Characterization of a Low Molecular Weight Glycolipid Antigen from Cryptosporidium parvum*

Cryptosporidium parvum, an Apicomplexan parasite of the mammalian gut epithelium, causes a diarrheal illness in a wide range of hosts and is transmitted by contamination of food or water with oocyst-laden feces from an infected animal. We have identified a glycosylinositol phospholipid from the sporozoite stage of the parasite that is frequently recognized by serum antibodies from human cryptosporidiosis patients. The humoral immune response is dominated by IgG1 subclass antibodies but can also include IgA and IgM antibodies. The glycosylinositol phospholipids were purified by butanol extraction of a Triton X-114-soluble fraction followed by octyl-Sepharose column chromatography and preparative high performance TLC and were shown to include at least 5 species. By using mass spectrometry and radiolabeled neutral glycan analysis, we found that the structure of the dominant glycosylinositol phospholipid antigen contained a C18:0 lyso-acylglycerol, a C16:0-acylated inositol, and an unsubstituted mannose3-glucosamine glycan core. Other diacyl species were also identified, most notably a series of glycosylinositol phospholipids having an acyl-linked C20:0 to C28:0 lipid on the inositol ring. Less abundant species having three acyl-linked fatty acids and species with an additional 1–3 hexoses linked to the mannose core were also observed. We are currently working to determine the role that these glycolipids may play in the development of disease and in the clearance of infection.

waterborne outbreaks have been linked to contaminated municipal water supplies (8 -11). A breakdown in the filtration system in Milwaukee in 1993 led to the infection of more than 400,000 residents in the greater metropolitan area (12,13). Infection in the immunocompetent host may be asymptomatic or may lead to a self-limiting diarrheal illness (14). However, in immunocompromised and immunosuppressed individuals, the disease is often severe and chronic and may contribute significantly to mortality (15).
Infection with C. parvum has been shown to elicit transient antibody responses that are directed mainly against two sporozoite antigens having apparent molecular masses of 27 and 17 kDa (16 -19). In a study of C. parvum infections in human volunteers, antibody responses to these antigens were associated with protection from diarrheal symptoms (20). Both of these antigens are associated with the sporozoite surface, and subsets of the antigens can be partially purified from sonicated oocysts by phase partitioning into Triton X-114 detergent (19). In earlier work we used this technique to purify the 17-kDa antigen for peptide sequence analysis and to generate a native antigen fraction suitable for use in an enzyme-linked immunoassay for the detection and quantitation of serum IgG antibodies (19,21). While analyzing the antibody response to the Triton X-114 detergent extract by Western blot, we noted that a number of cryptosporidiosis patients also reacted with a novel low molecular weight antigen. In this work we report that this antigen is actually a family of glycosylinositol phospholipids (GIPLs). 1 GIPLs and related structures that anchor some surface proteins into the membrane, glycosylphosphatidylinositol (GPI) anchors, are present in large quantities in the surface membranes of many protozoan parasites and have been recognized recently as important effectors of the host immune response during infection (reviewed in Refs. 22 and 23). GPI anchors and/or GIPLs from medically important parasites such as Trypanosoma cruzi, Leishmania mexicana, Trypanosoma brucei, Plasmodium falciparum, and Toxoplasma gondii have been shown to modulate immune system function as both suppressors and activators (23,24). A GPI-derived toxin has been implicated in pathogenesis in cases of severe malaria (25)(26)(27). In this work, we report that many cryptosporidiosis patients have a serum antibody response to sporozoite-derived GIPLs, and we report the purification and structural analysis of the sporozoite GIPL antigens.

MATERIALS AND METHODS
Purification of Antigen-C. parvum oocysts (Maine isolate) were purified from the feces of experimentally infected Holstein calves as described by Arrowood and Sterling (28). A crude antigen preparation was obtained by sonication and freeze/thaw of the oocysts as described previously (29). The crude antigen preparation was extracted with Triton X-114 using a modification of the method of Ko and Thompson (19,30). Antigens were acetone-precipitated from the detergent fraction (4 volumes of cold acetone with overnight incubation at Ϫ20°C) and dissolved in buffer containing 0.1% SDS, 25 mM Tris, pH 8.0, and 1 mM EDTA for a final protein concentration of ϳ1 mg/ml (BCA protein microassay; Pierce). Glycolipids were extracted from the antigen fraction with water-saturated 1-butanol (2 times, 1 volume) and then dried under vacuum. Butanol-extracted glycolipids were further purified by octyl-Sepharose (OS) chromatography (1 cm inner diameter ϫ 10 cm length) as described by McConville et al. (31). The fractions having the strongest orcinol/H 2 SO 4 reaction (1 l of each 1-ml fraction was spotted on a silica gel HPTLC plate (EM Science, Gibbstown, NJ)) (32) were pooled (fractions 86 -92), dried under vacuum, and dissolved in 40% 1-propanol. An aliquot of OS-purified glycolipid was further fractionated by preparative silica gel HPTLC using a solvent system of 10:10:3 (v/v/v) chloroform, methanol, 1.0 M NH 4 OH (31). A strip was cut from the side of the preparative lane and stained with orcinol/H 2 SO 4 reagent to locate the positions of the component glycolipid bands. The regions corresponding to each glycolipid band were scraped into a tube and extracted once with water-saturated butanol and once with chloroform/ methanol/water, 1:2:0.8. The combined eluates were dried under vacuum, dissolved in water-saturated butanol, and spotted onto polyvinylidene difluoride (PVDF) membranes (Immobilon P; Millipore Corp., Bedford, MA) for Western blot analysis.
Strips of PVDF membrane containing Triton X-114-soluble antigens were incubated 1-48 h at room temperature with 0 -25 mM NaIO 4 in 50 mM sodium acetate buffer, pH 4.5, in the dark, washed extensively, blocked, then incubated overnight at 4°C with diluted human patient serum, and developed for bound IgG antibodies as described above (35). A strip of PVDF membrane containing Triton X-114-soluble antigens was incubated overnight at 37°C in PBS containing 100 g/ml proteinase K. The strip was then washed twice for 1 h with PBS, 0.3% Tween 20 containing 1 mM phenylmethylsulfonyl fluoride, incubated overnight with diluted patient serum, and developed for bound IgG antibodies as described above.
Analysis of myo-Inositol and Phosphatidylinositol (PI) from C. parvum Antigens-Triton X-114-extracted antigens (120 g of protein) were resolved by gradient SDS-PAGE and transferred to PVDF as described previously (19). After staining the proteins with Amido Black, strips of membrane of approximately equal sizes were cut from the following: 1) above the 17-kDa antigen; 2) the stained 17-kDa antigen protein band near the 14.3-kDa marker; 3) immediately below the 17-kDa antigen band; and 4) a region between the 6-and 3-kDa markers. About 15% (20 g of protein) of each strip was subjected to 6 M HCl acid hydrolysis (16 h at 110°C) with 50 pmol of D 6 -myo-inositol added as an internal standard. The hydrolysis products were dried, derivatized with trimethylsilane, and analyzed by gas chromatography and selected ion gas chromatography-mass spectrometry (GC-MS) as described previously (36,37).
The remainder of each strip (about 100 g of total protein loaded) was used for PI analysis (38). The strips were treated with 0.5 M NaNO 2 in 0.15 M sodium acetate, pH 4.0, for 2 h at 37°C. The strips were then washed twice in water and extracted three times for 15 min with water-saturated butanol at 37°C. Samples were sonicated to improve the extraction efficiency. The combined butanol extracts were dried under vacuum, dissolved in 4:1 chloroform/methanol, and loaded onto 200-l silica gel mini columns (70 -230 mesh). After extensive washing with 4:1 chloroform/methanol, bound PI was eluted with 1:4 chloroform/ methanol for analysis.
Methylation of GIPL Glycan-Approximately 200 pmol of OS-purified GIPLs (based on GC-MS myo-inositol analysis) were deacylated in 1:1 (v/v) 40% methanol/ammonium hydroxide at 37°C for 1 h and then deaminated with NaNO 2 as described above. At the end of the deamination reaction, boric acid was added to a final concentration of 73 mM, and the pH was adjusted to 10.7 with NaOH. NaB 2 H 4 was added, and the deutero-reduction proceeded overnight at room temperature. The reaction was acidified by the addition of 0.5 volumes of 1 M acetic acid, and the reaction products were passed over a 0.5-ml Dowex AG50-X12(H ϩ ) column. The eluate was collected, dried by rotary evaporation, and sequentially dried from methanol, 1% acetic acid, toluene (2 times), and methanol (2 times). The deuterated products were permethylated with methyl iodide in NaOH/Me 2 SO as described previously (37).
C. parvum GIPL Neutral Glycan Labeling and Analysis-A membrane strip from region 4 (between 3 and 6 kDa) of a blot containing 140 g of Triton X-114-extracted antigen was minced into 2-mm squares and washed sequentially with excess methanol and water. The membrane fragments were pre-reduced with 200 mM NaBH 4 for 1 h at room temperature and then washed sequentially with excess water, 50 mM acetic acid, and water again. Antigens were deaminated for 2 h at 37°C with 0.5 M NaNO 2 in 0.15 M sodium acetate, pH 4.0, then washed with excess water (2 times) and with 0.1 mM NaOH. Those antigens still FIG. 1. Human IgG antibody responses to Triton X-114-soluble C. parvum antigens. Triton X-114-extracted antigens from sonicated C. parvum oocysts were resolved on a 10 -22.5% SDS-polyacrylamide gradient gel and were blotted onto a PVDF membrane. Sera from individual cryptosporidiosis patients (collected 43 days after a single exposure to contaminated food) were incubated at a 1:100 dilution in PBS, 0.3% Tween 20 with 2-mm strips overnight at 4°C. Bound IgG antibodies were visualized using a biotinylated mouse anti-human IgG monoclonal antibody and alkaline phosphatase-labeled streptavidin as described under "Materials and Methods." Locations of the 27-, 17-, and 3-6-kDa antigens and of the molecular weight markers are indicated.
bound to the membrane were labeled for 2 h at room temperature with 2.5 mCi of NaB 3 H 4 in 10 mM NaOH (70-l volume) (38). Unlabeled NaBH 4 (5 l of a 1 M solution) was added to the sample, and incubation was continued for an additional 1 h at room temperature. The membrane fragments were removed and washed sequentially with excess water, 50 mM acetic acid, PBS (2 times), and water again (2 times). HF treatment (50% aqueous solution at 4°C overnight) was used to release labeled neutral glycans from the membrane fragments. The HF solution was collected and lyophilized. The resulting labeled neutral glycans were re-N-acetylated with acetic anhydride and treated with neuraminidase (50 milliunits for 1 h at 37°C) prior to desalting on a column containing 0.2 ml of Chelex 100 (Na ϩ ), 0.2 ml of Dowex AG50-X12(H ϩ ), 0.4 ml of Dowex AG3-X4 (OH), and 0.2 ml of QAE-Sephadex-A25(OH Ϫ ) as described previously (37). Approximately 100,000 cpm were recovered in the final neutral glycan fraction.
Electrospray-Mass Spectrometry (ES-MS) Analysis-Mass spectra were collected on a Micromass Q-Tof2 mass spectrometer using nanoflow tips. For negative ion mode mass spectrometry, samples were dissolved in chloroform/methanol (2:3) at ϳ1 pmol/l. For positive ion mode mass spectrometry, OS-purified GIPLs were dissolved in 30% 1-propanol with 10% acetic acid, and permethylated glycans were dissolved in 80% acetonitrile with 1 mM sodium acetate. Capillary and cone voltages were 900 and 60 V, respectively. Collision-induced dissociation (CID) daughter ion spectra were collected using argon as collision gas at 2.5 ϫ 10 Ϫ3 torr and an accelerating voltage of 43-55 V.
Serum Specimens-Patient sera were available from individuals who were naturally infected during several C. parvum outbreaks. Samples were chosen for further analysis based upon the presence of Cryptosporidium-specific IgG antibody responses (19). Informed consent was obtained from patients prior to sample collection. This study was reviewed and approved by the Institutional Review Board at the Centers for Disease Control and Prevention.

Serum Antibodies from Cryptosporidiosis Patients Recognize a Family of Very Low Molecular Weight Antigens-
In previous work, we demonstrated that human infection with C. parvum elicits an IgG antibody response to the 17-and 27-kDa surface antigens within 10 -14 days of symptom onset (19,40). Although both of these antigens include post-translational carbohydrate and/or lipid modifications, we demonstrated that much of the antibody response was directed against the protein component (19,21). While examining the IgG antibody reactivity to the subset of those antigens that were extracted into Triton X-114 detergent (the 27-and 17-kDa antigen families include both soluble and membrane-bound forms), we noted that an additional antigen having a molecular mass of Ͻ6-kDa was recognized by serum IgG antibodies from ϳ50% of the patients ( Fig. 1 and data not shown). Although the resolution in that area of the Western blot was insufficient to allow the visual identification of distinct bands, we did note that there were differences in the apparent molecular weights of the antigens The bound IgA, IgM, IgG, and IgG subclass antibodies were detected using biotinylated class-and subclass-specific mouse anti-human monoclonal antibodies as described under "Materials and Methods." B, strips of PVDF membrane containing Triton X-114-soluble antigens were incubated at room temperature at the indicated times with the indicated concentrations of NaIO 4 in 50 mM sodium acetate buffer, pH 4.5, in the dark. The strips were then washed twice with buffer and incubated for 1 h at room temperature with 1% glycine in PBS. The strips were then incubated overnight at 4°C with a human patient serum (diluted 1:100 in PBS, 0.3% Tween 20) and developed for bound IgG antibodies as described previously. C, a strip of PVDF membrane containing Triton X-114-soluble antigens was incubated overnight at 37°C in PBS containing 100 g/ml proteinase K. The strip was then washed twice for 1 h with PBS, 0.3% Tween 20 containing 1 mM phenylmethylsulfonyl fluoride, incubated overnight with diluted patient serum as in B, and developed for bound IgG antibodies as described previously. Below 17-kDa antigen  8  4 3-6-kDa antigen 90 a Region 1 was cut from the vicinity of the 18-kDa marker. Region 2 was cut from the stained 17-kDa antigen protein band near the 14-kDa marker. Region 3 was cut immediately below the 17-kDa antigen protein band. 3-6-kDa antigen was cut from region 4 between the 3-and 6-kDa markers.
b The integrated intensities of the characteristic fragment ions at m/z of 307 and 321 (for D 6 -labeled internal standard) and at m/z of 305 and 318 (for the H 6 species) were used to calculate the relative amounts of myo-inositol in each membrane sample (37). recognized by the various patients; some patients had antibodies that recognized antigens spread between the 3-and 6-kDa markers, whereas others recognized a narrow band of antigen closer to the 3-kDa marker. A subclass analysis of the IgG antibody response to the 3-6-kDa antigen indicated that most of the antibodies were of the IgG 1 subclass, although IgG 3 and IgG 4 responses were present ( Fig. 2A). In contrast, most of the antibodies to the 27-kDa antigen were of the IgG 2 subclass. In addition to the IgG response, serum IgA and IgM antibodies to the 3-6-kDa antigens were also present in some patients ( Fig.  2A and data not shown). Recognition of the 3-6-kDa antigen by total IgG antibodies was abolished in a concentration-and time-dependent fashion by preincubation of the antigen in sodium periodate (Fig. 2B). As expected for a protein-directed response, antibody recognition of the 17-and 27-kDa antigens was largely unaffected by periodate treatment. In contrast, pretreatment of the antigens with proteinase K did not significantly affect the recognition of the 3-6-kDa antigens but completely eliminated antibody binding to both the 17-and 27-kDa antigens (Fig. 2C). Taken together, these results suggest that the antigens in the 3-6-kDa size range are composed mostly of carbohydrate or carbohydrate-dependent epitopes.
Identification of the 3-6-kDa C. parvum Antigens as GIPLs-A selected ion GC-MS technique that was previously used for the detection of GPI-anchored proteins (36) was used to demonstrate that the 3-6-kDa antigen region of blotted Triton X-114 extract contained myo-inositol. Analysis of a blot of Triton X-114-extracted antigens between 18 and 3 kDa demonstrated the presence of two distinct peaks of myo-inositol, one corresponding to the GPI-anchored 17-kDa antigen (apparent molecular mass of 12-14 kDa) and another corresponding to the newly identified 3-6-kDa antigen ( Table I).
The presence of myo-inositol in the region of the 3-6-kDa antigens and the sensitivity of the antigens to periodate digestion suggested that they might, in fact, be GIPLs. To confirm this hypothesis, we purified a GIPL fraction from the Triton X-114 extract by butanol extraction and OS column chroma-tography. As shown in Fig. 3, two sera from cryptosporidiosis patients (A, serum 1 and 2) and a monoclonal antibody that recognized antigens in the 3-6-kDa region (A, mAb 18.44) (34,41) also reacted strongly with OS-purified GIPLs (C, total GIPL). A serum from a cryptosporidiosis patient that reacted with the 17-and 27-kDa antigens but not with the 3-6-kDa antigens on Western blot (A, serum 3) reacted only weakly with the purified GIPLs (C, total GIPL). Fractionation of OS-purified GIPLs on a HPTLC plate revealed the presence of three major components (Fig. 3B, fractions 2-4) and at least two minor components (fractions 1 and 5). These components were differentially recognized on Western blots by the human sera (Fig. 3C); serum 1 recognized all 5 GIPLs, whereas serum 2 recognized only the GIPLS in fractions 4 and 5 and serum 3 (negative control) reacted very weakly with fraction 2. The monoclonal antibody reacted most strongly with the GIPLs in fraction 5. From these results, we can conclude that the 3-6-kDa antigens recognized by the human immune response are GIPLs and that individual human antibody responses are focused on different components of the GIPL fraction.
Analysis of the Lipid Components of the 3-6-kDa GIPL Antigens-Strips of PVDF membrane from the 3-to 6-kDa antigen region and from region 3 immediately below the 17-kDa antigen (see Table I) were deaminated with nitrous acid and extensively extracted with butanol to collect the released PIs. Negative ion ES-MS analysis of the extract from region 3 revealed that no PIs were present in the sample, as expected from the inositol analysis (data not shown). In contrast, the negative ion ES-MS spectrum of the extract from the 3-to 6-kDa antigen region (Fig. 4A)  lyso-acyl-(acyl)PI with the weaker carboxylate fatty acid ion attached to the 2-position of the inositol and the stronger carboxylate fatty acid ion attached to the glycerol backbone (usually at the sn-1 position). The presence of a fragment ion at m/z 419 confirms that the C18:0 fatty acid is indeed located at the sn-1 position of the glycerol. The weaker inositol 1,2-cyclic phosphate fragment ions at m/z 241 and 619 indicate that, at least in some fraction of the parent lyso-acyl-(acyl)PI species, the C26:0 fatty acid substituent is also located somewhere other than the 2-position of the inositol ring. The exact location of the C26:0 fatty acid on the inositol ring of these species could not be inferred from the collision spectrum, but acylation at the 3-position of the inositol ring has been reported previously (42) in another system.
A CID daughter ion spectrum for one of the triacylated PI species is shown in Fig. 4C. The carboxylate ion fragments at m/z 255, 281, and 283 in the daughter ion spectrum of the PI at m/z 1101.8 indicate that this species includes three different fatty acids, C16:0, C18:1, and C18:0, respectively. Because the C18:1 fatty acid substituent is not found on either the cyclic glycerophosphate or the cyclic phosphoinositol fragment ions, we believe that this fatty acid occupies the sn-2 position of the glycerol backbone. From the cyclic glycerophosphate fragment ions at m/z 391 and 419, we can infer that the sn-1 position of the glycerol is occupied by both C16:0 and C18:0 fatty acids. Similarly, from the inositol 1,2-cyclic phosphate fragment ions at m/z 479 and 507, we can infer that both of these fatty acids are also found linked to the inositol ring. We conclude from these results that the m/z 1101.8 parent ion is actually composed of two different isobaric diacyl-(acyl)PI species. The relative abundance of the fragment ions (m/z 479 versus 507 and m/z 391 versus 419) and the weak C16:0 carboxylate response (m/z 255) suggest that the species having a C18:0 at the glycerol sn-1 position and a C16:0 on the inositol ring is the more abundant of the two PI forms. The presence of both the m/z 153 and 241 cyclic phosphate fragment ions indicates that the acyl-linked fatty acid of the diacyl-(acyl)PI species, like that described above for the lyso-acyl-(acyl)PI species, is found both at the 2-position and at some other position of the inositol ring.
From the CID daughter ion spectra, we were able to use the rules described above to determine the fatty acid compositions of the various PI species identified in Fig. 4A. As shown in Table II, all of the PI species were acylated on the inositol ring with fatty acids ranging from 16 to 28 carbons in length. The sn-1 position of the glycerol backbone was occupied almost exclusively by C18:0 and C16:0 fatty acids, with the former generally being more abundant. The CID daughter ion spectra indicate that unsaturated fatty acids (C18:1) were found only on the glycerol backbone of the PI, usually at the sn-2 position. This position was also occupied by a C18:0 fatty acid and (rarely) by a C16:0 fatty acid.
Analysis of the Carbohydrate Components of the 3-6-kDa GIPLs-To examine the carbohydrate substituents of the OSpurified 3-6-kDa GIPLs, positive ion ES-MS and ES-MS-CID-MS were performed. As shown in Fig. 5A, the C. parvum GIPL antigen ES-MS spectrum was dominated by peaks at m/z 1459 and 1487. A CID daughter ion spectrum of the largest ion (m/z 1811) (Fig. 5B) suggested the presence of a PI with C16:0 and C18:0 fatty acid substituents, as well as one hexosamine and five hexose residues. A suggested fragmentation scheme for the CID spectrum in Fig. 5B is shown in C. No ethanolamine phosphate or 2-aminoethylphosphonate modifications were detected.
CID daughter ion spectra were used to determine the identities of many of the species observed in the OS-purified GIPL ES-MS spectrum in Fig. 5A. As shown in Table III, glycans composed of 0 -5 hexose residues along with one hexosamine were detected. The major ion peaks at m/z 1459 and 1487 were composed of a hexose 3 -hexosamine glycan with either two acyllinked C16:0 fatty acids or one C16:0 and one C18:0 acyl-linked fatty acid, respectively. Similar fatty acid configurations were observed for each of the possible glycan cores. All of the GIPLs that were examined by CID-MS had one fatty acid on the glycerol backbone and one fatty acid on the inositol ring. Interestingly, no GIPLs with three fatty acid substituents and no GIPLs with C20:0-C28:0 fatty acid substituents were observed in the positive ion spectrum (Table III). This contrasts with the negative ion spectrum of PIs extracted from blotted Triton X-114 extract shown in Fig. 4A and summarized in Table II. We would suggest that either the more hydrophobic GIPL components were inadvertently excluded from the OS column fraction pool used for positive ion ES-MS analysis or that the PI fraction used for negative ion ES-MS was contaminated with non-GIPL PIs that were also present in the 3-6-kDa region of the blotted Triton X-114 extract. We think the latter possibility is unlikely because negative ion ES-MS spectra of chloroform/methanolextracted phospholipids showed no evidence of C20:0-C28:0 acyl-linked fatty acids. 2 Another interesting feature of the positive ion GIPL analysis summarized in Table III is the presence of several diacyl-GIPL species having one unsaturated fatty acid. Because we did not collide any of these species, we cannot definitively locate the C18:1 fatty acid. However, since all of the GIPL species that were analyzed by ES-MS-CID-MS had one fatty acid on the inositol ring and one fatty acid on the glycerol backbone and since no C. parvum PIs have been observed to have an unsat- The positions of the substituents on the glycerol (R 1 ϭ sn-1 and R 2 ϭ sn-2) were assigned by the presence or absence of the fatty acid on the cyclic glycerophosphate and inositol 1,2-cyclic phosphate fragment ions in the CID daughter ion spectra.
b The presence of both cyclic glycerophosphate and inositol 1,2-cyclic phosphate daughter ions in all of the CID spectra indicates that these fatty acids were located at the 2-position of the inositol ring and at another position on the ring. c Predominant species. d A CID daughter ion spectrum was not obtained on this species, and the identities and locations of the substituents are speculative. e The CID daughter ion spectra do not indicate the presence of any unsaturated fatty acids on the fragments derived from the acylated inositol.
urated fatty acid on the inositol ring, we think it likely that the C18:1 fatty acid is alone on the glycerol backbone of these GIPLs. We did note the presence of a weak ion in the negative ion PI spectrum that might correspond to a C18:0, C18:1 diacyl species, but we did not obtain a daughter ion spectrum (Table II).  Table III. Note that GIPLs with three acyl-linked fatty acids and lyso-acyl-GIPLs with C20:0 to C28:0 acyl-linked fatty acids on the inositol ring are missing from this spectrum. These species were likely not present in the orcinol-positive peak fractions that were pooled for analysis. B, ES-MS-CID-MS daughter ion spectrum of the m/z 1811 parent ion is shown. C, fragmentation scheme for the daughter ions produced by collision induced dissociation of the m/z 1811 parent ion shown in B.

Structures of the Carbohydrate Components of the 3-6-kDa
GIPLs-To examine the structures of the carbohydrates present in the OS-purified GIPLs, the deaminated and reduced neutral glycans were permethylated, and the reaction products were analyzed by positive ion ES-MS and ES-MS-CID-MS. As shown in Fig. 6A, four neutral glycan species were detected at m/z 1264.6, 1060.5, 856.4, and 698.4. These species correspond to the monosodium ions of deuterated 2,5-anhydro-mannitol (AHM) with 2-5 hexose substituents (see Table IV), with the hexose 3 species being the most abundant. These results are similar to those obtained earlier by positive ion ES-MS of the intact GIPL fraction. The CID daughter ion spectrum of the m/z 1060.5 parent ion is shown in Fig. 6B, and the fragment ion assignments in Fig. 6C suggest that the hexose 4 species contains both branched and linear carbohydrate chains; fragment ions at m/z 624.3 and 420.2 could only be produced by the removal of two terminal tetra-O-methyl-hexoses, whereas the fragment ion observed at m/z 434.2 could not be produced if the molecule was branched at that point (Table IV). Similarly, the presence of fragment ions at m/z 828.4, 624.3, and 420.2 and the absence of a fragment ion at m/z 434.2 in the CID daughter ion spectrum of the m/z 1264.6 hexose 5 parent ion (data not shown) would suggest that all of these molecules were branched at the hexose adjacent to the deuterated AHM (Table  IV). These molecules also appear to be heterogeneous in that the observed m/z 638.3 and 624.3 fragment ions could only be produced from molecules having one hexose and two hexoses, respectively, at the branch point. In contrast, the hexose 3 species appears to be both homogeneous and linear (Table IV), because none of the characteristic ions derived from a branched molecule were observed in the daughter ion spectrum (data not shown).
To determine the identities and linkage arrangements of some of the hexose components, C. parvum GIPLs in the 3-6-kDa size range were deaminated, labeled by reduction with tritiated NaBH 4 , recovered from the blot, and digested with various exoglycosidases. When the labeled neutral glycans were chromatographed on an HPTLC plate, four bands were visible (Fig. 7A, 3-6-kDa GIPL NG) as follows: a major band (band a) that migrated at the same position as a Man␣1-2-Man␣1-6-Man␣1-4-AHM standard (Man 3 Std.) and three other bands (bands b-d) that ran closer to the origin. None of these three additional bands appeared to migrate at the same position as the linear mannose 4 standard (see Man 4 Ladder lane, bottom band). Their spacing relative to band a suggests the presence of a mannose 3 core with 1-3 additional carbohydrate residues in a branched configuration. When the labeled neutral glycans were treated with JBAM, band a was completely digested, and the intensities of bands b and c were diminished (Fig. 7A, GIPL NG ϩ JBAM). Three new bands were observed in the JBAM digest as follows: band aЈ migrated at the same position as AHM as would be expected following the removal of three mannoses from parent band a, and bands bЈ and cЈ migrated at positions that would be consistent with the removal of two mannose residues from parent bands b and c, respectively. The mannoses adjacent to the AHM in bands bЈ and cЈ are likely not sensitive to JBAM cleavage because of the presence of a side-chain substituent. Similarly, three new bands were observed following treatment of the labeled neutral glycans with ASAM: bands aЉ, bЉ, and cЉ migrated at positions that were consistent with the removal of one mannose residue from the parent a, b, and c glycans, respectively. Band cЉ is not likely to represent undigested band a since the conditions used in the reaction were sufficient to digest completely the Man 3 standard in a parallel reaction (data not shown). The presence of undigested bands b-d glycans in both the JBAM and ASAM digests suggests the presence of a terminal carbohydrate residue other than mannose on some of the components of these bands. We were unable to identify this carbohydrate because the glycans appeared to be resistant to the other commercially available exoglycosidases, namely, ␣and ␤-galactosidase and ␤-mannosidase.
Based on the exoglycosidase results described above, we suggest that the components of the neutral glycan bands in Fig. 7A are consistent with the structures presented in B. The linear mannose 3 species in band a is consistent with the linear hexose 3 -deutero-AHM species detected by permethylation analysis. We believe this is the predominant glycan found on the C. parvum GIPLs and that this is the structure found on the fraction 2 GIPLs purified in Fig. 3B. As shown in Fig. 7B, band b is likely composed of both a linear hexose 4 species that is resistant to ␣-mannosidase cleavage and a branched species that can be cleaved to the branch point by JBAM. These structures are entirely consistent with the permethylation analysis of the 1060.5 parent ion shown in Fig. 6, B and C. Similarly, the structures shown for band c in Fig. 7B are consistent with the permethylation analysis results in that all of the molecules are branched. We believe these glycans are the ones purified in fraction 4 of Fig. 3B and that these are the glycans that dominate the human immune response to the GIPLs. The structures proposed for band d in Fig. 7B must be considered more speculative than the others because no definitive digestion products could be linked to these parent compounds. We also have no corroborating evidence on these two proposed structures from the ES-MS analyses described previously. However, we believe that the glycans seen in band d are not likely to be artifacts because of the presence of the fraction 5 GIPLs shown in Fig. 3B. The fraction 5 GIPLs were the least abundant by mass, but they reacted very strongly with monoclonal antibody 18.44 (Fig. 3C) and were also recognized by both of the positive human sera. DISCUSSION In this work, we report the presence of a family of GIPLs in the membrane fraction of C. parvum sporozoites, and we demonstrate that the most abundant species of GIPL contains a C18:0 lyso-acylglycerol, a C16:0 acylated inositol, and an unsubstituted mannose 3 -glucosamine glycan core. Because the  triacyl species containing two fatty acids on the glycerol backbone, and species with an additional 1-3 hexoses on the mannose core were also detected but were in lower abundance.
Consistent with our previous observation that treatment of Triton X-114 extract with phosphatidylinositol-specific phospholipase C failed to expose a cross-reacting determinant to antibody detection (36), all of the GIPL species we analyzed carried an acyl-linked fatty acid on the inositol ring. On some of the GIPL species, the fatty acid was located at the 2-position of the inositol, whereas on other species it was located elsewhere on the ring (as indicated by the presence of an inositol 1,2-cyclic phosphate in the daughter ion spectra). The presence of an acyl-linked fatty acid at more than one position on the inositol ring is not without precedent. Earlier work by Ferguson (42) on the T. brucei procyclic acidic repetitive protein GPI anchor demonstrated that the acyl-linked fatty acid was located at both the 2-position (40%) and at the 3-position (60%) of the inositol ring. Ferguson (42) also suggested that acidic conditions might promote acyl migration between the 2-and 3-posi-tions since they are cis to each other on the inositol ring. We would propose that either the inositol is acylated at both the 2-position and at another position (likely the 3-position) in native C. parvum GIPLs or that the fatty acid migrated away from the 2-position during the silica gel purification of the PI.
Whereas GIPLs and GPI anchors with acyl-linked C16:0 or C18:0 fatty acids on the inositol are common (on the T. brucei procyclic acidic repetitive protein, for example) (43), we were surprised to find fatty acids of 20 -28 carbons in length at this position. To our knowledge, this is the first report of C20:0 to C28:0 acylated inositols in a GIPL. The longer chain length would certainly increase the hydrophobicity of the anchor, but it is not clear what advantage this would provide to the organism, especially in light of the fact that the lyso-acyl-GPI anchor of the C. parvum 17-kDa surface antigen does not contain any C20:0 to C28:0 fatty acids. 2 The presence of GIPLs in the membrane of C. parvum was not unexpected. Riggs and co-workers (34, 44) developed a monoclonal antibody, 18.44, that was capable of neutralizing the infectivity of C. parvum sporozoites, and then used the antibody to identify a low molecular weight antigen on the surface of both merozoites and sporozoites. An analysis of the target antigen, named CPS-500, suggested that it was a mannose-and inositol-containing glycolipid (34,41). Although they did not determine the structure of the CPS-500 antigen, they did show that antibody recognition was abolished by ␤-mannosidase treatment of the antigen but not by treatment with mixed ␣-glycosidases, ␤-glucosidase, or ␤-xylosidase. Because monoclonal antibody 18.44 recognizes one or more of the minor GIPLs present in our Triton X-114 extract, we believe that the CPS-500 antigen was included in the GIPL fraction that we analyzed. In fact, among the minor GIPL species that we characterized, we identified several that had a mannose 3 core glycan along with an additional 1-3 hexoses in a linear or FIG. 7. Exoglycosidase and HPTLC analysis of tritiated C. parvum GIPL neutral glycans. A, neutral glycans were purified from the 3to 6-kDa region of a blot after nitrous acid deamination, NaB 3 H 4 reduction, and HF treatment as described under "Materials and Methods." After an ion exchange purification step, the labeled neutral glycans were subjected to exoglycosidase cleavage with either JBAM or ASAM. Reaction products were purified by ion exchange chromatography and resolved on an HPTLC plate using a solvent system of 10:10:3 (v/v/v) chloroform, methanol, 1.0 M NH 4 OH. Lane identifications are as follows: Dextran, labeled dextran ladder; Man 3 Std., a labeled Man␣1,2-Man␣1,6-Man␣1,4-AHM standard; Man 4 Ladder, a labeled Man␣1,2-Man␣1,2-Man␣1,6-Man␣1,4-AHM standard that was partially acid-hydrolyzed; 3-6 kDa GIPL NG, untreated, labeled C. parvum neutral glycans; GIPL NG ϩ JBAM, JBAM-treated neutral glycans; GIPL NG ϩ ASAM, ASAM-treated neutral glycans. Proposed structures of bands labeled a-d as well as JBAM reaction products (aЈ, bЈ, and cЈ) and ASAM reaction products (aЉ, bЉ, and cЉ) are given in B. Hex 5 -deutero-AHM ϩ Ϫ a A branch point in the carbohydrate structure is indicated by the presence of fragment ions at m/z 828, 624, or 420 in the CID daughter ion spectrum.
b The absence of a branch point at the hexose adjacent to the deuterated AHM is indicated by the presence of a fragment ion at m/z 434. c Not applicable.
branched configuration. We were, however, unable to demonstrate the presence of a ␤-mannose in our labeled neutral glycan preparation despite repeated attempts using ␤-mannosidase from two different suppliers (data not shown). One possible explanation for this apparent discrepancy is that the neutral glycan from the CPS-500 antigen may have been at too low a concentration in our preparation to be detected by the analysis. This possibility is supported by the finding that mAb 18.44 was most reactive with the least abundant GIPL (fraction 5) in the preparation analyzed. In the future, we hope to better characterize some of the minor species since they may play an important role in immune recognition. Whereas serum IgG antibody responses to the C. parvum GIPLs were not universal among cryptosporidiosis patients, they were quite common. In the few patients we examined in detail, the anti-GIPL antibody response was dominated by the IgG 1 subclass with lesser amounts of IgG 3 and, occasionally, IgG 4 . This result contrasts with the IgG 3 subclass dominant antibody response recently reported for P. falciparum GIPLs in human patients (45). The antibody responses to the C. parvum GIPLs did not appear to depend on the particular strain of the parasite that caused the infection since they were found in patients infected with either the type 1 "human" strain or the type 2 "bovine" (zoonotic) strain (46). 2 Given that GPI anchors and GIPLs are being increasingly recognized as modulators of host immune system function, it would be interesting to determine whether the C. parvum GIPLs play some role in the development of clinical disease. Human volunteer studies have suggested previously (20) that individuals with pre-existing serum antibodies to the low molecular mass 17-and 27-kDa C. parvum surface protein antigens are more likely to develop asymptomatic infections when challenged, whereas those volunteers with no pre-existing antibody are more likely to develop overt cryptosporidiosis. Similarly, recent work (25-27) on P. falciparum has shown that an IgG response to the GIPLs correlates with protection from severe cerebral malaria symptoms. We are currently working to collect the samples and symptom information necessary to address the role (if any) of GIPLs in diarrheal illness during C. parvum infection.