Candida albicans Phospholipomannan, a New Member of the Fungal Mannose Inositol Phosphoceramide Family*

The pathogenic yeast Candida albicanshas the ability to synthesize unique sequences of β-1,2-oligomannosides that act as adhesins, induce cytokine production, and generate protective antibodies. Depending on the growth conditions, β-1,2-oligomannosides are associated with different carrier molecules in the cell wall. Structural evidence has been obtained for the presence of these residues in the polysaccharide moiety of the glycolipid, phospholipomannan (PLM). In this study, the refinement of purification techniques led to large quantities of PLM being extracted from Candida albicans cells. A combination of methanolysis, gas chromatography, mass spectrometry, and nuclear magnetic resonance analyses allowed the complete structure of PLM to be deduced. The lipid moiety was shown to consist of a phytoceramide associating a C18/C20 phytosphingosine and C25, C26, or mainly C24 hydroxy fatty acids. The spacer linking the glycan part was identified as a unique structure: -Man-P-Man-Ins-P-. Therefore, in contrast to the major class of membranous glycosphingolipids represented by mannose diinositol phosphoceramide, which is derived from mannose inositol phosphoceramide by the addition of inositol phosphate, PLM seems to be derived from mannose inositol phosphoceramide by the addition of mannose phosphate. In relation to a previous study of the glycan part of the molecule, the assignment of the second phosphorus position leads to the definition of PLM β-1,2-oligomannosides as unbranched linear structures that may reach up to 19 residues in length. Therefore, PLM appears to be a new type of glycosphingolipid, which is glycosylated extensively through a unique spacer. The conferred hydrophilic properties allow PLM to diffuse into the cell wall in which together with mannan it presents C. albicansβ-1,2-oligomannosides to host cells.

Sphingolipids are ubiquitous and essential components of living cells found mainly on the outer leaflet of plasma membranes (1)(2)(3). Along with their role in membrane permeability and fluidity, they have also been shown to act as second messengers produced in response to various stress situations, which regulate basic processes such as cell cycle control, apoptosis (4), cell-cell interactions (1), and immune response (5).
Sphingolipids are ceramide structures composed of a long chain base whose amino group is amide-linked to various fatty acids. This basic structure as well as those derived by the addition of polar groups such as phosphocholine or carbohydrates to the ceramide may vary according to the species (6 -8). It has been established that sphingolipids from fungal and mammalian cells display important differences in their fine structure and biosynthetic pathways (9). Fungal species incorporate phytosphingosine (PHS) 1 and presumably dihydrosphingosine in the ceramide moieties instead of the sphingosine used mainly by mammals and also have longer fatty acids (C 24 -C 26 instead of C 16 -C 18 ) (9). Fungi then preferentially add inositol to the ceramide group, leading to the family of inositol phosphoceramides composed of inositol phosphoceramide (IPC), mannose inositol phosphoceramide (MIPC), and mannose diinositol phosphoceramide (M(IP) 2 C) instead of phosphocholine in mammalian cells (9). These differences between mammalian cells and fungi represent a target for antifungal drugs such as aureobasidin A (10) and khafrefungin (11), which inhibit IPC synthase. Along with sphingomyelin of mammals and IPCs of fungi and plants, glycosylceramides are another important class of sphingolipids described in fungi and mammals in which the sugar moiety is directly linked to the ceramide (6). The biosynthetic pathway of fungal sphingolipids has only been studied in the yeast Saccharomyces cerevisiae. Numerous steps in the pathway are still unknown (12), and the enzymes involved have not yet been purified or characterized. This lack of knowledge contrasts with the key role of these molecules in cell growth (13) and viability (10), resistance to various environmental stresses (14), remodeling of glycosyl phosphatidylinositol anchors (15), endocytosis (16), and calcium homeostasis (17).
Although Candida albicans is the prominent opportunistic fungal pathogen causing mucocutaneous and systemic infections (18,19), almost nothing is known about its sphingolipids with the exception that a glucosylceramide has been described previously (20) and that IPCs, MIPC, and M(IP) 2 C have recently been characterized in its hyphal form (21). In a series of studies, some immunochemical properties of a C. albicans glycolipid containing long chains of ␤-1,2-linked mannose residues have been demonstrated and characterized previously (22). C. albicans ␤-1,2-oligomannosides, which act as adhesins (23,24), induce cytokine production, and generate protective antibodies (25,26), are considered to be among the virulence factors of this species. Using a model of C. albicans macrophage interaction, it was shown that phospholipomannan (PLM) shed by C. albicans acted as a vector for ␤-1,2-oligomannosides in inducing strong cellular stimulation (27).
The aim of this study was to investigate the complete structure of this pathophysiologically relevant molecule of C. albicans. A purification procedure was developed to obtain large quantities of soluble PLM, and phase gas chromatography (GC) coupled to mass spectrometry (MS) analysis of methanolysis products, extensive NMR, and MS analyses were carried out. A combination of the results of these analyses allowed the complete structure of PLM to be determined and demonstrated that PLM is a new type of mannose inositol phosphoceramide.

EXPERIMENTAL PROCEDURES
Strain and Growth Conditions-C. albicans serotype A, strain VW32, was used throughout this study. For large scale cell production, cells were preincubated on Sabouraud's agar for 18 h at 28 or 37°C and then inoculated at a concentration of 10 6 cells/ml in 1-liter Erlenmeyer flasks containing 500 ml of Sabouraud's broth. Growth was performed to stationary phase at 28 or 37°C on an orbital shaker (180 rpm). Cells were finally collected by centrifugation.
Purification of PLM-Cells were broken with a French press (Aminco) at 20,000 p.s.i, dialyzed, and lyophilized. PLM was then extracted and purified as described previously (22) with the exception that care was taken to avoid micelle formation to improve PLM solubility for NMR studies. Modifications to the method were mainly the lyophilization of most water phases or fractions instead of centrifuge concentration and the reduced use of highly concentrated water phases. The main steps in PLM purification were as follows: (i) extraction with various chloroform/methanol and chloroform/methanol/water mixtures; (ii) extensive butanol/water partitions; and (iii) hydrophobic interaction on phenyl-Sepharose using increasing concentrations of ethanol (1-40%) for elution. Purification and control of PLM fractions were followed by thin-layer chromatography (TLC) (22) and Western blot analysis using the mouse IgM monoclonal antibody, DF9-3 (28).
Monosaccharide and Lipid Composition of PLM-Samples (1-5 g) of purified PLM were lyophilized in conical heavy walled Pyrex tubes (2.0 ml) with Teflon lined screw caps. Methanolysis reagent (0.25-0.5 ml) was then added, and the closed vessels were left for 20 h at 80°C. The methanolysis reagent was obtained by dissolving anhydrous gaseous HCl (up to 0.5 M) at Ϫ50°C in anhydrous methanol previously redistilled on magnesium turnings (29). Gaseous HCl was prepared by the dropwise addition of concentrated sulfuric acid onto crystallized sodium chloride. After methanolysis, samples were evaporated to dryness under a light stream of nitrogen in a ventilated hood followed by the addition of 200 l of acetonitrile and 25 l of heptafluorobutyric anhydride with a pipette with plastic tips. The closed vessels were heated for 15 min at 150°C in a sand bath. After cooling to room temperature, the samples were evaporated in a light stream of nitrogen in a ventilated hood to eliminate excess reagent, and samples were then submitted to a further methanolysis followed by acylation with heptafluorobutyric anhydride as described above. This second methanolysis and acylation step was found necessary for complete cleavage of phosphodiester bonds (and when present, of the linkage between glucosamine and inositol which is observed in glycosylphosphatidylinositols). After drying, the sample was taken up in 200 l of acetonitrile previously dried on calcinated calcium chloride, and an aliquot was introduced into the Ross injector of the GC apparatus.
GC/MS Analysis-For GC/MS analysis, the GC separation was performed on a Carlo Erba GC 8000 gas chromatograph equipped with a 25-m ϫ 0.32-mm CP-Sil5 CB Low Bleed/MS capillary column 0.25-m film phase (Chrompack France, Les Ullis, France). The temperature of the Ross injector was 260°C, and the samples were analyzed using the following temperature program: 90°C for 3 min and then 5°C/min until reaching 240°C. The column was coupled to a Finnigan Automass II mass spectrometer. The analyses were performed in the electron impact mode (ionization energy 70 eV, source temperature 150°C). As described previously (29 -31), this method allowed the identification and the quantification of all constituents of glycolipids in a single GC/MS analysis.
Mass Spectrometry-Electrospray mass measurements were carried out in negative ion mode on a triple quadrupole instrument (Micromass Ltd, Altrincham, United Kingdom) fitted with an atmospheric pressure ionization electrospray source. A mixture of polypropylene glycol was used to calibrate the quadrupole mass spectrometer. The samples were dissolved in dimethyl sulfoxide and further diluted in methanol to obtain a final concentration of ϳ1 g/l. Solutions were infused using a Harvard syringe pump at a flow rate of 3 l/min. The quadrupole was scanned from 600 to 2200 Da with a scan duration of 5 s and a scan delay of 0.1 s. The samples were sprayed using a 3.5-kV needle voltage, and the declustering cone was set at 70 V.
The one-dimensional proton 1 H spectra were measured using a 90°t ipping angle for the pulse and 1.5 s as a recycle delay between each 32 acquisitions of 2.4 s. The spectral width of 4006 Hz was collected in 16,000 complex data points. One-dimensional 13 C was recorded using a spectral width of 20,161 Hz, and 32,768 data points were collected to obtain a field induction decay resolution of 0.6 Hz/point. The 31 P spectra of both compounds were acquired with a spectral width of 16,233 Hz collected in 16,384 data points. Both experiments were carried out with composite pulse decoupling during acquisition using globally optimized alternating phase rectangular pulse sequence at the carbon or phosphorus frequency. An exponential transformation (line broadening factor ϭ 6 Hz for 13 C and 4 Hz for 31 P) was applied prior to processing data points in the frequency domain. Two-dimensional homonuclear ( 1 H-1 H) COSY, one-step relayed COSY, two-step relayed COSY and ROESY, and heteronuclear 1 H-13 C HMQC, 1 H-31 P HMQC, and HMQC-HO-HAHA spectroscopy were carried out using standard Bruker pulse programs (cosy, cosyrl, cosyr2, roesytp, inv4tp, inv4mlevtp, respectively). All of the experiments were carried out in the time-phase proportional increment with the exception of COSYs, which were conducted in a magnitude mode. Finally, two-dimensional HMQC-HOHAHA spectroscopy was recorded using Malcom Lewitt-17 mixing sequences with a mixing time of 120 ms for both 13 C and 31 P nuclei. Two-dimensional ROESY spectra were obtained with mixing times of 400 ms. All of the parameters including special delays were carried out in each experiment.

Analysis of Molar Composition of PLM Revealed That the ␤-Oligomannoside Chain Is Anchored to a Ceramide Moiety-
Our GC/MS analysis of the methanolysis products of PLM allowed us to establish in a single experiment the identification and respective ratio of each PLM component (Table I). It revealed for PLMs synthesized by C. albicans strain VW32 at 28°C (PLM-28) or 37°C (PLM-37) an average composition of 12 or 13 mannose residues for one inositol, one long chain base and one fatty acid. The major long chain bases were identified by their electron impact spectra ( Fig. 1) as the 1,3,4-trihydroxy-2-amino-15-methyl-heptadecane (ramified C 18 phytosphingosine or rC 18 PHS) (Fig. 1c) and 1,3,4-trihydroxy-2-amino-15methyl-nonadecane (rC 20 PHS) (Fig. 1d). Minor components were identified as 1,3,4-trihydroxy-2-amino-octadecane and 1,3,4-trihydroxy-2-amino-eicosane (Fig. 1e), i.e. the corresponding non-ramified PHSs. The major fatty acid (FA) was 2-hydroxy-21-methyl-tricosanoate, a ramified hydroxylated fatty acid with 24 carbon atoms (OH rC 24 FA) (Fig. 1b). Other fatty acids were the homologues with 25 and 26 carbon atoms all hydroxylated on carbon 2 and ramified in the (n-3) position (OH rC 25 FA and OH rC 26 FA, respectively). Linear fatty acids (hexadecanoate and octadecanoate) were also present at a lower level. No trace of monosaccharides other than mannose was detected. In contrast, a small peak with a typical fragmentation pattern revealed the presence of mannose-6-phosphate.
The molar composition of PLM-28 and PLM-37 appeared to be similar as did their OH rC 24 FA ratio ( Table I). The main modifications concerned an inversion of the relative ratios of OH rC 25 FA and OH rC 26 FA and a marked inversion of rC 18 PHS and rC 20 PHS, the latter becoming the most abundant in PLM-37.

MS Analysis of PLM Revealed Crossed Heterogeneity Arising from Its Carbohydrate and Lipid
Moieties-Electrospray mass spectra of PLM-37 ( Fig. 2a) and PLM-28 (Fig. 2b) demonstrated a series of groups of five peaks that covered the entire spectra. Among these groups, those spaced by an average m/z ratio of 81 were observed. These corresponded to (M Ϫ 2H)/2 molecularrelated ions and consequently to the varying degrees of polymerization of the glycan chain of PLM seen in a previous study (22). Conversely, the five peaks within each group arose from heterogeneity of the lipid part of PLM, because they displayed spaced m/z ratios of 7, which therefore corresponded to a mass change of 14 (one CH 2 group). According to the results of a See the details of the fatty acids composition in the column Fatty Acids. b See the details of the long chain base composition in the column Long Chain Base. c r refers to a ramification of the chain in the position (n Ϫ 3).

FIG. 1. GC/MS analysis of methanolysis products from purified PLM-28.
Phospholipomannan was subjected to two consecutive runs of methanolysis and heptafluorobutyric anhydride treatment and analyzed by GC/MS in the electron impact mode of ionization. a, total ion count profile. rC 18 PHS and rC 20 PHS refer to ramified phytosphingosines with a chain length of 18 and 20 carbon atoms, respectively. The major peaks corresponded to the furanic forms produced by dehydration between carbons 1 and 4 during anhydrous acid-catalyzed methanolysis. OH rC 24  GC/MS analysis of methanolysis products and to the relative ratios of their lipid components, these peaks can be attributed to the different compositions of the ceramide moiety as shown in Table II for PLM-37. This interpretation of the results was confirmed when the PLM-28 and PLM-37 spectra were compared (Figs. 2, a and b), because at 28°C, for each degree of  Table II because of smoothing of the spectra. polymerization of the glycan chain, an inversion of the relative importance of the two main peaks of each group was observed, which arose from changes in the ceramide moiety (i.e. rC 18 PHS Ϫ OH rC 24 FA and rC 20 PHS Ϫ OH rC 24 FA). This change must be related to the inversion of the relative molar ratios of the long chain bases as revealed by methanolysis. This observation was also confirmed by inversion of the minor peaks corresponding to rC 20 PHS Ϫ OH rC 25 FA and rC 20 PHS Ϫ OH rC 26 FA in agreement with methanolysis analysis. A second family of groups of five peaks was also observed mainly in the left part of the spectra (Fig. 2, a and b), which most probably corresponded to (M Ϫ 2H ϩ Cl)/3 molecular-related ions. These groups of peaks were spaced by an average m/z ratio of 54 (1 Man/3) and therefore also arose because of the varying degrees of polymerization of the glycan part of PLM (Table II). Peaks within each group were spaced by an average m/z ratio of 4.66, which corresponded to a mass variation of 14 arising from the lipid part (Table II). As confirmed from their respective spacing, groups of peaks from the two families regularly overlapped, leading to common groups of peaks, i.e. groups with a main peak at m/z 1166.19 or 1327.51 on Fig. 2a. This finding was also confirmed by a slight shift of the m/z values (data not shown) and from a modification of the relative ratios of the peaks within these groups explained by the heterogeneous overlapping of the families, because they vary respectively according to M/3 and M/2. Therefore, the two families of groups of peaks observed in the two spectra appeared to originate from the same native molecules. According to these results, PLM may contain 3-20 Man residues/molecule with PLM containing 14 Man residues being the most abundant. In both spectra, the two main peaks at m/z ϳ678 and ϳ1357 appeared unrelated to the PLM families and seemed to correspond to (M Ϫ 2H)/2 and (M Ϫ H) molecularrelated ions from the same molecule, which according to its mass and the results of NMR analysis (see below) is probably M(IP) 2 C. As observed for PLM, M(IP) 2 C is also heterogeneous and seems to express the same variability in its ceramide moiety.
NMR Analysis of PLM Revealed That a Specific Spacer Is Used to Anchor the ␤-1,2-Man Polymer to the PHS of the Ceramide Moiety-Low solubility of glycolipids leads to micelle formation, which affects the relaxation time of molecules in NMR experiments. Therefore, previous NMR analysis of PLM has been restricted to water-soluble glycans released by alkaline treatment (22). A new PLM preparation procedure, avoiding micelle formation, and the use of Me 2 SO-d 6 , known to improve the solubility of glycolipids, have allowed NMR studies of the native molecule. In complementation to previous analysis of the glycan part performed in D 2 O, this study mainly focused on the structure of the spacer that links glycans (␤-1,2oligomannosides) to the lipid moiety and resolves previous ambiguity regarding the position of the second phosphate group of PLM.
NMR analyses have consisted of: (i) a series of COSY experiments to define the 1 H chemical shifts in the sugar, inositol, fatty acid, and phytosphingosine units; (ii) 1 H-13 C and 1 H-31 P HMQC correlation experiments to define the 13 C chemical shifts and the nature of the phosphodiester bridges; and (iii) a ROESY experiment that has established the linkage anomery of the sugar units, their sequence, and their linkage with inositol and phytosphingosine.
The 1 H NMR spectra of PLM-28 and PLM-37 of C. albicans were almost identical, and the anomeric signals of the sugar units were designated A to G according to the decreasing chemical shift of their resonances (Fig. 3, a and b). IN, SP, and FA were also used to indicate myo-inositol, phytosphingosine, and fatty acid signals, respectively, to avoid excessive spectra overload. Because the coupling constants of the anomeric signals could not be measured accurately, the anomeric configuration of Man units A to G was established on the basis of their intraresidual connectivities, H-1/H-2 for ␣-Man and H-1/H-3, H-5 for ␤-Man by ROESY experiments (see below). The anomeric configuration of ␣-Man units A and B was also revealed by their characteristic downfield-shifted 1 H-5 resonance as well as by their upfield-shifted 13 C-5 resonance (see Table III). In this way, sugar units A and B on the one hand and C to G on the other hand were identified as ␣and ␤-Man units, respectively. Moreover, the relatively high coupling (J 3 ϭ 7.2 Hz), which splits the H-1 resonance of ␣-Man unit A, is most probably attributed to the presence of a phosphate group as confirmed by heteronuclear 1 H-31 P NMR spectroscopy.
Because two-dimensional COSY, ROESY, and 1 H-13 C or 1 H-31 P HMQC spectra of both PLMs were essentially identical (data not shown), the assignment of 1 H and 13 C chemical shifts was performed using two-dimensional NMR spectra of PLM-28 or PLM-37.
The starting point for interpretation of the COSY (Fig. 4) and one-step or two-step relayed COSY spectra (data not shown) was the H-1 signal of sugar residues A to G and the H-   . 4) and was attributed to heterogeneity of the material as described below. The two-dimensional 1 H-13 C HMQC spectrum of PLM-28 (Fig. 5a) allowed the assignment of most of the 13 C resonances (Table III) with the exception of those belonging to ␣-Man unit B, which were deduced from a comparison of the 1 H-1 H COSY, 1 H-13 C, and 1 H-31 P HMQC spectra (data not shown). The complete assignment of the 13 C resonances was achieved by interpretation of the 1 H-13 C HMQC-HOHAHA spectrum (Fig. 5b). This experiment provided individual C-2 to C-6 resonances of sugar units G, C, D, A, F, and E, starting from their respective H-2 atom resonance as explained in Fig. 5. The complete series of 1 H and 13 C resonances of the inositol unit was also observed, and the IN C-4 and C-6 resonances in particular were clearly assigned although they were masked on the 1 H-13 C HMQC spectrum by C-3 resonances of sugar units C, D, F, and G. Moreover, the complete series of 1 H resonances belonging to sugar unit B was also observed.
The 1 H-31 P HMQC spectrum of PLM-28 (data not shown) revealed the presence of two 31 P resonances connected to the four protons indicated by arrows in Fig. 6. Two of these protons were unambiguously assigned to A1 and IN1 by comparison   Table III. Labels A to G, IN, FA, and SP refer to signals from Man units inositol, fatty acid, and phytosphingosine, respectively. On the HMQC spectrum (a), the x axis displayed the 1 H shifts of protons as previously determined by COSY experiments. Their projection on the y axis allowed us to define the corresponding 13 C chemical shift. On the HMQC-HOHAHA spectrum (b), the 1 H/ 13 C correlations were read vertically, i.e. from 1 H-2 E (E2) at ␦ ϭ 3.781, the 13 C-6, 13 C-4, 13 C-2, 13 C-3, and 13 C-5 signal resonances (noted E6, E4, E2, E3, and E5) were assigned, respectively. 13  with the COSY spectrum. The assignment of SP1a and SP1b 1 H resonances was also inferred from this comparison and was in agreement with previous NMR data (7).
The 1 H-31 P HMQC-HOHAHA spectrum exhibited a complex panel of correlations for the two 31 P resonances (Fig. 6). By comparison with the COSY spectra, the A2, A3, and B2 signals could easily be connected with the 31 P signal at ␦ ϭ Ϫ0.9. This observation proved the presence of a phosphodiester bridge between the anomeric carbon of sugar unit A and one of the carbons of sugar unit B. From the 1 H-13 C HMQC spectrum (Fig. 5a) and comparison with the 1 H-31 P HMQC spectrum, the 1 H resonance observed at ␦ ϭ 3.88, which correlated with a 13 C signal at ␦ ϭ 65.85, could therefore be defined as the H6a-H6b resonance of sugar unit B. This assertion was confirmed by the observation of B2, B3, B4, B5, and B6 horizontal alignment at ␦ ϭ 71.85 on the 1 H-13 C HMQC-HOHAHA spectrum (Fig. 5b). This observation was consistent with the establishment of the sequence A-P-B, namely Man␣-P-6-Man␣. The second line of 1 H-31 P correlations at ␦ ϭ 3.2 exhibited the complete IN proton spin system as well as the H-1a, H-1b, H-2, and H-3 atom resonances of the phytosphingosine unit (SP1a, SP1b, SP2, and SP3). From this observation, the linking structure IN-1-P-SP was established.
The definitive sequence of PLM was finally achieved through ROESY experiments. The sequences G-F-A and B-IN were suggested from the interresidual correlations between G1-F2, F1-A1, and B1-IN2 (Fig. 7). Because the 1 H-31 P HMQC-HO-HAHA spectra previously analyzed (Fig. 6) have shown that the two phosphate groups were part of the two sequences IN-P-SP and A-P-6-B, the sequence of the linking structure was fully established as ␣-D-Manp-13 P-(O36)-␣-D-Manp-(132)-IN-1-P-(O31)-SP. Concerning the glycan part whose ␤-1,2-oligomannoside structure was established previously (22), a comparison of the proton and carbon chemical shift values (Table III) provided evidence for the non-reducing terminal position of the ␤-Man E unit. However, for Man units C and D involved in the repeated sequence of ␤-1,2-linked mannoses, no more information could be deduced regarding their exact position because of the polydispersity of the molecule.
These results allowed us to define the respective positions of units A to G, IN, SP, and the phosphate groups (Fig. 8a) and to propose a schematic structure of PLM, which specifies its heterogeneity (Fig. 8b) The split of IN2, IN3, IN6, and B1 signals, previously referred to as IN2*, IN3*, IN6*, and B1* and observed mainly in PLM-28 (Figs. 4 and 7) appeared to arise from the same molecule, which may correspond to the (M Ϫ 2H)/2 molecularrelated ion observed on MS spectra at m/z 677.94 (Fig. 2b). Its deduced mass closely fits the following sequences: Man-␣-P-Man-␣-Ins-P-Cer or Ins-P-Man-␣-Ins-P-Cer (M(IP)2C). We suggest that this structure is most probably M(IP) 2 C because: (i) examination of the COSY and HMQC spectra of PLM-28 revealed that ␤-Man unit E was the only one that occurred in the terminal position; (ii) NMR data revealed the absence of a terminal non-reducing ␣-Man unit; and (iii) the HMQC spectrum of PLM-28 clearly showed that the IN* unit was terminal according to its C-2 atom resonance observed at ␦ ϭ 73.5 (Fig. 5). DISCUSSION The poor solubility of glycolipids leading to the formation of micelles is a major limitation to their physicochemical analysis. Therefore, a previous structural study of C. albicans PLM has been restricted to its highly soluble glycan part released by saponification (22). This current NMR and MS study has confirmed the two unusual properties of this molecule that were expected from previous immunochemical analyses, i.e. the ␤-anomery of the linkages and the high degrees of polymerization of the oligomannoside chains (22,28). The presence of two phosphate groups in the molecule was also demonstrated, most probably the first connecting the glycan chain to the lipid moiety and the second presumed to be used to branch a short oligomannoside chain. The presence of this second phosphate group was confirmed by fluorhydric acid hydrolysis that led to a marked modification of the major polymer migration on TLC together with the liberation of a dimannoside or trimannoside (data not shown). The most appropriate method to address the question of the position of the phosphate groups is phosphor/ proton NMR analysis. This required: i) the purification of large quantities of PLM (Ͼ5 mg), ii) the analysis of the whole molecule since saponification was shown to affect the environment of at least one phosphate group and to induce an additive heterogeneity of the mass spectra (22), and iii) high solubility of the whole molecule in a solvent system. These problems were overcome by improving the step-by-step purification procedure and by using as the solvent for NMR studies dimethyl sulfoxide, which allowed a better solubility of PLM than H 2 O. Methanolysis of PLM revealed the presence of rC 18 or rC 20 PHS and rC 24 , rC 25 , or rC 26 hydroxylated fatty acids in addition to mannoses and inositol. These results showed that PLM is related to the family of IPC, MIPC, and M(IP) 2 C previously observed in C. albicans (21). Nevertheless, the electron impact spectra of methanolysis products newly evidenced predominant branched aliphatic chains in both the long chain bases and fatty acids of the ceramide moiety.
As shown previously, the glycan chain was the major source of heterogeneity of the molecule because electrospray mass spectrometry revealed 3-20 mannose units/molecule, the majority of PLM molecules presenting Ͼ11 mannose units. Such heterogeneity could be constitutive or may result from the different steps of biosynthesis. The first hypothesis is more coherent with Western blot analysis of cell wall extracts of C. albicans that are supposed to contain mature PLM, which nevertheless revealed heterogeneity of the molecule. 2 However, mass spectrometry together with analysis of methanolysis products also demonstrated the heterogeneity of the ceramide moiety. This was based mainly on the presence of both rC 18 and rC 20 PHS and concerned all PLM molecules, irrespective of the degree of polymerization of their glycan part. This finding suggests that the constant heterogeneity of the ceramide moiety results from the first steps of synthesis rather than from remodeling of the mature glycolipids as described for S. cerevisiae GPI anchors (15). However, the major difference observed in rC 18 PHS and rC 20 PHS ratios by both methods was related to growth temperature, the rC 20 PHS becoming predominant when temperature shifted from 28 to 37°C. Such an increase in  A to G, IN, SP, and FA. b, general structure of PLM pointing out the heterogeneity of its glycan moiety. The ceramide moiety (Cer) also heterogeneous is mainly composed of rC 18 PHS-OH rC 24 FA for PLM at 28°C or rC 20 PHS-OH rC 24 FA for PLM at 37°C. the C 20 /C 18 PHS ratio has also been reported for S. cerevisiae during heat shock. It was transient for free PHS (15 min) but more stable for PHS found in phytoceramides (at least 1 h) (32) and was proposed to act as a key signaling event in the heat stress response including the induction of trehalose accumulation (32), a cellular thermoprotectant (33). Therefore, it seems that C. albicans and S. cerevisiae respond similarly to temperature by increasing C 20 PHS. However, this increase is continuous, and rC 20 /rC 18 PHS ratios are more important in C. albicans for which the consequences of this shift on both yeast cell metabolism and host-parasite relationships remain to be determined.
Similar modifications linked to growth temperature were also observed in a molecule associated with our PLM preparations with the characteristics of M(IP) 2 C, the main sphingolipid of plasma membranes (3). This was deduced from MS analysis, which displayed the expected masses related to variability of the ceramide moiety similar to that of PLM, whereas NMR studies revealed additional IN2*, IN3*, and IN6* signals arising from unsubstituted inositol. The presence of this structurally related molecule was not surprising because our purification procedure was designed to address to the whole PLM family, which encompasses a number of molecules that diverge according to their amphipathic character. Among these, poorly glycosylated PLMs have physicochemical properties closely related to M(IP) 2 C. We may also expect in the peaks allocated to M(IP) 2 C a minor interference of Man-P-Man-IPCs that display the same masses and correspond to the first biosynthesis step specific for PLM However, because of the selectivity of the NMR signals, the presence of M(IP) 2 C did not change the interpretation of our results and allowed us to determine the general structure of PLM, particularly the position of the two phosphate groups. As suggested previously, the first phosphate group links the inositol to the PHS of the lipid moiety. The second group, formerly thought to be used for branching a short lateral oligomannoside chain, appears to be a component of the spacer that links the ␤-1,2-oligomannoside chain to the lipid anchor. The structure of this spacer (-Man-P-Man-Ins-P-), although closely related to the polar group of M(IP) 2 C, -Ins-P-Man-Ins-P-, has never been described for a glycosphingolipid and appears to be more complex than those of other fungal glycoinositol sphingolipids (8) including those characterized in higher mushrooms (7). Thus, the synthesis of PLM appears to follow the biosynthetic pathway of major membrane sphingolipids from fungi, IPC, MIPC, and M(IP) 2 C from which it should diverge at the MIPC step by further addition of mannosyl phosphate instead of inositol phosphate. In contrast to the addition of inositol phosphate that leads to M(IP) 2 C, an end-biosynthesis product that is not mannosylated further, the addition of mannosyl phosphate is followed by stepwise ␤-1,2-mannosylations, leading to unusual linear chains containing up to 19 Man units. These chain lengths are indeed unique when compared with the glycan part of other fungal glycoinositol phosphosphingolipids that never exceed six sugar units (7,8). They are also longer than those of ␤-1,2-oligomannosides from C. albicans mannan, which never exceed 14 Man units (34). These chains confer on PLM marked hydrophilic properties, which differentiate this molecule from IPC, MIPC, and M(IP) 2 C whose hydrophobic characteristics are coherent with their membrane location. These hydrophilic properties explain why PLM may be released from the cell wall and is ultimately shed from the C. albicans cell surface. In the presence of macrophages, PLM is used by C. albicans as a vector for ␤-1,2-oligomannosides, and shedding of this molecule induces intense signal transduction pathways (27). From a pathophysiological point of view, PLM appears to be a new member of the inositol containing glycolipids described in microbial pathogens, which favor their survival in the host and/or modify its immune response (35)(36)(37)(38).
Elucidation of the complete structure of PLM (Fig. 8b) was a necessary step to determine its biosynthetic pathway and to specify its role in C. albicans, particularly its role in yeast-yeast or yeast-host cells interactions (related to hydrophobicity) (39). This structure also provides the basis for the identification of yeast genes involved in PLM biosynthetic pathways with the ultimate aim of constructing C. albicans PLM-deficient mutants (12). Such studies are currently underway and will help to determine the importance of PLM to the fungal cell and as a virulence factor.
Specific inhibition of mannan ␤-mannosylation related to growth of C. albicans at acidic pH (40) or high temperature (41,42) has previously been demonstrated. However, neither conditions had any affect on PLM ␤-1,2-mannosylation. Interestingly, a comparison of mannan and PLM structures now shows that ␤-mannosylation of both molecules starts from the same -Man␣-P-6-Man␣-group. Whether the nature of the acceptor molecule for ␤-1,2-oligomannosides is regulated by transcription pathways will be an interesting question to address.
In conclusion, this study shows that the prominent opportunistic fungal pathogen C. albicans uses the biosynthetic pathway of membrane sphingolipids to express an unusual linear mannose polymer of ␤-1,2-linked mannose on its surface. As ␤-1,2-oligomannosides are known to have an important role as adhesins, induce cytokine production, and generate protective antibodies, it can be anticipated that this alternative mode of presentation to host cells may have some significance in terms of pathogenesis.