Analysis of glycosylinositol phosphorylceramides expressed by the opportunistic mycopathogen Aspergillus fumigatus.

Acidic glycosphingolipid components were extracted from the opportunistic mycopathogen Aspergillus fumigatus and identified as inositol phosphorylceramide and glycosylinositol phosphorylceramides (GIPCs). Using nuclear magnetic resonance sppectroscopy, mass spectrometry, and other techniques, the structures of six major components were elucidated as Ins-P-Cer (Af-0), Manp(α1→3)Manp(α1→2)Ins-P-Cer (Af-2), Manp(α1→2)Manp(α1→3)Manp(α1→2)Ins-P-Cer (Af-3a), Manp(α1→3)[Galf(β1→6)]Manp(α1→2)-Ins-P-Cer (Af-3b), Manp(α1→2)-Manp(α1→3)[Galf(β1→6)]Manp(α1→2)Ins-P-Cer (Af-4), and Manp(α1→3)Manp(α1→6)GlcpN(α1→2)Ins-P-Cer (Af-3c) (where Ins = myo-inositol and P = phosphodiester). A minor A. fumigatus GIPC was also identified as the N-acetylated version of Af-3c (Af-3c*), which suggests that formation of the GlcNα1→2Ins linkage may proceed by a two-step process, similar to the GlcNα1→6Ins linkage in glycosylphosphatidylinositol (GPI) anchors (transfer of GlcNAc, followed by enzymatic de-N-acetylation). The glycosylinositol of Af-3b, which bears a distinctive branching Galf(β1→6) residue, is identical to that of a GIPC isolated previously from the dimorphic mycopathogen Paracoccidioides brasiliensis (designated Pb-3), but components Af-3a and Af-4 have novel structures. Overlay immunostaining of A. fumigatus GIPCs separated on thin-layer chromatograms was used to assess their reactivity against sera from a patient with aspergillosis and against a murine monoclonal antibody (MEST-1) shown previously to react with the Galf(β1→6) residue in Pb-3. These results are discussed in relation to pathogenicity and potential approaches to the immunodiagnosis of A. fumigatus.

Of .185 fungal species to which humans are exposed, aspergilli are perhaps the most common; conidia (spores) from a variety of Aspergillus species, such as A. versicolor, A. niger, A. terreus, A. flavus, and A. fumigatus, can be found at high concentrations in both rural and urban environments, outdoors as well as within human habitations (1). Of these species, A. fumigatus is considered the most pathogenic and is now recognized as the most prevalent airborne fungal pathogen in developed countries (2,3), even in hospital environments, where it may contribute only a small fraction of the aerial spore count (4). Along with the general increase in the number of immunocompromised patients as a result of the use of corticoids (5), organ transplants, human immunodeficiency virus infection, and cancer (6), the high thermotolerance of A. fumigatus, which is able to grow in temperatures varying from ,20jC to at least 50jC, as well as its high sporulating capacity and the small size of its spores (2-3 mm) (1), have contributed to the prevalence of A. fumigatus in invasive and noninvasive human aspergillosis worldwide.
Currently, the most commonly used drugs for the treatment of aspergillosis are amphotericin B, which binds membrane sterol and creates transmembrane channels leading to an increase in membrane permeability to monovalent cations (7), and itraconazole, which competes for oxygen at the catalytic heme site of cytochrome P450/ lanosterol 14a-demethylase, leading to an inhibition of ergosterol biosynthesis and the accumulation of 14amethylated sterols in the fungal plasma membrane (8). However, these two drugs have major disadvantages, including the nephrotoxicity of amphotericin B and the lack of itraconazole preparations for intravenous application, limiting its use in patients with impaired bowel absorption. In addition, there are increasing reports documenting the development of resistance to these antifungal drugs among a variety of species, including Aspergillus (8). Thus, there exists a compelling interest in the discovery of new targets for the development of antifungal therapeutics: novel functional molecules and the pathways by which they are generated and interact with other components in processes essential for fungal reproduction and growth or host colonization.
Studies carried out during the 1990s demonstrated that many species of fungi are vulnerable to inhibitors of sphingolipid biosynthesis (reviewed in Ref. 9). This has stimulated increased interest in the structure, biosynthesis, and functional roles of fungal sphingolipids as potential targets for antifungal agents. A particularly interesting target is the fungal inositol phosphorylceramide (IPC) synthase (10), inhibitors of which are highly toxic to many mycopathogens but exhibit low toxicity in mammals (11)(12)(13). IPC synthase, generally believed to be encoded by orthologous AUR1/IPC1 genes discovered in Saccharomyces cerevisiae, Schizosaccharomyces pombe, Candida species, Aspergillus species, and Cryptococcus neoformans (10,(14)(15)(16)(17)(18), catalyzes the transfer of myo-inositol-1-phosphate from phosphatidylinositol to ceramide. Neither IPC, IPC synthase, nor any orthologous AUR1/IPC1 genes have been found in mammals, consistent with the lack of susceptibility of mammalian species to inhibitors of this enzyme.
Further processing of IPC by glycosyltransferases yields glycosylinositol phosphorylceramides (GIPCs), a class of complex anionic glycosphingolipids widely distributed among fungal species (19). Isolation and detailed characterization of GIPCs from a variety of fungi has begun to reveal an extensive structural diversity (see Discussion), the expression of which is considerably dependent on species and, apparently, at least in some mycopathogens, strongly regulated during morphogenesis (20)(21)(22)(23)(24). A total of ?30 distinct GIPC structures have been identified to date, with the discovery rate increasing during the last 5-6 years, based on further elaboration of three wellconfirmed core structures distinguishable at the monoglycosyl level (Scheme 1).
Antigenic glycoside determinants are expressed on some fungal GIPC components, suggesting their potential as targets for immunodiagnostic reagents and the possibility of therapy based on stimulation of the mammalian humoral response (20,21,(25)(26)(27)(28)(29). Elucidation of such im-munological interactions calls for further detailed knowledge of the structures of these compounds, which is still limited compared with what is known about glycosphingolipids of animal species.

Fungal isolates and growth conditions
A. fumigatus American Type Culture Collection strain 9197 was used for most preparations. Also used was A. fumigatus strain 237, originally cultured from open lung biopsy from a patient with invasive aspergillosis at Hope Hospital in Manchester, UK (a gift of Dr. David Holden, Hammersmith Hospital, London, UK). Approximately 5 3 10 9 spores were inoculated to 200 ml or 1 liter of complete medium (1% glucose, 0.2% peptone, 0.1% yeast extract, 0.1% casamino acids, nitrate salts, trace elements, and 0.01% vitamins, pH 6.5; trace elements, vitamins, nitrate salts, and amino acid supplements are described in the appendix to Ref. 32), incubated at 30jC or 37jC with shaking (200 rpm) for 24 h, filtered, and processed as described. Yields were 6-8 g wet weight per 200 ml of medium or 25-35 g wet weight per 1 liter of medium.

HPTLC immunostaining
Immunostaining of purified fungal GIPCs separated by HPTLC was performed by the procedure of Magnani, Smith, and Ginsburg (33), as modified by Kannagi et al. (34), with some additional changes to the HPTLC protocol. Briefly, separation by HPTLC was carried out in three steps. After application of GIPCs, the HPTLC plates were developed twice with chloroform-methanol (4:1, v/v; solvent E) and once with chloroform-methanol-water [30:18:4 (v/v/v), containing 0.020% (w/v) CaCl 2 ; solvent F). After HPTLC development, plates were dried, soaked in 0.5% polymethacrylate in hexane, dried again, and blocked for 2 h with 1% BSA in 0.01 M PBS, pH 7.2. Plates were then incubated with monoclonal antibody MEST-1 or aspergillosis serum overnight; this was followed by sequential incubation with either rabbit antimouse IgG or rabbit anti-human IgG, respectively, and 125 I-labeled protein A (4 3 10 5 cpm/ml) (35).

Extraction and purification of glycosphingolipids
Extraction and purification of glycosphingolipids were carried out as described previously (27,36,37), but with some modifications and additional steps introduced for the purpose of reducing the amounts of irrelevant substances to be dealt with earlier in the protocol. Briefly, glycosphingolipids were extracted by homogenizing mycelia (25-35 g wet weight) in an Omni-mixer (Sorvall, Inc., Wilmington, DE) three times with 200 ml of solvent A and three times with 200 ml of solvent B. The six extracts were pooled, dried on a rotary evaporator, dialyzed against water, and lyophilized. The dried residue was partitioned between water and 1-butanol presaturated with water (200 ml each) with vigorous shaking in a separatory funnel. The lower (water) layer was removed and similarly extracted three more times with equal volumes of water-saturated 1-butanol. The four 1-butanol extracts were combined in a round-bottom flask, evaporated to dryness, resuspended in a minimal volume of solvent C, and applied to a column of DEAE-Sephadex A-25 (Ac 2 form). Neutral glycosphingolipids were eluted with 5 volumes of solvent C. Acidic glycosphingolipids were eluted with 5 volumes of 0.5 M sodium acetate in methanol. The acidic fraction was dried, dialyzed exhaustively against deionized water, redried, and treated with 20 ml of methanol-water-1-butanol (4:3:1, v/v/v) containing 25-30% methylamine at 55jC for 4 h (flask tightly stoppered), with occasional agitation. After removal of the reagent solution by rotary evaporation, the acidic lipids were further fractionated by repetitive preparative-scale HPLC, using columns of either 60 cm 3 4.6 mm Iatrobeads (Iatron Chemical Co., Tokyo, Japan) 6RS-8010 or 50 cm 3 4.6 mm Sphereclone (Phenomenex, Torrance, CA) 10 mm porous spherical silica. The mobile phase was a 2-propanolhexane-water gradient programmed from 55:40:5 to 55:25:20 over 120 min, followed by isocratic elution for 40 min, with a flow rate of 0.5 ml/min. Generally, 40 3 2 ml fractions were collected for first-stage purifications and 80 3 1 ml fractions were collected for second-stage purifications, where required. The identity and purity of each fraction were assessed by analytical HPTLC as described above, and by one-dimensional (1-D) 1 H-NMR spectroscopy, before further characterization by the full range of NMR and MS techniques described below.
hanced heteronuclear single-quantum correlation (gHSQC) experiments were performed at 35jC on Varian (Palo Alto, CA) Unity Inova 500 MHz (Department of Chemistry, University of New Hampshire) and 600 and 800 MHz (Complex Carbohydrate Research Center, University of Georgia) spectrometers using standard acquisition software available in the Varian VNMR software package. Proton chemical shifts are referenced to internal tetramethylsilane (y 5 0.000 ppm). Carbon chemical shifts are referenced to solvent DMSO (y 5 40.0 ppm); in some cases, these were not obtained from directly detected 1-D spectra but measured in F1 of the gHSQC experiment; therefore, the values are given to single decimal places only.

Positive ion mode electrospray ionization mass spectrometry
Mass spectrometry of the IPC fraction (Af-0) was performed via electrospray ionization (ESI) in the positive ion mode ( 1 ESI) on a linear ion trap instrument (LTQ; Thermo-Finnigan, San Jose, CA). Mass spectrometry of all other fractions containing GIPCs was performed in the positive ion mode on a Micromass Global (Manchester, UK) hybrid ESI-Q /q-oa-TOF-MS instrument (Q-TOF). All (G)IPC samples were introduced into the ion source via direct infusion in 100% methanol (?100 ng/ml). The flow rate was usually 0.5 ml/min, but for analysis of samples available in limited amounts, a nanospray capillary tip was used, from which the flow rate was estimated to be ?200 nl/min. To generate [M(Li)1Li] 1 adducts of (G)IPC molecular species, LiI (10 mM) in methanol was added to the analyte solution until the observed ratio of [M(Li)1Li] 1 adducts to mixed Na 1 /Li 1 adducts in MS profile mode was .5:1; the necessary LiI concentration was generally in the range 3-5 mM (39,40). In general, spectra represent summations of 50-350 scans ( experiments. Additional matrix-assisted laser desorption timeof-flight (MALDI-TOF) mass spectrometry was performed on an Axima CFR (Shimadzu Biotech, Columbia, MD) instrument. Lithiation was accomplished by doping 2,5-dihydroxybenzoate matrix with LiI (41).

Fragment ion naming conventions and interpretation
Nominal, monoisotopic m/z values are used in the labeling and description of 1 ESI-MS results. Fragment naming conventions and interpretation of spectra derived from [M(Li)1Li] 1 adducts of GIPC molecular species are based on those of Adams and Ann (42) and Singh, Costello, and Beach (43), as described previously (39)(40)(41); additional naming conventions for ceramide fragments are derived from Hsu and Turk (44). Essential characteristics of these spectra are summarized below with reference to Scheme 2 (see also Schemes 3-7 below accompanying the mass spectra). In general, 1 ESI-MS/CID-MS spectra of GIPCs acquired [via 1 ESI-Q/q(CID)-oa-TOF-MS or other tandem MS/CID-MS configurations (30,(39)(40)(41)  Monosaccharide, inositol, fatty acid, sphingoid, and linkage analysis by GC-MS Monosaccharides were analyzed as their per-O -trimethylsilyl methyl glycosides. A re-N-acetylation step was used after methanolytic depolymerization to ensure that hexosamines, if present, would be detected as their acetamido derivatives. Ceramidederived sphingoid bases and 2-hydroxy fatty acids were detected as their N-acetyl-per-O -trimethylsilyl and 2-O -trimethylsilyl methyl ester derivatives, respectively. Myo -inositol (Ins) was detected in the monosaccharide analysis as its per-O -trimethylsilyl derivative, although optimal conditions for its release (20) were not used. All derivatives were prepared according to protocols described previously (30). Linkage analysis was performed by the method of partially methylated alditol acetates. Permethylation was performed by microscale adaptation of the method of Ciucanu and Kerek (45); hydrolysis, reduction, and acetylation to produce partially methylated alditol acetates were performed as described by Levery and Hakomori (46). Instruments used for GC-MS were a GCQ (Thermo-Finnigan) and a GC3800-MS1200 (Varian), operated in electron ionization mode. All component analyses were performed on a 30 m DB-5 (or equivalent) bonded-phase fused silica capillary column, temperature programmed from 160jC to 200jC at 2j/min and from 200jC to 260jC at 10j/min (for per-O -trimethylsilyl methyl glycosides) or from 140jC to 320jC at 4j/min (for fatty acid methyl esters and N -acetyl-per-Otrimethylsilyl sphingosines). Partially methylated alditol acetates were analyzed with the same system, temperature programmed from 160jC to 260jC at 2j/min. All derivatives were identified by comparison of retention times and mass spectra compared with authentic standards and published data.

RESULTS
Preliminary comparison of the immunoreactivity of A. fumigatus and A. nidulans GIPCs with monoclonal antibody MEST-1 In Fig. 1 are reproduced HPTLC profiles of crude fungal acidic lipid fractions stained with orcinol (panel A), which reacts with hexose residues characteristic of GIPCs, and the monoclonal antibody MEST-1 (panel B), which specifically recognizes terminal Galf b residues (25). 2 These compare the reactivity profiles of GIPCs from A. fumigatus strains 9197 and 237 (cultured at 37jC; lanes 1 and 2, respectively) with those from P. brasiliensis strain 18 (lane 3) and A. nidulans strain A28 (lane 4). The characterizations of the components from P. brasiliensis and A. nidulans have been described previously (30,40), except that the component formerly named Pb-1 is here named Pb-3. 3 The structural basis for naming the component bands from A. fumigatus will be described below.
, was previously observed to react with MEST-1 (25). The branching Galf(b1Y6) residue is known to be essential for the MEST-1 reactivity of Pb-3. Here, comigrating bands corresponding to Af-3b and Pb-3 from A. fumigatus and P. brasiliensis, respectively, are both clearly stained with MEST-1 (Fig. 1B, lanes 1-3), suggesting that Af-3b may have a structure similar or identical to that of Pb-3. Interestingly, bands corresponding to Af-4 from both strains of A. fumigatus also reacted, along with at least two other components with lower R f in each strain. This suggests that Af-4 contains the common branching Galf(b1Y6) residue responsible for MEST-1 reactivity. No other components of P. brasiliensis appeared to be reactive with MEST-1 ( Af-3a appears to be unstained by MEST-1; although this is not completely clear because of the small difference in R f with respect to Af-3b, the result would be consistent with absence, or unreactive presentation, of the Galf b residue.

Fractionation of major A. fumigatus GIPCs
The HPTLC results in Fig  the highest R f value (Af-2) comigrated with authentic M 2 IPC components previously characterized from P. brasiliensis (Pb-2) (30) and A. nidulans (An-2) (40), but the most abundant component (second band, Af-3a) had a slightly higher R f than the other major GIPC from P. brasiliensis (Pb-3). A third, less abundant component (Af-3b) migrated at the same R f as Pb-3, and the fourth band (Af-4) had a slightly lower R f , consistent with three to four monosaccharide units attached to IPC. Af-3b and Af-4 appeared to migrate, respectively, slightly above and slightly below the major GIPC component of A. nidulans, Mana6(Mana3)Mana2IPC (M 3 IPC 5 An-3) (40). A fifth component, designated Af-3c, exhibiting a much lower R f value by HPTLC, appeared in some preparations with a higher R f and, as was discovered later, sometimes appeared in the neutral (DEAE-Sephadex pass-through) fraction. The reason for this, that it is a zwitterionic GIPC component, will be discussed below. Acidic lipids from strain 9197 exhibited a qualitatively similar profile, but with less Af-3a and more Af-3b and Af-4 (data not shown). This A. fumigatus acidic fraction appeared to lack some of the additional low R f components that were observed when the fungus was cultured at 37jC (see above).
Crude acidic lipids from both strains 237 and 9197 were fractionated by preparative-scale HPLC, with each of the first four major GIPC components obtained in at least one fraction as a single orcinol-stained band (Fig. 2B, lanes 1-4); these components, designated Af-2, Af-3a, Af-3b, and Af-4, respectively, were subjected to characterization by 1 H-NMR, MS, and GC-MS techniques as described below.
An amount of Af-3c sufficient for extensive 2-D NMR analysis was accumulated from three HPLC fractionations of crude neutral lipids. An additional component, eluting before Af-2 in HPLC but not stained by orcinol, was also collected and analyzed (it will be referred to as Af-0). Some intermediate fractions were found to be mixtures; analysis of such mixtures also produced useful results, as reported below.
Monosaccharide, inositol, fatty acid, and sphingoid component analysis of A. fumigatus GIPCs Mannose was essentially the only monosaccharide identified by GC-MS analysis of the trimethylsilylated methyl glycosides produced after methanolysis of Af-2 components from strains 237 or 9197. The Af-3a (237) fraction was observed to consist mainly of mannose, although a trace of galactose was also detected. Af-3b (9197) contained both galactose and mannose in an ?1:2 ratio. In Af-3c, 2-deoxy-2-amino-glucose was detected (as N-acetylglucosamine), along with mannose, in a ratio of ?1:2. Ins was detected in all four fractions. The major long-chain fatty acid detected was h24:0, identified by its characteristic retention time and fragments observed in its electron ionization mass spectrum (47). Small amounts of nonhydroxylated 16:0, 18:0, and 24:0 fatty acids were also detected, along with traces of h25:0, h26:0, and the dihydroxy fatty acid 2h24:0. Major sphingoid components detected were t18:0 and t20:0 4-hydroxysphinganines (phytosphingosines), in ratios varying between the different fractions. Under conditions of ion trap detection, electron ionization spectra of the 4-hydroxysphinganine-derived N-acetyl-1,3,4-tri-O -trimethylsilyl-4-hydroxysphinganines were qualitatively similar in most respects than those described previously by Thorpe

Structural analysis of A. fumigatus IPC (Af-0)
A partial 1-D 1 H-NMR spectrum of fraction Af-0, reproduced in Fig. 2, exhibited resonances characteristic for a Ins spin system as well as downfield resonances characteristic of the highly hydroxylated portion of phytoceramide, which we previously observed in GIPCs, but no resonances were observed characteristic for monosaccharide residues. Upfield resonances (data not shown) included a composite multiproton alkyl/acyl CH 2 signal at ?1.23 ppm and a characteristic 6H triplet for both the alkyl and acyl CH 3 at 0.853 ppm. Therefore, this fraction appears to contain IPC, the obligate intermediate in GIPC synthesis. The resonances were assigned as shown in Fig. 3 and Table 1. The resonances assigned to Ins H-1 and Sph-1a/1b all exhibited additional three-bond coupling to the phosphorus atom of the phosphodiester (Table 1).
Because NMR analysis is less useful for determining ceramide size and relative sphingoid/fatty-N-acyl chain length distributions, the Af-0 fraction was further analyzed by 1 ESI-MS n in a linear ion trap (Fig. 4). Two major lipoforms were detected as a pair of [M(Na)1Na] 1 salt-adduct ions observed at m/z 970 and 998, consistent with IPC having ceramides consisting of t18:0 and t20:0 4-hydroxysphinganines in combination with h24:0 fatty-N-acylation (Fig. 4A). To confirm that the m/z 28 (C 2 H 4 ) increment is in the sphingoid and not the fatty-N-acyl moiety, further MS n steps were carried out to isolate and fragment the ceramide ions; because we have observed that phytoceramide ions are somewhat difficult to generate and fragment under low-energy conditions but that fragmentation is facilitated by lithium adduction (39)(40)(41), the fraction was first treated with excess LiI and reinfused into the ESI source. As shown in

Structural analysis of A. fumigatus GIPCs
Af-2 from strains 237 and 9197. A 1-D 1 H-NMR spectrum of an Af-2 fraction from strain 237 is reproduced in Fig. 5A. Aside from the presence of minor impurities, it is essentially identical to a spectrum previously published for Mana3Mana2IPC (Pb-2) from P. brasiliensis and for which most of the resonances were assigned by 2-D homonuclear COSY and TOCSY experiments (30). For confirmation, Af-2 was subjected to a similar sequence of 2-D 1 H-NMR analyses; the resonance assignments are listed for reference in Table 1 the Ins H-1, H-2, and H-3 resonances with respect to their values in InsPCer (Dy 5 0.13, 0.18, and 0.10 ppm, respectively), consistent with increments found by comparison with data previously reported for Mana2InsPCer (22). The minor anomeric resonances at 5.030 and 4.905 ppm are consistent with a small contamination by a triglycosyl-IPC component (see below) incompletely separated from Af-2 in this fraction. A 1-D 1 H-NMR spectrum of the Af-2 component from strain 9197 was of somewhat lower quality, but it lacked contamination from the triglycosyl-IPC component (data not shown). All of the salient features discussed above were visible in this spectrum, confirming its identity to the Mana3Mana2IPC isolated from strain 237.
For further characterization, particularly of the ceramide moiety, which could differ from those of other species and is not well defined by NMR methods, a molecular profile of Af-2 from strain 9197 was acquired via 1 ESI-Q/oa-TOF-MS. These data are described in detail in Af-3a from strain 237. The 1-D 1 H-NMR spectrum of a fraction of Af-3a (Fig. 5B) showed it to be not a homogeneous GIPC component but a mixture of Af-2 and an additional triglycosyl-IPC having a set of three anomeric protons at 4.905, 5.029, and 5.064 ppm. The anomeric protons from Af-2 are observed as lower intensity signals at 4.897 and 5.064 ppm, the latter coincident with one of the signals from the major triglycosyl-IPC component, as indicated by the greater relative intensity of that signal. Conversely, resonances from the major Af-3a component are observable as minor signals in the spectrum of Af-2 from strain 237 (cf. Fig. 5A). Unfortunately, because of the insufficient amount and the presence of two closely re-lated structures in the mixture, we were not able to obtain high-quality 2-D NMR data specific for the Af-3a component from this fraction, resulting in only fragmentary resonance assignments. Signals assignable by apparent analogy were Ins H-3 and H-5 resonances at 2.948 and 3.231 ppm, respectively; sphingoid H-1b and H-2 reso- Glycosidically linked 13 C are given in boldface. nances at 4.047 and 3.847, respectively; and a fatty-N-acyl H-2 resonance at 3.827 ppm. The similarity in chemical shift of two of the H-1 signals to those of Af-2 suggested that it could be a derivative of Mana3Mana2IPC, although potential glycosylation-induced changes in chemical shifts preclude reliable assignments based purely on analogy. A triglycosyl structure consisting solely of a-Manp residues would be consistent with the similar 3 J 1,2 values (1.5-2.0 Hz) of all H-1 resonances observed in the spectrum. A linkage analysis by GC-MS detected derivatives corresponding to T-Man, Y3Man, and Y2Man. The first two derivatives would be produced from the Af-2 known to be present, but the latter, assuming that Af-3a is a mannosylated product of Af-2, as the NMR data suggest, would only be consistent with addition of the third Man residue in 1Y2 linkage. Such a structure would also produce the same T-Man and Y3Man derivatives. Therefore, the structure Mana2Mana3Mana2IPC is proposed for Af-3a. This structure could be consistent with the NMR spectrum, provided that the signal at 5.029 ppm is assigned to H-1 of the penultimate Mana3 residue, because it is known that glycosylation of an a-Man residue by Mana2 induces a substantial downfield shift of H-1 of the substituted residue (51). The resonance at 4.905 ppm, therefore, is assigned to H-1 of the nonreducing terminal Mana2 residue, with H-1 of the reducing end Mana2 residue remaining at 5.064 ppm, apparently unaffected by the substitution ( Table 2).
A molecular profile of Af-3a was acquired via 1 ESI-Q/ oa-TOF-MS (Fig. 6A) Thus, in agreement with the sphingoid and fatty acid analysis, these spectra indicate that the m/z 28 increment between the two predominant molecular species is attributable primarily to a corresponding C 2 H 4 difference in the sphingoid rather than the N-acyl chain. Therefore, the [M(Li)1Li] 1 adducts at m/z 1,424 and 1,452 correspond to lipoforms containing t18:0 and t20:0 4-hydroxysphinganines, respectively, with h24:0 fatty-N-acylation.
Af-3b from strain 9197. The 1-D 1 H-NMR spectrum of fraction Af-3b showed it to be a nearly homogeneous GIPC component, although a few major peaks from non-GIPC impurities could be observed (data not shown). Three anomeric proton resonances are clearly observed in the downfield region of the spectrum (Fig. 7A) at 5.005 ppm ( 3 J 1,2 , 2.0 Hz), 4.889 ppm ( 3 J 1,2 , 2.0 Hz), and 4.828 ppm ( 3 J 1,2 5 2.0 Hz). These values are almost identical to those previously observed for the branched Galf b6containing triglycosyl-IPC isolated from P. brasiliensis, Mana3(Galf b6)Mana2IPC (30), referred to here as Pb-3. This structure would be consistent with its HPTLC R f value and immunoreactivity toward MEST-1 (see below). There were some differences in the spectra, however. These could be attributable to impurities, but to be more certain of the resonance and structure assignments, a series of 2-D 1 H-1 H and 13 C-1 H spectra were acquired, because off-diagonal correlations originating from H-1 resonances in 2-D 1 H-1 H spectra of glycoconjugates can generally be assigned in the presence of obscuring impurity peaks, and from these monosaccharide ring 1 H spin system assignments most 13 C-1 H correlations can then be interpreted.
Thus, almost all monosaccharide, inositol, and ceramide resonances were unambiguously assignable from 2-D 1 H-1 H gCOSY and TOCSY experiments (data not shown; Table 3). Analysis of approximate 3 J i, j proton-coupling constants around the three monosaccharide spin systems confirmed the presence of two a-Man residues, particularly recognizable by their signature small values for 3 J 1,2 and 3 J 2,3 . Definitive analysis of the b-Galf spin system was more problematic, but it was characterized by its essentially identical appearance to that observed previously for this residue in Pb-3 (30); acquisition of 13 C resonance NMR data (which had not been obtained previously) proved helpful in eliminating any doubt about the identity of this residue in particular (see below). Analysis of approximate 3 J i,j coupling constants around the cyclic Ins spin system confirmed it as Ins (all hydroxyl groups equatorial except that at C-2), whereas the chemical shift pattern, compared with previously acquired data, was consistent with glycosylation of Ins at O-2 (22,23,30,52). An interesting characterizing feature of the Af-3b 1 H-NMR spectral data is the downfield shift of the Mana2 H-5 (4.053 ppm), which was observed previously for Pb-3 (30), as well as another GIPC having the Mana2 residue substituted at O-3 and O-6 [i.e., Mana3(Mana6)Mana2IPC (An-3) from Aspergillus nidulans (40)].
Finally, 13 C resonance assignments (Table 3), made from a 2-D 1 H-detected, 13 C-1 H gHSQC spectrum (data not shown), were also consistent with the linkage of two glycosyl a-Manp residues to the core Mana2 residue at O-3 and O-6; this is supported by a pattern of substantial downfield shift increments (a-effects) for Mana2 C-3 and C-6, along with upfield shift decrements (b-effects) (53) for Mana2 C-2, C-4, and C-5, compared with 13 C spectral data for the parent compound, Mana2IPC (22). A striking feature of the 13 C-NMR spectral data is the extreme downfield position of the C-1 resonance correlated with H-1 at 4.828 ppm; its chemical shift of 108.3 ppm essentially confirms the identification of a b-Galf residue (53). All other 13 C chemical shifts are consistent with the proposed Mana2IPC core structure, including those characteristic for the Ins residue substituted at O-2 (22,40). Based on all of these data, the structure of Af-3b was proposed with high confidence to be Mana3 (Galf b6)Mana2IPC.    Af-4 from strain 9197. In the downfield 1-D 1 H-NMR spectrum of fraction Af-4 (Fig. 7B), four major anomeric proton signals can be observed at 4.826, 4.901, 5.000, and 5.023 ppm (all 3 J 1,2 ? 1.5-2.0 Hz), suggesting that it represents a tetraglycosyl-IPC. Two of these chemical shifts (4.901 and 5.023) are similar to those observed for H-1 resonances of Af-3a, whereas the other two (4.826 and 5.000) are similar to those observed for H-1 resonances of Af-3b, suggesting that it might contain features common to both triglycosyl-IPCs. In terms of glycosylation-induced chemical shift changes for the H-1 resonances, Af-4 could be thought of as having the same relationship with Af-3a as Af-3b has with Af-2 (addition of branching Galf b6 to the Mana3 residue), or the same relationship with Af-3b as Af-3a has with Af-2 (addition of terminal Mana2 to the Mana3 residue); for this reason, the structure Mana2 Mana3(Galf b6)Mana2IPC was proposed.
A molecular profile of Af-4 was acquired via 1 ESI-Q / oa-TOF-MS (Fig. 9A)  lower abundance in higher m/z regions of the spectra (data not shown  1 ions (data not shown) all agreed with the sphingoid and fatty acid analysis, which indicates that the m/z 28 increment between the two predominant molecular species is primarily attributable to a corresponding C 2 H 4 difference in the sphingoid rather than the N-acyl chain. Thus, the [M(Li)1Li] 1 adducts at m/z 1,586 and 1,614 again correspond to lipoforms containing t18:0 and t20:0 4-hydroxysphinganines, respectively, with h24:0 fatty-N-acylation.
The low ratios of the relative abundances of m/z 435 versus m/z 597 (0.18) in both CID spectra appear to be consistent with a structure branched at the inner a-Man residue. Interestingly, however, the m/z 273 appeared to be highly attenuated with respect to its abundance in all other spectra acquired to date; the significance of this difference is unclear. In all other respects, the fragmentation spectra are consistent with the proposed structure of Af-4, Mana2Mana3(Galf b6)Mana2IPC (Scheme 5).
Af-3c from strain 9197. A significant amount of this component was isolated from both the acidic and neutral fractions of A. fumigatus lipids. Component analysis by GC-MS indicated that the low-R f fraction Af-3c contained mannose and glucosamine (detected as GlcNAc) in a 2:1 ratio; other components detected were Ins, t18:0 and t20:0 4-hydroxysphinganines (phytosphingosines), and h24:0 fatty acid (major). The relevant downfield section of a 1-D 1 H-NMR spectrum of fraction Af-3c (corresponding material isolated from neutral fraction lipids) is reproduced in Fig. 10A; a section of the TOCSY spectrum corresponding to this region is reproduced in Fig. 10B. These data show a marked similarity to those obtained previously for Mana3Mana6GlcNa2IPC (Ss-Y6) previously isolated and characterized from the yeast form of Sporothrix schenckii (23). We still consider this structure particularly unusual (see Discussion) and therefore sought to confirm its occurrence in this context with a thorough reanalysis of fraction Af-3c, supporting it with additional heteronuclear experiments not reported previously.
As described above, chemical shift/connectivity assignments of all 1 H signals, as well as approximate measurements of 3 J i, j coupling constants, in the monosaccharide, inositol, and proximal part of the ceramide were obtained from sequential application of 2-D 1 H-1 H gradient-COSY and TOCSY experiments. Complete assignments of 1 H resonances derived from this analysis are listed in Table 4. The Ins residue is recognized as a cyclic spin system in which all 3 J i, j are large except for 3 J 1,2 and 3 J 2,3 , as H-2 is the only equatorial proton in the 1,2,3,4,5,6-hexahydroxycyclohexane ring. The three monosaccharide residues are recognized by their connectivity/coupling patterns, starting from the most downfield signal (H-1) of each spin system. Aside from the presence of two Man residues recognizable by their signature small values for 3 J 1,2 and 3 J 2,3 , the NMR spectrum of Af-3c is characterized in particular as exhibiting a sugar H-2 signal shifted far upfield at 2.460 ppm, part of a spin system, originating from H-1 at 5.011 ppm, having 3 J i, j coupling patterns consistent with an a-glucopyranosyl configuration. The upfield-shifted signal is consistent with H-2 attached to a carbon atom bearing an amino group, although in the S. schenckii GIPC, it was observed significantly farther upfield, at 2.380 ppm. The chemical shift of this proton may reflect the zwitterionic status of the GIPC, because the spectral data were obtained from an Af-3c fraction isolated from the neutral fraction of A. fumigatus lipids, whereas in the case of the S. schenckii component reported previously, it was isolated from the anionic fraction (23). Presumably, in the latter case, the deprotonated amino form would have been isolated, which condition should be consistent with a more upfield shift of the vicinally linked proton, H-2, compared with that in the zwitterionic form, in which the amino group is protonated and therefore more electron-withdrawing.
The linkage assignments were established by observation in the nuclear Overhauser effect spectroscopy spectrum (Fig. 10C) of strong interresidue correlations between H-1 of the a-GlcN residue at 5.034 ppm and Ins H-2 (4.000 ppm), between a-Man H-1 at 4.622 ppm and a-GlcN H-6 (3.564 ppm), and between a-Man H-1 at 4.883 ppm and H-3 of the first a-Man residue (3.618 ppm). Weak interresidue correlations were also observed from a-Man H-1 to a-GlcN H-5 (4.040 ppm) and other H-6 (3.675 ppm). Further unambiguous confirmation of these linkages was obtained by collection of directly detected 13 C data and correlation of these with already assigned 1 H resonances via one-bond correlations observed in a 2-D 13 C-1 H-HSQC experiment (data not shown; Table  4). Relative downfield shifts of Ins C-2, GlcNa2 C-6, and Mana6 C-3 are consistent with glycosidic linkage points on those residues. In a subsequent 13 C-1 H-heteronuclear multiple bond correlation experiment (Fig. 11), a threebond correlation was clearly observed between the a-GlcN H-1 resonance at 5.034 ppm and the uniquely assignable Ins C-2 resonance at 79.96 ppm. Additional three-bond correlations could be observed between the a-Man H-1 at 4.622 ppm and a-GlcN C-6 at 65.73 ppm and between the second a-Man H-1 at 4.883 ppm and C-3 of the first a-Man at 79.31 ppm.
In preliminary 1 MALDI-TOF-MS analysis (data not shown), two major [M(Na)1Na] 1 salt-adduct ions for Af-3c  465 and 1,493 (Fig. 12). Corresponding 1 ESI-MS/ CID-MS product ion spectra were acquired with the selection of each molecular adduct precursor; the spectrum of products from m/z 1,465 is reproduced in Fig. 12    the expected structure (data not shown). Similarly, the higher m/z [M(Li)1Li] 1 ions (m/z 1,424 and 1,452) are those consistent with the presence of smaller, but significant, amounts of triglycosyl-IPCs, either Af-3a, Af-3b, or both. This assessment was confirmed by 1 ESI-MS/CID-MS with sequential selection of each of these ions ( Fig. 13B; only the result from m/z 1,424 is shown). In this profile, an additional ion resulting from underlithiation is clearly observable at m/z 1,440; strikingly, however, the underlithiated "satellite" for m/z 1,452, expected at m/z 1,468, is strongly overlapped by a disproportionately more abundant ion at m/z 1,465. Moreover, the odd m/z of the ion suggested the presence of another N atom in the structure (as prescribed by the "nitrogen rule"), such as would be contributed by an amino sugar. This was confirmed by 1 ESI-MS/CID-MS of the ion. As shown in Fig. 13C, the resulting product ion spectrum is essentially identical to that obtained previously for the artificially N-acetylated Af-3c (Fig. 13B), suggesting the presence in A. fumigatus GIPCs of a small amount of incompletely de-N-acetylated GlcNAca2IPC precursor (Af-3c*), apparently also processed by the subsequent mannosylation steps.
This was confirmed by an appraisal of the 1 H-NMR spectrum of this fraction (data not shown), which showed the expected preponderance of Af-2, as indicated by the pair of a-Manp H-1 resonances at 5.057 and 4.894 ppm. The spectrum also exhibited a number of lower intensity resonances that appeared to correlate with those expected for triglycosyl-IPCs Af-3a (5.026 ppm) and Af-3b (4.993 and 4.829 ppm), but in addition a minor resonance was observed at 4.631 ppm, which is near to that for Mana6 H-1 of Af-3c. Subsequently, a spectrum of chemically Nacetylated fraction Af-3c was acquired; this still contained some residual unacetylated material, which facilitated the detection and measurement of small chemical shift differences attributable to the derivatization (data not shown).

Immunoreactivity of A. fumigatus GIPCs with aspergillosis serum
In Fig. 14 are reproduced HPTLC profiles of crude and purified fungal GIPCs stained with orcinol (panel A) and the serum of a patient with aspergillosis (panel B). These compare the reactivity profiles of GIPCs from A. fumigatus strains 9197 and 237 (cultured at 37jC; lanes 1 and 2, respectively) with those from P. brasiliensis strain 18 (lane 3) as well as the reactivity with purified A. fumigatus Af-3a (lane 4) and A. nidulans components An-2 and An-3 (lanes 5 and 6, respectively). These show more extensive staining, including most of the significant bands from both A. fumigatus strains (Fig. 14B, lanes 1 and 2) and P. brasiliensis Pb-3 (Fig. 14B, lane 3) and the purified Af-3a (Fig. 14B,  lane 4). The latter staining shows unambiguously in this case that Af-3a is reactive with the serum, along with components Af-3b/Pb-3 and Af-4. A variety of lower R f components from A. fumigatus also reacted with the serum. The M 2 IPC and M 3 IPC components (An-2 and An-3, respectively) from the nonpathogenic A. nidulans appeared to be nonreactive (Fig. 14B, lanes 5, 6).

DISCUSSION
The results presented here show that A. fumigatus synthesizes, in addition to InsPCer, a complex mixture of GIPCs containing ?10-12 components apparent on HPTLC analysis. In this work, five of the most abundant and least complex of these components were characterized: Af-2, Manpa3Manpa2InsPCer; Af-3a, Manpa2Manpa3Manpa2-InsPCer; Af-3b, Manpa3(Galf b6)Manpa2InsPCer; Af-4, Manpa2Manpa3(Galf b6)Manpa2InsPCer; and Af-3c, Man-pa3Manpa6GlcNpa2InsPCer. In addition, the likely biosynthetic precursor to Af-3c, in which the GlcN residue is N-acetylated, was detected as a minor component in a mixture of triglycosyl-IPCs (referred to as Af-3c* and discussed further below). Of these GIPCs, Af-3a, Af-3c*, and Af-4 have structures that were unreported previously.
Af-2 represents an apparently common intermediate dimannosyl-IPC that has now been found in a number of mycelial or dimorphic euascomycete fungi (Scheme 7, Eu-2a) (20-22, 30, 39, 40) 4 but has not been reported in hemiascomycetous yeasts such as S. cerevisiae and Candida albicans (54), nor in any basidiomycete species (52,55,56), with the sole exception of Cantharellus cibarius (chantarelle) (52). In S. cerevisiae and C. albicans, GIPC biosynthesis diverges (Scheme 7, IPMa2IPC) by addition of a second mole of myo-inositol-1-phosphate to the Man residue of mannosyl-IPC, a process shown in S. cerevisiae to be dependent on the IPT1 gene (57). In basidiomycetes, on the other hand, the dominant structural modification of mannosyl-IPC appears to be conversion to the alternative intermediate diglycosyl-IPC Galp(b1Y6)Manp(a1Y2) InsPCer (Scheme 7, Ba-2), which precedes further glycosylation steps in the formation of diverse "basidiolipids" which are found in mushrooms (52,55), as well as in GIPCs of the yeast-like basidiomycete mycopathogen Cr. neoformans (56). The Galpb6-and Manpa3-transferases responsible for the further glycosylation of Manp(a1Y2) InsPCer in euascomycetes and basidiomycetes, respec-4 Although Barr, Laine, and Lester (21)   tively, remain unidentified. In A. nidulans, the major GIPC was found to be the product of An-2 (5 Eu-2a) with a Mana6-transferase to yield a branched tri-a-mannosyl structure (Scheme 7, An-3) similar to that of the Man 3 GlcNAc 2 core of protein N-linked glycans (except in that case the branching Man residue is b-, rather than a-).
Among the triglycosyl-IPCs found in A. fumigatus, Af-3b has been reported previously as a component of both yeast and mycelial forms of the dimorphic mycopathogen P. brasiliensis (30). A closely related structural variant, identical to Af-3b but possessing a branching a-, rather than b-, Galf residue was reported among the GIPCs of the dimorphic mycopathogen Histoplasma capsulatum (20,21). 4 The H. capsulatum antigen was strongly reactive with sera from patients with histoplasmosis, but, unlike the case with P. brasiliensis, it was detected only in GIPCs from the yeast form. The glycosylinositol structure of Af-3a is novel, apparently unreactive with MEST-1 but reactive with polyclonal sera from a patient with aspergillosis. The tetraglycosyl-IPC Af-4, reactive with both MEST-1 and aspergillosis serum, appears to be a product of further glycosylation of the Manpa3 unit in Af-3a or Af-3b by Galf b6 or Manpa2 res-idues, respectively; putative convergent pathways to Af-4 via these intermediate triglycosyl-IPC structures are illustrated in Scheme 8.
The identification of Af-3c, which contains a nonacetylated GlcNp residue linked a1Y2 to Ins, is highly significant in several respects. A GIPC with this structure was originally isolated and characterized from the yeast form of the dimorphic mycopathogen S. schenckii (23); at that time, it was pointed out that the GlcNpa1Y2Ins linkage is isomeric to the GlcNpa1Y6Ins linkage found in the core of all protein and phosphoglycan glycosylphosphatidylinositol (GPI) anchors (58)(59)(60). Since that time, GPI anchors of A. fumigatus proteins have been characterized and found to have hexaglycosylinositol phosphorylceramide core structures: Mana3Mana2Mana2Mana6Mana4GlcNa6 InsPCer (61). In addition, a 30 kDa galactomannan from A. fumigatus was found to be anchored by a GPI with a similar core structure, although in this case the linkage of GlcN to Ins was not defined unambiguously: Mana2 Mana2Mana6Mana4GlcN[x]InsPCer (62). It is worth noting that the GlcN-containing GIPCs reported here differ from the structure of GPI anchors not only in the GlcN-Ins linkage position but also in the linkage sequence of the attached Man residues. Thus, they appear to be synthesized by a pathway altogether distinct.
On the other hand, it was hypothesized that, similar to the case with GlcNpa1Y6Ins of GPI anchors, the GlcNpa1Y2Ins structure might be formed by a process requiring two enzymes: transfer of GlcNAc to the Ins moiety, followed by de-N-acetylation of the GlcNAc residue (23). More recently, a series of novel GIPCs, also based on the GlcNpa1Y2Ins core structure, was isolated from mycelia of Acremonium species (31). In this case, the alternative triglycosyl-IPC structure Manp(a1Y6)Manp(a1Y6) GlcNp(a1Y2)InsPCer was identified, along with all of the presumptive intermediates. In addition, a remarkable finding was the addition of phosphocholine to the 6-hydroxyl of the terminal a-Manp residues of both Manp(a1Y6) GlcNp(a1Y2)InsPCer and Manp(a1Y6)Manp(a1Y6) GlcNp(a1Y2)InsPCer. The suggestion of a two-step conversion of InsPCer to the core intermediate GlcNp(a1Y2) InsPCer was again invoked (31). We propose that the small amount of triglycosyl-IPC observed in a mixed fraction of A. fumigatus GIPCs (Af-3c*), clearly containing a HexNAc residue directly linked to Ins, is Manp(a1Y3)Manp (a1Y6)GlcNAcp(a1Y2)InsPCer: in other words, a structural analog of Af-3c derived from incompletely de-N-acetylated GlcNAcp(a1Y2)InsPCer. Although this needs further confirmation, it lends support to the suggestion that a two-step process, similar to that observed for GlcNpa1Y6Ins in the GPI anchor pathway, also applies to GlcNpa1Y2Ins in the GIPC pathway. Whether the same de-N-acetylase could be involved is an interesting question that should be answered by further in vitro and in vivo studies. A more important question, of course, is whether any special functional role(s) might be associated with this zwitterionic class of GIPCs, with or without the addition of phosphocholine. In this regard, it is significant that the Acremonium species GIPCs were found to induce cell death in suspension-cultured rice cells (31).
Thus, we consider that three core motifs have been confirmed at the monosaccharide level for fungal GIPCs (by identification in structural studies from at least two different laboratories); these are Mana1Y2Ins, Mana1Y 6Ins, and GlcNa1Y2Ins (Scheme 1). Of these, the Mana1Y 6Ins core has been reported only for GIPCs of S. schenckii (22,24); remarkably, S. schenckii expresses GIPCs with all three core structures (23). As shown in the current study, A. fumigatus is sufficiently versatile at least to use two of these. Notably, the GlcNa1Y2Ins core has not been observed in GIPCs from the related model fungus A. nidulans; to date, only oligo-a-mannosyl-IPCs, based on the Manp(a1Y3)Manp(a1Y2)InsPCer core, have been isolated from A. nidulans, with no evidence for the expression of antigenic determinants containing b-Galf residues (40). This interesting dichotomy suggests that wider versatility in GIPC biosynthesis may be one factor in the survival of A. fumigatus in highly diverse environments, including the entire range of mammalian host cells, tissues, and organs.