strains of R. equi were used in this study as follows: R. equi 103 ϩ (foal isolate) and R. equi 28 ϩ (pig isolate), which are clinical isolates containing plasmids associated with increased virulence (27); and the attenuated type strain R. equi ATCC 6939. Stock cultures were maintained on slopes of brainheart infusion (BHI) agar (Oxoid, Unipath Ltd., Basingstoke, UK) at 4°C. Cultures were maintained by routine subculture onto BHI agar and growth at 37°C for 18 h. Broth cultures were grown in BHI broth incubated at 37°C with shaking (200 rpm). for 18 h. Growth was then harvested, washed twice in PBS, and lyophilized.
Extraction of Lipoglycans-Lyophilized cells were delipidated with chloroform/methanol (1:1 v/v; 50 mg/ml) at ambient temperature for 18 h (28). The cells were then recovered by centrifugation (4000 rpm for 10 min) and washed twice with phosphate-buffered saline (PBS). Cells were then permeabilized with mutanolysin (50 units/ml) and lysozyme (25 mg/ml) according to the method of Assaf and Dick (29). An equal volume of phenol (90% w/v) was then added to the cell suspension in lysozyme buffer, and the mixture was incubated with shaking at 68°C for 1 h to extract lipoglycans, which were subsequently recovered into the aqueous phase formed on refrigerated centrifugation as described previously (30).
Purification of the Lipoglycans-The crude aqueous extract was purified using a modification of the hydrophobic interaction chromatography (HIC) method (22,31). Briefly, the crude extract was taken up into 6 ml of equilibration buffer (100 mM sodium acetate buffer, pH 4.5, containing 15% v/v n-propyl alcohol) and loaded onto a column (1.25 ϫ 17 cm) of octyl-Sepharose CL-4B (Amersham Biosciences). The column was eluted with 40 ml of this buffer prior to gradient elution with a 120-ml gradient of 15-65% v/v n-propyl alcohol in 100 mM sodium acetate buffer, pH 4.5. Fractions were assayed for carbohydrate using the method of Fox and Robyt (32). HIC fractions were also monitored by SDS-PAGE. Samples (30 l) of alternate fractions were prepared and electrophoresed. With consideration of SDS-PAGE analysis, carbohydrate-containing peak fractions were pooled, dialyzed extensively, and lyophilized. To remove a persistent protein contaminant from the lipoglycan, a 1 mg/ml aqueous solution of lipoglycan was treated with an equal volume of 90% phenol (w/v). The mixture was heated at 68°C for 1 h and then separated into distinct aqueous and phenol phases by centrifugation (30 min, 4000 rpm, 4°C). The upper aqueous phase was recovered and re-extracted with phenol. After further centrifugation, the aqueous phase was recovered and extensively dialyzed before being lyophilized and stored at Ϫ20°C. Lipoglycan-containing fractions (PK3) were pooled, dialyzed, and lyophilized. Contaminants were removed by gel filtration. A sample (16 mg) was dissolved in 0.2 M NaCl, 0.25% sodium deoxycholate (w/v), 1 mM EDTA, and 10 mM Tris, pH 8, to a final concentration of 200 mg/ml, incubated 2 days at room temperature, and loaded on a gel permeation Bio-Gel P-100 column (52 ϫ 3 cm) eluted with the same buffer at a flow rate of 5 ml/h. Electrophoresis and Western Blotting Procedures-SDS-PAGE was performed as described by Laemmli (33) using 15% acrylamide resolving gels. The lipoglycan bands were characterized by silver staining with polysaccharide-specific periodic acid oxidation according to Tsai and Frasch (34). Western blotting and lectin blotting using concanavalin A were performed as described previously (23).
Analysis of Lipoglycan Carbohydrate and Fatty Acid Composition-Lipoglycan samples (1 mg) were acid-hydrolyzed with 2 M trifluoroacetic acid (250 l) at 110°C for 2 h in a 8.5 ml of polytetrafluoroethylene screw-capped tube. The hydrolysates were then neutralized in vacuo over sodium hydroxide pellets, and the dried residue was taken up in distilled water (250 l) and then lyophilized. The monosaccharides were derivatized as alditol acetates and analyzed by GLC as described previously (23). Fatty acid composition of the lipoglycan samples (1 mg) was analyzed by GLC of the acid hydrolysate following derivatization to form fatty acid methyl esters as described previously (23). An Immu-nopure® recombinant mannose-binding protein (MBP) column (Pierce and Warriner Ltd., Chester, UK) was used to demonstrate interaction of lipoglycan with MBP as described previously (23).
Permethylation Analysis of Lipoglycan Samples-Lipoglycan (2 mg) from R. equi strains 103 ϩ and 28 ϩ was deacylated according to the method of Beachey et al. (35). The deacylated lipoglycans were permethylated according to the method of Dell et al. (36). The permethylated samples were cleaned using a C 18 Sep-Pak Cartridge (Waters Associates). Each sample was taken up into chloroform/water (1:1 v/v, 1 ml). The cartridge was conditioned by sequential washing with distilled water (5 ml), acetonitrile (5 ml MeCN) and distilled water (10 ml). The sample was then loaded onto the cartridge and distilled water (5 ml) and 2 ml each of 15, 35, 50, and 75% aqueous MeCN and MeCN (2 ml) were used to elute the sample. Fractions collected at each step were assayed for carbohydrate as described previously (23) and by TLC in MeCN/H 2 O (85:15, v/v), visualizing with ␣-naphthol. Fractions that were found to contain carbohydrate were pooled and evaporated under nitrogen.
The permethylated samples were hydrolyzed with 2 M trifluoroacetic acid at 110°C for 2 h as above. The methylated monosaccharides were then reduced with sodium borodeuteride and acetylated according to the method of Saddler et al. (37). Gas chromatography-mass spectrometry analysis of the samples was performed on a Carlo-Erba 8060 MS gas chromatograph connected to a Micromass Trio 2000 mass spectrometer. Samples were injected with a split injector (split rate of 50:1). The injection port temperature was 250°C and the transfer line 250°C. The column was a 30 m ϫ 0.32 mm internal diameter BPX-5 fused silica column with helium (50 kPa) as the carrier gas. The oven was programmed to hold 140°C for 1 min followed by a 10°C/min rise to 280°C and a 10-min hold. The mass spectrometer operated in electron ionization mode (70 eV) and was set to scan from 20 to 650 atomic mass units.
MALDI-TOF Analysis-Analysis by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) was carried out on a Voyager DE-STR (Perspective Biosystems, Framingham, MA) using linear mode detection. All samples were irradiated with UV light (337 nm) from an N 2 laser and were analyzed with the instrument operating at 20 kV in the negative ion mode. The matrix used was 2,5-dihydroxybenzoic acid.
Acetolysis Procedure-10 g of samples were treated with 15 l of acetic anhydride/acetic acid/sulfuric acid (10:10:1, v/v/v) mixture for 3 h at 40°C. The reaction was quenched by addition of 40 l of water. Acetolysis products were extracted twice with 40 l of chloroform and, after drying, were deacetylated with 20 l of a methanol, 20% aqueous ammonia solution (1:1 v/v) at 37°C for 18 h. The reagents were removed under a stream of nitrogen. The samples were then submitted to APTS tagging (see below).
APTS Derivatization-Dried hydrolysis (complete hydrolysis, 2 M trifluoroacetic acid at 110°C for 2 h; mild hydrolysis, 0.1 M HCl at 110°C for 30 min) or acetolysis (38) products were mixed with 0.4 l of 0.2 M 1-aminopyrene-3,6,8-trisulfonate (APTS) (Interchim, Montluçon, France) in 15% (v/v) acetic acid and 0.4 l of a 1 M sodium cyanoborohydride solution dissolved in tetrahydrofuran (39). The reaction was performed for 90 min at 55°C and was quenched by addition of 20 l of water. Dilutions of 1-5 l of the APTS derivatives were prepared in 20 l of total water before analysis by capillary electrophoresis (CE).
Capillary Electrophoresis-The electropherograms were acquired and stored on a Dell XPS P60 computer using the System Gold software package (Beckman Instruments). APTS derivatives were loaded by applying 0.5 pounds/square inch (3.45 kPa) vacuum for 5 s (6.5 nl injected). Separations were performed using an uncoated fused-silica capillary column (Sigma) of 50-m internal diameter with 40-cm effective length (47 cm total length). Analyses were usually performed on a P/ACE capillary electrophoresis system (Beckman Instruments) with the cathode on the injection side and the anode on the detection side (reverse polarity) (Fig. 3, a, c, and d). They were carried out at a temperature of 25°C with an applied voltage of Ϫ20 kV and using 1% acetic acid (w/v), 30 mM triethylamine in water, pH 3.5, as running electrolyte. For Fig. 3b, the electropherogram was recorded in normal mode, at a temperature of 25°C with an applied voltage of ϩ25 kV and borate buffer (20 mM, pH 9.2) as running electrolyte.
Detection system consisted in a Beckman laser-induced fluorescence (LIF) equipped with a 4-milliwatt argon-ion laser with the excitation wavelength of 488 nm and emission wavelength filter of 520 nm.
NMR Spectroscopy-Prior to NMR spectroscopic analysis, fractions were exchanged in D 2 O (99.9% purity, Eurisotop, Saint Aubin, France) at room temperature with intermediate freeze-drying, and then dissolved in 400 l of D 2 O or Me 2 SO-d 6 (99.8% purity, Eurisotop, Saint Aubin, France). ReqLAM from strain 28 ϩ (15 mg) and strain 103 ϩ (3 mg) was analyzed in a 200 ϫ 5-mm 535-PP NMR tubes at 313 K on a Bruker DMX-500 500 MHz NMR spectrometer equipped with a double resonance ( 1 H/X)-BBi z-gradient probe head. Data were processed on a Bruker-X32 work station using the xwinnmr program. Proton and carbon chemical shifts are expressed in ppm downfield from internal acetone (␦ H /TMS 2.225 and ␦ C /TMS 34.00). The one-dimensional phosphorus ( 31 P) spectra were measured at 202.46 MHz and phosphoric acid (85%) was used as the external standard (␦ P 0.0). 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 method (40). Four two-dimensional Homonuclear Hartmann-Hahn (HOHAHA) spectra were recorded using MLEV-17 mixing sequences of 9, 43, 82, and 113 ms (41). The 1 H-13 C and 1 H-31 P single-bond correlation spectra (HMQC) were obtained using Bax's pulse sequence (42). The GARP sequence (43) at the carbon or phosphorus frequency was used as a composite pulse decoupling during acquisition. The pulse sequence used for 1 H-detected heteronuclear relayed spectra (HMQC-HOHAHA) was that of Lerner and Bax (44) and for HMBC was that of Bax and Summers (45).
Macrophage Culture-Peripheral blood macrophages were isolated from a healthy adult horse using Percoll (Sigma) and resuspended at 5 ϫ 10 6 cells/ml in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) with 10% fetal calf serum (FCS) supplemented with penicillin (100 g/ml), streptomycin (80 g/ml), and gentamicin (20 g/ml). 1 ml of the suspension was placed in each well of a 24-well tissue culture plate, with 10 wells used for each treatment. Following overnight (18 h) incubation at 37°C in a humidified atmosphere containing 5% CO 2 , nonadherent cells were removed by washing the plate with warm DMEM/FCS with or without antibiotics for R. equi and lipoglycan treatments, respectively. Following 4 -6 h of culturing, 2.5 ϫ 10 6 cells remained in each well and were used for either bacterial infection or lipoglycan stimulation.
Bacterial Infection and Lipoglycan Stimulation-Initial studies were done to optimize lipoglycan concentration. Infection of horse macrophages with R. equi was carried out as described by Giguère and Prescott (46). A multiplicity ratio of 5 bacteria per macrophage was used. After 40 min of incubation to allow phagocytosis, the macrophages were washed three times and then incubated in DMEM/FCS with antibiotics in order to kill extracellular bacteria and to prevent further extracellular bacterial growth and reinfection of macrophages. For the time course study of lipoglycan stimulation, 10 l of lipoglycan in phosphate-buffered saline, pH 7.2 (PBS), with 200 g/ml polymyxin were added to each well to yield a final lipoglycan concentration of 5 g/ml. Macrophages were harvested for RNA extraction at 0.7, 4, 8, 12, and 24 h following lipoglycan stimulation or exposure to R. equi for infection. Thus, the first time point for the R. equi-infected macrophages represents the time point at which infection was stopped by replacement with DMEM/FCS medium containing antibiotics. Untreated macrophages cultured under the same conditions were used as controls.
Quantitation of Cytokine mRNA Expression by Real Time PCR-Macrophages harvested at the times described above were centrifuged at 400 ϫ g for 5 min and washed with warm PBS, and total RNA was extracted using Qiagen RNeasy Mini Kit (Qiagen Inc., CA). All RNA samples were treated with amplification grade DNase I (Invitrogen) to remove any traces of genomic DNA contamination. 1 g of total RNA was used for cDNA synthesis in a volume of 25 l using the Thermoscript RT-PCR System kit (Invitrogen). After cDNA synthesis by reverse transcription, the reaction was diluted to 80 l. Gene-specific primers and internal oligonucleotide probes for IL-1␤, IL-6, IL-8, IL-10, IL-12p40, IL-18, TNF-␣, and IFN-␥ and glyceraldehyde-3-phosphate dehydrogenase (G3PDH) were selected based on equine cDNA sequences (Table III) using the Primer Express Software (Applied Biosystems, Foster City, CA). The internal probes were labeled at the 5Ј end with the reporter dye 6-carboxyfluorescein, and at the 3Ј end with the quencher dye 6-carboxytetramethylrhodamine. Amplification of 2 l of cDNA was performed in a 25-l PCR containing 900 nM of each primer, 250 nM of Taqman probe, and 12 l of TaqMan Universal PCR Mastermix (Applied Biosystems). Amplification and detection were performed using the ABI Prism 7700 Sequence Detection System (Applied Biosystems) with initial incubation steps at 50°C for 2 min and 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Each sample was assayed in triplicate, and the mean value was used for comparison. Samples without cDNA were included in the amplification reactions to determine background fluorescence and check for contamination. cDNA from 24-h concanavalin A-stimulated equine blood mononuclear cells was used as positive control. To account for variation in the amount and quality of starting material, all the results were normalized to G3PDH expression. Relative quantitation between samples was achieved by the comparative threshold cycles method and is reported as the n-fold difference relative to cytokine mRNA expression in unstimulated macrophages (47).

Purification of the Lipoglycan Fraction of R. equi
In order to minimize contamination with extractable lipids (particularly phosphatidylinositol dimannoside (PIM 2 )) and to maximize yields, R. equi cells were delipidated and permeabilized prior to extraction with hot water/phenol (28,48). The carbohydrate profile for the HIC purification of a crude hot water/phenol extract of R. equi 28 ϩ is shown in Fig. 1a. Similar results, with a reduced proportion of PK2, were obtained for the purification of extracts of R. equi 103 ϩ and ATCC 6939.
Initial elution of the column with equilibration buffer removed hydrophilic contaminant material that includes nucleic acids, polysaccharides, and proteins (PK1, Fig. 1a). Two carbohydrate-containing peaks eluted within the propanol gradient (PK2 and PK3, Fig. 1a). Each fraction was analyzed by SDS-PAGE and modified silver staining which revealed PK3 to contain lipoglycan (ReqLAM). However, the final few fractions were contaminated with a low molecular weight mannose-containing lipid thought to be phosphatidylinositol mannosides (PIM) larger than PIM 2 . Moreover, the fractions that composed PK2 were not revealed using this staining method and were subsequently shown to contain a very high molecular weight polysaccharide, the composition of which varied between strains, and no fatty acids. Consequently PK2 was thought to derive from capsular polysaccharide and as such was not studied further. Lipoglycan-containing fractions (PK3) were pooled, excluding those containing the low molecular weight contami- Subsequently, NMR studies revealed that PK3 was still contaminated by small mannose-containing molecules. The PK3 lipoglycan was further purified by gel filtration in presence of sodium deoxycholate buffer. Gel filtration chromatographic profile shows two peaks, I and II (Fig. 1b). Peak I was tentatively assigned to ReqLAM, based on its electrophoretic mobility on SDS-PAGE (Fig. 2). However, ReqLAM shows an electrophoretic behavior slightly different from those of M. tuberculosis and BCG ManLAM, in agreement with MALDI-TOF MS spectrum showing a broad peak centered at m/z 8000 (data not shown) indicating a molecular mass for the most abundant ReqLAM molecular species of 8 kDa. A molecular mass around 17 kDa was established for the BCG ManLAM (15,38).

Structural Characterization of R. equi LAM
ReqLAM Polysaccharidic Backbone-The ReqLAM (strain 28 ϩ ) was first hydrolyzed (2 M trifluoroacetic acid for 2 h at 110°C), and the resulting monosaccharides were derivatized with APTS and analyzed by CE (12). The electropherograms (Fig. 3, a and b) are dominated by one peak assigned to Man-APTS and a peak of lower intensity attributed to Ara-APTS in both running electrolytes used. From peak integration, the relative composition of the ReqLAM polysaccharide backbone was 86% Manp and 14% Araf. Glycerol and inositol, components of a putative phosphatidylinositol anchor, were also detected by GC analysis. These data suggested that the R. equi lipoglycan may represent an unusual variant on the LAM archetype (ReqLAM).
Western blotting of the lipoglycan preparations with polyclonal anti-LAM antibody (raised against ManLAM from M. tuberculosis, strain H37Rv) demonstrated a weak positive cross-reaction only with material from R. equi 103 ϩ (results not shown). This was further indication that the lipoglycan from R. equi represents a structural variant of the mycobacterial LAM.
In order to investigate the glycosidic linkages present in the ReqLAM, the sample was deacylated, permethylated, hydrolyzed, and derivatized as alditol acetates. The various methylated alditol acetates were routinely identified by gas chromatography-mass spectrometry as terminal  (48). Mild hydrolysis leads to selective cleavage of Araf links and consequently to the release of mannose caps with 1 Ara unit at the reducing end (12,14,38). The electropherogram (Fig. 3c) showed mainly two peaks assigned to Ara-APTS and Man-APTS supporting, as expected, the absence of (t-Manp3Araf) sequence.
We then investigated whether t-Manp and t-Araf cap the 2-O-linked Manp or are directly linked to the C-2 of linear 6-O-linked Manp. ReqLAM was submitted to acetolysis, which allows preferential cleavage of 6-O-linked hexopyranose and the Araf linkages (38), followed by deacetylation, APTS derivatization, and CE analysis. Three peaks of interest (Fig. 3d) were observed assigned to Ara-APTS, Man-APTS, and Manp␣132Man-APTS derivatives in the relative abundance of 20, 45, and 35%. This result indicated the existence of side chains composed of a single Manp capped selectively by t-Araf and not-by t-Manp. This implied that t-Manp is linked directly to the 6-O-linked backbone. Moreover, a percentage of branch-  Purified ReqLAM from strain 28 ϩ was then analyzed by NMR. The one-dimensional 1 H NMR anomeric zone (Fig. 4b) was composed of a multitude of signals, assignment of which required more sophisticated experiments. A complete NMR strategy, involving two-dimensional 1 H-1 H COSY, HOHAHA with different mixing times, ROESY, 1 H-13 C HMQC, HMQC-HOHAHA, and HMBC, was undertaken in order to characterize the different spin systems that compose the ReqLAM (Table  I) and to determine the sequence of these monosaccharidic units. This strategy was realized through NMR analysis of the Ins-P-Gro-Ac 2 , phosphatidyl-myo-inositol anchor, with Ac corresponding to palmitic and tuberculostearic acids (C 16 /C t:19 7/3 for 103 ϩ ReqLAM and 6/2 for 28 ϩ ReqLAM). The diacyl form represents the major acyl form. The linkage of the backbone to the myo-inositol anchor was postulated by analogy to the mycobacterial ManLAM anchor structure. In the same way, ␣-Manp unit was postulated in position 2 of the phosphatidyl-myoinositol anchor based on the characterizations of PIM 2 which are considered as LAM precursors. NMR data indicate two domains, one corresponding to 6-O-linked ␣-Manp and the other to 2,6-O-linked ␣-Manp. From the molecular mass (8 kDa), the percentage of branching (44%) and the fact that one side chain in two was demonstrated capped with t-Araf, m and n could be estimated to 6 and 8, respectively. X corresponds to the terminal unit, which could be assigned to either t-Araf or t-Manp.  (14) in Me 2 SO-d 6 .
The different anomeric protons were characterized by 1 H-1 H HOHAHA (Fig. 5b) and 1 H-13 C HMQC experiments, evidencing 9 spin systems, noted as I to IX in Table I. The anomeric area of the 1 H-13 C HMQC spectrum (Fig. 5c)  The spin system II was unambiguously assigned to t-␣-Araf. The ␣ anomeric configuration was based on the C-1 chemical shifts (␦ C-1 Araf, ␣:109.2/␤:103.1). This spin system could be defined up to H-5/C-5. The different chemical shifts proved that this unit was terminal. The furanose ring was deduced from the C-4 chemical shift (␦ 86.7) and from the intense cross-peak observed on the 1 H-13 C HMBC spectrum between the H-1 (␦ 5.20) and C-4 (␦ 86.7) and between C-1 (␦ 112.3) and H-4 (␦ 4.12) (not shown). The spin systems I, III, and IV typified 2-O-linked ␣-Manp. Indeed, the C-2s were found at 80.4 for system I and at 81.7 ppm for system III and IV on the twodimensional 1 H-13 C HMQC-HOHAHA spectrum (not shown). The pyranose ring of system I was confirmed by the cross-peak observed on the 1 H-13 C HMBC between the H-1 (␦ 5.21) and C-5 (␦ 76.3). Concerning systems III and IV, their 13 C reso-nances were identical, highlighting a glycosylated C-6 (␦ 69.9). They were assigned to 2,6-O-linked ␣-Manp. The chemical shifts of systems III and IV were found very similar to the ones described for the 2,6-O-linked ␣-Manp of the mycobacterial arabinomannan (49). System V had approximately the same anomeric chemical shifts (␦ C-1 at 101.3 and ␦ H-1 at 5.10) as systems III and IV (Table I) and was then attributed to ␣-Manp. However, the C-2 was not found around 82 ppm on the two-dimensional 1 H-13 C HMQC-HOHAHA spectrum but at 73.2 ppm showing that this spin system is not C-2 glycosylated. This chemical shift was similar to the one of the C-2 of system VI (␦ 73.1) which was attributed to terminal ␣-Manp (t-␣-Manp). Spin system V was then tentatively attributed to the 4-O-linked ␣-Manp observed by the permethylation analysis. Then, systems VII, VIII, and IX were found to correspond to 6-O-linked ␣-Manp by analogy to the mycobacterial LAM. 2 Glycosylation in position C-6 was proved by the C-6 chemical shift (␦ 68.7).
The glycosylated carbons were defined from the 13 C chemical shifts. The next step was to determine the sequence of units from 1  In summary, the alditol acetate, permethylation, NMR, and CE data taken together allow us to propose the three following structural features for the polysaccharidic backbone ( Fig. 6) 3) allowed us to deduce that one side chain in two was capped with t-␣-Araf. Moreover, the alditol acetate and permethylation data are consistent with the same structure being present in the ReqLAM from R equi strain 103.
Phosphatidyl-myo-inositol Anchor Acylation State-The phosphatidyl-myo-inositol anchor structure was investigated from one-and two-dimensional phosphorus NMR. The onedimensional 31 P spectrum of ReqLAM exhibited broad unresolved signals in D 2 O (not shown) consistent with multiacylated ReqLAM. Indeed, the one-dimensional 1 H NMR spectrum (Fig. 4a) evidenced the presence of fatty acids by the signals at 0.88 and 1.30 ppm. This was consistent with the presence of fatty acids as analyzed by GLC. The predominant fatty acids found were hexadecanoic acid (56%) and 10-methyloctadecanoic acid (tuberculostearic acid) (19%), whereas heneicosanoate, octadecenoic acid, heptadecanoic acid (C17), and a C16-2 M. Gilleron, T. Brando, and G. Puzo, unpublished results.

TGTCGCAGGAGCCACCACGCT
branched fatty acid (possibly 10-methylhexadecanoic acid) are present in smaller amounts (11,6,4, and 5%, respectively). The fatty acid composition of the lipoglycan reflected that of the whole bacterial cells (data not shown) although a reduction in the relative proportion of unsaturated fatty acids was noted, as has been observed previously (22,25,26) for other LAM-like lipoglycans of the mycolata. As ReqLAM exhibited broad unresolved signals in D 2 O, no connectivities between phosphate and protons could be obtained by two-dimensional 1 H-31 P HMQC and HMQC-HO-HAHA. Recently, it was shown that Me 2 SO-d 6 is a suitable solvent for recording high resolution one-dimensional 31 P NMR spectra of multiacylated mycobacterial ManLAM (13). Onedimensional 31 P spectrum of ReqLAM dissolved in Me 2 SO-d 6 mainly showed three signals at ␦ 1.83, ␦ 3.50, and ␦ 4.05. From their chemical shifts and by comparison with those of the 31 P signals observed in the one-dimensional 31 P spectrum of the M. bovis BCG cellular ManLAM (13) (Fig. 7), two of these signals were tentatively assigned to P3 (␦ 1.83) and P5 (␦ 3.50).
The 31 P resonance assignments were confirmed with help of two-dimensional 1 H-31 P NMR spectroscopy. The 1 H-31 P HMQC-HOHAHA spectrum of ReqLAM (not shown) exhibited a complex panel of correlations. P3 showed correlations with downfield resonances at ␦ 5.13 in F 2 dimension, which were assigned to methine protons H-2 of di-acylated glycerol according to previous results (11,13,14). A similar downfield correlation was not observed for P5; instead the Gro H-2 resonance is superimposed with the Gro H-3/H-3Ј, indicating lack of acylation of O-2 in this minor species. So, only P3 corresponds to 1,2-di-acyl-3-phospho-sn-glycerol unit, and P5 corresponds to 1-acyl-2-lyso-3-phospho-sn-glycerol unit. The inspection of the different cross-peaks (chemical shifts, multiplicity of the signals, and 3 J HH coupling constants), by comparison with the 1 H-31 P HMQC-HOHAHA of M. bovis BCG cellular ManLAM (11,13,14), allowed their attribution to the different protons of glycerol and myo-inositol spin systems (Table II). From these data, it can be proposed that P3 typified a di-acylated Gro anchor. Although the P5 signal was weak, we were able to deduce the absence of an acyl residue on C-2 of the Gro. P3 and P5 corresponded to phosphatidyl-myo-inositol anchor devoid of fatty acid on their myo-Ins unit.
The linkage of the backbone to the myo-inositol anchor was postulated by analogy to the mycobacterial ManLAM anchor structure. In the same way, ␣-Manp unit was postulated in position 2 of the phosphatidyl-myo-inositol anchor based on the characterization of PIM 2 by MALDI (Fig. 8) as PIM 2 are considered as LAM precursors.

Biological Activities of the Lipoglycan
ReqLAM showed a strong reaction on Western blotting with Hypermune-RE, an anti-R. equi antiserum produced commercially for the prophylactic treatment of foals (data not shown), confirming the antigenicity of the lipoglycan in vivo. Moreover, ReqLAM gave reactions of varying intensity with four out of four convalescent sera from foals that had recovered from R. equi infection (data not shown). Mycobacterial LAM, LM, and PIM have been shown to interact with MBP (50,51). ReqLAM was applied to a column of immobilized MBP, and retention of lipoglycan material was monitored by electrophoresis followed by blotting and probing with concanavalin A (23). ReqLAM was retained by the MBP (Fig. 9), eluting only after the application of buffer containing EDTA to disrupt calcium-dependent interactions between the lipoglycan and the MBP carbohydrate recognition domain.
Further study of the biological activities of the ReqLAM focused on its ability to stimulate cytokine production by equine peripheral blood macrophages. Cytokines selected for study were representative of the major cytokines produced by macrophages and indicative of the development of different T cell responses. In preliminary experiments, various concentrations (1, 5, 10, or 20 g/ml) of ReqLAM were used to stimulate equine macrophages, and TNF-␣ and IL-12 p40 mRNA expression was quantitated. For both cytokines, ReqLAM at 5 g/ml induced the highest level of mRNA expression (data not shown), and therefore this concentration was used in subsequent experiments. Treatment of 5 g/ml ReqLAM with 200 and 1,000 g/ml polymyxin showed no inhibitory effect at either concentration on TNF-␣ and IL-12 p40 mRNA cytokine production over that caused by polymyxin alone (data not shown). This confirmed that ReqLAM was not contaminated by endotoxin.
Subsequently we examined cytokine induction in horse macrophages by virulent R. equi or by ReqLAM (5 g/ml), at various time points. Quantitative expression of cytokine mRNA at different times following infection or lipoglycan stimulation are shown in Figs. 10 and 11. For all the representative inflammatory cytokines assayed, with the exception of TNF-␣, ReqLAM induced a greater early response than the whole cells ( Fig. 10) but that at later time points the overall profile of cytokine induction was similar for both R. equi and ReqLAM, albeit with typically higher response to the R. equi whole cells.
Likewise, for the representative regulatory cytokines assayed, ReqLAM again induced a greater early response compared with the whole cells (Fig. 11). At later time points in the assay of the regulatory cytokines, the whole cells typically induced a greater response than the ReqLAM, with the exception of IL-18 that was not induced by either stimulant (Fig. 11B). The ability of ReqLAM to induce production of both inflammatory and regulatory cytokines was also confirmed in preliminary experiments with ReqLAM from strain 103ϩ (data not shown). DISCUSSION The R. equi lipoglycan ReqLAM has a structure related to, but clearly distinct from, that of mycobacterial ManLAM from BCG and M. tuberculosis (7). ReqLAM is smaller than Man-LAM; MALDI-TOF mass spectrometry revealed the molecular mass of ReqLAM to be centered around 8 kDa. By using a similar method, Venisse et al. (15) measured the average size of ManLAM of M. bovis BCG as 17.4 kDa and that of LM as 6 kDa. 2 The broad diffuse band observed by SDS-PAGE and the spread of molecular mass revealed by mass spectrometry indicate that like ManLAM, ReqLAM is heterogeneous in size.
The small size of ReqLAM as compared with ManLAM is consistent with the reduced arabinose content, ϳ11% of the carbohydrate moiety compared with 55% for BCG ManLAM (15). The ␣1-6 mannan backbones of ManLAM of M. tubercu- losis Erdman and M. bovis BCG were demonstrated to be highly branched although the degree of branching and the size were heterogeneous (52,53). Results of permethylation and CE analysis indicate that ReqLAM has a similar ␣1-6 mannan backbone, with ϳ44% of branching. Acetolysis and study of derivatized products by CE revealed the branches to contain one 2-O-linked mannose residue either as t-␣-Manp or capped by t-␣-Araf. The majority of the arabinose residues was detected by permethylation analysis as t-␣ Araf in agreement with the absence of mannose caps that typify the BCG and M. tuberculosis ManLAM (12). A structural model of ReqLAM can be proposed here, consisting of a similar mannan backbone to that of ManLAM but in which single arabinose residues decorate the mannan backbone instead of elaborate arabinan branches (Fig. 6). In agreement with its molecular mass of 8 instead of 17 kDa for the ManLAM, the arabinan domain of the ReqLAM is restricted to single ␣-Araf capping the 2-O-linked ␣-D-Manp. This structural feature reveals that the biosynthesis of an extensive arabinan domain comparable with that of mycobacterial ManLAM (by addition of 5-O-linked ␣-Araf) does not occur in R. equi, probably due to the absence of a 5-arabinosyltransferase. The reduced arabinose content of the ReqLAM may explain the relatively weak or negative crossreactions of the lipoglycan with a polyclonal anti-LAM antiserum as the arabinose residues appear to be the principal antigenic determinants in ManLAM (7,54). However the ReqLAM does elicit antibody production in foals and adult horses, as judged by its reaction with hyperimmune and convalescent antisera.
The presence of terminal mannose units within the ReqLAM structure (Fig. 7) may have considerable relevance with respect to the biological significance of this lipoglycan, notably through interaction with components of the innate immune response. The interaction of ReqLAM with recombinant MBP (Fig. 9) is consistent with the previously observed interaction of MBP with a variety of mannoconjugates including mycobacterial ManLAM, LM, and PIM (51) and the lipomannan of Micrococcus luteus (55). Binding of MBP by ReqLAM in vivo may activate complement C3b deposition onto R. equi via the lectin pathway (56) thereby helping promote the previously described (3) complement-receptor Mac-1-mediated uptake of R. equi into macrophages. ReqLAM may be able to bind other collectins as both LM and ManLAM have been identified as ligands for human pulmonary surfactant protein A (57), and human pulmonary surfactant protein D binds ManLAM (58), although only binding of the former surfactant promotes binding of M. tuberculosis to phagocytes. Equine pulmonary surfactant protein A binds mannose (59), and equine pulmonary surfactant protein D has been shown to bind both mannose and phosphatidylinositol (60). Thus ReqLAM may bind either or both of these surfactants in the foal lung. LAM may also influence mycobacterial uptake into host cells by interacting with other receptors including macrophage mannose receptors (8,(17)(18)(19), and similar interactions may be possible for ReqLAM. Entry via some pathways, notably macrophage mannose receptors (20), may influence intracellular survival by circumventing the activation of macrophage antimicrobial responses. Consequently, the multifaceted potential of ReqLAM for promoting bacterial entry into host cells warrants further investigation.
We show here that the pro-inflammatory and immune cytokine response of equine macrophages to ReqLAM largely parallels the response to infection with live virulent R. equi. Activation of resident macrophages is one of the earliest responses to microbial invasion, with macrophage-derived cytokines playing a critical role in initiating the inflammatory response as well as in regulating the immune response. The results shown here suggest that much of the early macrophage cytokine response occurring after infection with R. equi can be attributed in part to its ReqLAM component. The effect of ReqLAM observed herein was not the result of contamination with lipopolysaccharide, because the effect could not be inhibited by use of polymyxin, which inactivates lipopolysaccharide. These findings extend and complement the results of earlier studies with infected murine macrophages, which failed to demonstrate any cytokine response that could be specifically attributed to possession of the virulence plasmid (46).
The ManLAM of M. tuberculosis has been shown to produce a wide spectrum of immunomodulatory functions in vitro, but the biological implications of these effects are still largely undefined. These effects include suppression of T-cell proliferation, inhibition of interferon (IFN)-␥-induced functions including microbicidal activity, scavenging of cytotoxic free oxygen radicals, and complement activation (7,8,61). However, only LAM lacking mannose caps (PILAM and AraLAM) has been found to induce significant TNF-␣ and IL-12 expression by human and murine macrophages (7,8). Our studies with ReqLAM also show the marked effect of this lipoglycan on macrophage cytokine induction. It is intriguing that ReqLAM induces both pro-inflammatory cytokines (Fig. 10) and macrophage deactivating cytokines (Fig. 11). The lack of immediate TNF-␣ response to ReqLAM was a notable difference from the immediate TNF-␣ response to whole cells of R equi (Fig. 10D). Darrah and co-workers (62) have shown that both TNF-␣ and IFN-␥ are required to activate macrophages in order to kill R equi by peroxynitrite. Finally, the observation of a marked optimal dose effect of ReqLAM on selected cytokine mRNA transcription (data not shown) was unexpected. Further study is needed to support the intriguing possibility that ReqLAM could produce an immunosuppressive effect through an intracellular bacterial load effect on macrophage cytokine expression. Possible effects attributable to ReqLAM thus need to be considered in future studies of the cytokine responses of equine macrophages to infection with virulent and avirulent R. equi.