Structural Characterization of NETNES, a Novel Glycoconjugate in Trypanosoma cruzi Epimastigotes*

The unicellular stercorarian protozoan parasite Trypanosoma cruzi is the etiological agent of Chagas' disease. The epimastigote form of the parasite is covered in a dense coat of glycoinositol phospholipids and short glycosylphosphatidylinositol (GPI)-anchored mucinlike molecules. Here, we describe the purification and structural characterization of NETNES, a relatively minor but unusually complex glycoprotein that coexists with these major surface components. The mature glycoprotein is only 13 amino acids in length, with the sequence AQENETNESGSID, and exists in two forms with either four or five post-translational modifications. These are either one or two asparagine-linked oligomannose glycans, two linear α-mannose glycans linked to serine residues via phosphodiester linkages, and a GPI membrane anchor attached to the C-terminal aspartic acid residue. The variety and density of post-translational modifications on an unusually small peptide core make NETNES a unique type of glycoprotein. The N-glycans are predominantly Manα1–6(Manα1–3) Manα1–6(Manα1–3)Manβ1–4GlcNAcβ1–4GlcNAcβ1-Asn; the phosphate-linked glycans are a mixture of (Manα1–2)0–3Man1-P-Ser; and the GPI anchor has the structure Manα1–2(ethanolamine phosphate)Manα1–2Manα1–6Manα1–4(2-aminoethylphosphonate-6)GlcNα1–6-myo-inositol-1-P-3(sn-1-O-(C16:0)alkyl-2-O-(C16:0)acylglycerol). Four putative NETNES genes were found in the T. cruzi genome data base. These genes are predicted to encode 65-amino acid proteins with cleavable 26-amino acid N-terminal signal peptides and 26-amino acid C-terminal GPI addition signal peptides.

ted to mammals, including humans, via the reduviid bug insect vector. The parasite undergoes biochemical and morphological changes during its complex life cycle. Within the midgut of the insect vector, the parasite multiplies as a non-infective epimastigote form that differentiates into an infectious metacyclic trypomastigote form upon migration to the hindgut. It is this non-dividing form that is transmitted to the mammalian host through fecal contaminative infection of the blood meal skin lesion. The metacyclic trypomastigote forms invade host cells, where they differentiate into small round amastigote forms that divide in the cytoplasm. Some amastigotes differentiate into non-dividing bloodstream trypomastigote forms that, once liberated by host cell lysis, propagate the infection by invading new host cells. These same trypomastigotes can also be ingested by an insect vector during a blood meal and subsequently differentiate into epimastigote forms in the insect midgut, thus completing the life cycle.
In this study, we have purified and determined the complete structure of a novel low abundance GPI-anchored glycoprotein from epimastigote stage cells of T. cruzi. The glycoprotein (NETNES) is a 13-amino acid peptide with up to five posttranslational modifications, including one or two N-linked glycans, two phosphate-linked mannose chains, and a GPI anchor. The NETNES glycoprotein is one of the most unusual and complex molecules yet characterized in any eukaryote.

EXPERIMENTAL PROCEDURES
Cell Culture-Epimastigotes of the T. cruzi CL-Brener strain were grown at 28°C in liver infusion tryptose medium (24) supplemented with 10% heat-inactivated fetal calf serum (PAA Laboratories).
Purification and Hydrophobic Chromatography of Glycoconjugates-Glycoconjugates were extracted from cultures of ϳ2 ϫ 10 11 cells using organic solvents and purified on an octyl-Sepharose column (100 ϫ 5 mm) as described previously (10). The fractions containing GPI-anchored mucins and NETNES were combined and purified twice more by octyl-Sepharose chromatography (the final time using a gradient of 10 -45% propan-1-ol over 60 ml) to achieve further separation. Sample purity was assessed by SDS-PAGE (Invitrogen precast 4 -12% gel) and periodate-Schiff staining.
Periodate-Schiff Staining of SDS-Polyacrylamide Gels-After electrophoresis, gels were rinsed with water, incubated with a fixing solution of acetic acid/methanol/water (10:35:25) for 15 min, and washed three times for 5 min with water. This was followed by incubation with 1% sodium periodate in 3% acetic acid for 30 min. The water washes were repeated with subsequent incubations for 1 h with Schiff's reagent (Sigma) and for 30 min with a reducing solution of 1% sodium metabisulfite and three additional water washes. Development of glycoconjugate staining improved overnight.
Mild Acid Hydrolysis-Dried samples were treated with 50 l of 40 mM trifluoroacetic acid for 10 min at 100°C. After hydrolysis, samples were dried in a SpeedVac concentrator.
Aqueous Hydrogen Fluoride Dephosphorylation-Dried samples were treated with 50 l of 48% aqueous HF for Ͼ48 h at 0°C before drying in a SpeedVac concentrator to remove the aqueous HF. Samples were then dried twice from 50 l of water to remove residual HF.
Partial Acetolysis-Dried radiolabeled samples were acetylated with 50 l of pyridine/acetic anhydride (1:1) at 100°C for 30 min in a heating block and then dried under a nitrogen stream. The samples were incubated with 100 l of acetic anhydride/acetic acid/sulfuric acid (10: 10:1) at 37°C for 6 h in a heating block. The reaction was quenched and neutralized by the addition of 75 l of pyridine and 500 l of water and incubation at room temperature for 30 min. The acetylated acetolysis products were recovered by partitioning into 250 l of chloroform. After separation, the upper aqueous phase was removed, and the remaining chloroform phase was washed three times with 500 l of water before being dried under a nitrogen stream. The samples were incubated with 200 l of methanol and 35% aqueous ammonia (1:1) at 37°C for Ͼ16 h in a heating block to de-O-acetylate the products and dried under a nitrogen stream.
Jack Bean ␣-Mannosidase Digestion-Dried samples were treated with 750 milliunits of dialyzed jack bean ␣-mannosidase (Roche Applied Science) in 20 l of 0.1 M sodium acetate (pH 5.0) at 37°C for Ͼ16 h and subsequently boiled in a heating block at 100°C for 5 min.
Aspergillus saitoi ␣-Mannosidase Digestion-Dried samples were treated with 10 microunits of A. saitoi ␣-mannosidase (Oxford Glycosystems) in 10 l of 20 mM sodium acetate (pH 5.0) at 37°C for Ͼ16 h and subsequently boiled in a heating block at 100°C for 5 min.
Endoglycosidase H Digestion-Dried samples were boiled at 100°C for 2 min in 10 l of 0.04% SDS and allowed to cool before the addition of 5 l of 100 mM sodium acetate (pH 5.0) and 25 milliunits of endoglycosidase H (Roche Applied Science) and subsequent incubation at 37°C for Ͼ16 h. Samples were then dried in a SpeedVac concentrator.
Peptide N-Glycosidase F (PNGase F) Digestion-Dried samples were boiled at 100°C for 10 min in 8 l of water before 1 l of 0.25 M sodium phosphate (pH 7.5) and 500 units of PNGase F (New England Biolabs Inc.) were added, followed by incubation at 37°C for Ͼ16 h. An additional 250 units of PNGase F were added, and incubation at 37°C was continued for 3 days. Samples were then dried in a SpeedVac concentrator. Note that exhaustive digestion with PNGase F was required to quantitatively de-N-glycosylate NETNES.
Matrix-assisted Laser Desorption Ionization Time-of-Flight (MALDI-TOF) Mass Spectrometry-Negative ion MALDI-TOF mass spectra were acquired on an ABI Voyager DE-STR instrument in linear mode using 2,5-dihydroxybenzoic acid as the matrix. 0.5-l aliquots (ϳ100 pmol) of native/trifluoroacetic acid-treated/aqueous HF-treated NETNES were mixed 1:1 with 2,5-dihydroxybenzoic acid. Spectrometry was performed with an extraction delay time of 300 ns and a laser intensity of 2861 V for the native sample, 200 ns and 2761 V for the dephosphorylated sample, and 300 ns and 2911 V for the mild acidtreated sample.
Nitrous Acid Deamination and Radioactive Reduction of the GPI Neutral Glycans and Recovery of the Phosphatidylinositol (PI) Moieties-This procedure is an adaptation of a process outlined previously (10). An aliquot (2.5 nmol) of NETNES was freeze-dried, resuspended in 50 l of water, washed three times with 100 l of butan-1-ol saturated with water, and freeze-dried once more. The sample was dissolved in 15 l of 0.3 M sodium acetate (pH 4.0), and 7.5 l of 1 M sodium nitrite were added and incubated for 1 h at room temperature. An additional 15 l of 0.3 M sodium acetate (pH 4.0) and 7.5 l of sodium nitrite were added and incubated for 2 h at 37°C.
The released PI moieties were recovered by three extractions with 100 l of butan-1-ol saturated with water. The butan-1-ol extracts were pooled, dried in a SpeedVac concentrator, resuspended in 50 l of chloroform/methanol (2:3), and analyzed by negative ion electrospray mass spectrometry (ES-MS) on a Micromass Quattro Ultima triple quadrupole instrument as described below.
The deaminated molecules remaining in the aqueous phase were reduced with NaB 3 H 4 as described previously (25); the reaction was stopped by the addition of 20-l aliquots of 1 M acetic acid; and the product was freeze-dried. The samples were dissolved in 50 l of water, dialyzed in a microdialyzer for Ͼ3 h, and freeze-dried. The deaminated and reduced samples were dephosphorylated with aqueous HF as described above, freeze-dried, re-N-acetylated with 100 l of 1 M sodium hydrogen carbonate and three 2.5-l aliquots of acetic anhydride, incubated twice at 0°C for 10 min and finally at room temperature for Ͼ1 h. Samples were desalted, and boric acid and acetic acid were removed as described previously (25). The GPI neutral glycans were purified from radiochemical contaminants by downward paper chromatography as described previously (25), but without the high voltage paper electrophoresis step; dried by rotary evaporation; and dissolved in 100 l of water. The 3 H-labeled glycans (ϳ35,000 cpm) were analyzed by Dionex high performance anion exchange chromatography, and the elution positions of the radiolabeled glycans were expressed in Dionex units by linear interpolation of the elution position between adjacent glucose oligomer internal standards (25).

ES-MS and Electrospray Tandem Mass Spectrometry (ES-MS/ MS)-Electrospray
mass spectra were recorded on either the Micromass Quattro Ultima triple quadrupole instrument or a Micromass Q-ToF2 instrument. Samples were introduced into the mass spectrometers using nanospray tips. In positive ion mode, the capillary and cone voltages were 0.7-1.2 kV and 25-35 V, respectively. Daughter ion ES-MS/MS spectra were recorded using collision voltages of 35-40 V for most samples, but 35-60 V for permethylated glycan samples. In negative ion mode, the capillary and cone voltages were 0.9 -1.1 kV and 40 -60 V, respectively, and collision voltages were 40 -60 V. In all cases, the collision gas was argon at 3 ϫ 10 Ϫ3 torr. All data were collected and processed with MassLynx software.
Edman Sequencing-Aliquots (50 pmol) of NETNES were treated with either aqueous HF or PNGase F (as described above) or were left untreated. Glycopeptides were adsorbed onto polyvinylidene difluoride membrane, washed with 0.1% trifluoroacetic acid in water, and sequenced on an Applied Biosystems 494C Protein Sequencer. The PNGase F-treated samples were also sequenced after coupling to the solid phase support system Sequelon AA (product no. GEN920033, Applied Biosystems), as it was noticed that the glycopeptide washed out of the Sequencer after removal of the N-linked glycans.
ES-MS/MS Sequencing-An aliquot (100 pmol) of NETNES was dephosphorylated with aqueous HF before being treated with PNGase F to remove N-glycans. The sample was dried in a SpeedVac concentrator and redissolved in 20 l of 1% formic acid before being loaded onto a C18 StageTip (Proxeon Biosystems) that had been preconditioned with 20 l of 50% acetonitrile and 1% formic acid and equilibrated in 20 l of 1% formic acid. The tip was washed twice with 20 l of 1% formic acid, and the peptide was eluted with 2 l of 50% acetonitrile and 1% formic acid. The peptide was sequenced by collisioninduced fragmentation on the Micromass Q-ToF2 instrument in positive ion mode as described above.
BLAST Search and Post-translational Modification Prediction-tBLASTn searches were carried out using the data base for predicted T. cruzi open reading frames Ͼ50 amino acids on the TcruziDB web site (available at tcruzidb.org/). Predictions for signal sequences, GPI anchor modification, subcellular localization, and N-and O-glycosylation were carried out using the algorithms available on the web sites for Analysis of N-Linked Glycans by Permethylation and ES-MS-An aliquot (400 pmol) of NETNES was treated with PNGase F or endoglycosidase H as described above in a 2-ml glass vial, dried, and permethylated by an adaptation of a procedure described previously (25), where 50 l of Me 2 SO (AnalaR, VWR International) were added to the dried sample and incubated at room temperature with gentle agitation for 10 min. 50 l of a 120 mg/ml slurry of sodium hydroxide (Aristar, VWR International) in Me 2 SO were added to the sample and gently agitated for an additional 20 min. Following this, the sample was methylated by three additions of methyl iodide (Aldrich), two of 10 l and one of 20 l, each followed by incubation with occasional gentle agitation at room temperature for 10, 10, and 20 min, respectively. 250 l of chloroform were added, followed by 1 ml of fresh sodium thiosulfate, with vortexing, and the samples were then allowed to settle into two phases. After removal of the upper phase, the lower phase was washed five times with 1 ml of water. The lower phase was dried, and the residue was dissolved in 25 l of 80% acetonitrile. 2 l were removed to a microcentrifuge tube, and 2 l of 80% acetonitrile and 1 mM sodium acetate were added to give a final concentration of 80% acetonitrile in 0.5 mM sodium acetate. This was then analyzed on the Micromass Q-ToF2 instrument in positive ion mode as described above.
Release and Purification of P-Linked Glycans-An aliquot (10 nmol) of NETNES was hydrolyzed with mild acid as described above using 0.5 ml of 40 mM trifluoroacetic acid. The sample was dried and redissolved in 500 l of 10% propan-1-ol and 100 mM ammonium acetate. The glycans released through this reaction were purified from the remaining glycoprotein by hydrophobic interaction chromatography using a 7 ϫ 18-mm octyl-Sepharose column pre-equilibrated in 10% propan-1-ol and 100 mM ammonium acetate. The released glycans and glycoprotein were eluted with a 30-ml gradient of 10 -60% propan-1-ol. The fractions containing released glycans (flow-through fractions) and glycoprotein were detected using orcinol staining of 0.5% aliquots spotted onto Silica Gel 60 plates. An aliquot (2%) was subjected to monosaccharide analysis as described above. The remainder of the released glycans and the glycoprotein sample were dried under a nitrogen stream. The glycoprotein sample was resuspended in 50 l of 40% propan-1-ol, and 1% of this was analyzed by MALDI-TOF mass spectrometry as described above. The released P-glycans were dissolved in 50 l of water.
Radiolabeling and Microsequencing of Released P-Glycans-An aliquot (20%) of the released P-glycan sample was radiolabeled through NaB 3 H 4 reduction as described above alongside a standard of 4 nmol of Man␣1-3Man mannobiose (Dextra Laboratories). The radiolabeled samples were dried and resuspended in 500 l of water. Aliquots (10,000 cpm) of the released radiolabeled P-glycans and mannobiose were dried and microsequenced as described for the GPI glycan core, but using only jack bean ␣-mannosidase and A. saitoi ␣-mannosidase digests.
Permethylation, ES-MS, and GC-MS Linkage Analyses of the Released P-Glycans-The remaining 80% of the non-radiolabeled sample 2 R. Gupta, E. Jung, and S. Brunak, manuscript in preparation. was deuteroreduced using 200 l of 0.5 M NaB 2 H 4 at 4°C for Ͼ16 h. Excess reductant was destroyed with acetic acid, and the sample was desalted by passage through 0.4 ml of AG-50 Hϩ, drying, and evaporation with methanol. Deuteroreduced samples were mixed with 20% (ϳ36,000 cpm) of the tritium-reduced material (see above) to act as tracer, dried, and permethylated as described above. A sample (1%) of the permethylated mixture was dried and dissolved in 80% acetonitrile and 0.5 M sodium acetate for positive ion ES-MS and ES-MS/MS analyses on the Micromass Quattro Ultima triple quadrupole instrument as described above. The remainder of the chloroform phase of the permethylated released P-glycan sample was dried under a nitrogen stream, dissolved in 20% acetonitrile, and separated by HPLC on a Hichrom Kromasil 100-5 C18 column (150 ϫ 3.2 mm) using a gradient of 20 -65% acetonitrile in water. The radioactivity of the fractions was measured in a Beckman LS6000SE scintillation counter, and three peaks of radioactivity (corresponding to di-, tri-, and tetrasaccharitols) were pooled. Aliquots (1%) of the samples were analyzed directly by positive ion ES-MS on the Micromass Q-ToF2 instrument. The remainder of the samples were subjected to the rest of the methylation linkage analysis protocol as described previously (25). GC-MS was performed on the Hewlett-Packard 6890-5973 system using the Agilent Technologies HP5 column (30 m ϫ 0.25 mm).

NETNES Is a Highly Glycosylated GPI-anchored
Glycoprotein-While purifying mucins and GIPLs from solvent extracts of the CL-Brener strain of T. cruzi epimastigotes, we noticed that these cells contained additional carbohydrate-containing glycoconjugates that migrated as a doublet upon SDS-PAGE between the mucins and the GIPLs. Hydrophobic interaction chromatography on octyl-Sepharose indicated that these bands with apparent molecular masses of 13 and 15 kDa (that we now call NETNES) could be partially resolved from the mucins and the GIPLs (Fig. 1A). Rechromatography of the NETNES-enriched fractions twice more using shallower gradients resulted in a fraction highly enriched in NETNES as judged by SDS-PAGE and periodate-Schiff staining for carbohydrate (Fig. 1B,  lane 1). To identify the true mass of NETNES, this fraction was analyzed by negative ion MALDI-TOF mass spectrometry, which showed that the molecule was present as two polydisperse forms with principal components of 5.6 and 6.8 kDa (Fig.  2A). These principal components were flanked by other ions differing in mass by multiples of 162 Da, suggesting glycosyl microheterogeneity. Analysis for non-N-acetylated glucosamine (a hallmark of GPI anchors) by conversion to 2,5-anhydromannitol (AHM) and quantification by selected ion moni- toring GC-MS (Fig. 3A) (14) suggested that GPI-anchored molecules were present in the fraction. Monosaccharide analysis by GC-MS identified mannose, galactose, and N-acetylglucosamine in the sample, and the two analyses suggested a Man:Gal:GlcNAc:GlcN ratio of ϳ21:3:3:1, where some or all of the Gal content may be due to the traces of Gal-rich mucins in the fraction (Fig. 1B). Assuming one GlcN/molecule, these data suggested that carbohydrate alone could account for ϳ4.2 kDa (75 and 62% by mass, respectively) of the 5.6-and 6.8-kDa glycoconjugates. The presence of stoichiometric amounts of myo-inositol, as determined by GC-MS (25), supported the notion that these glycoconjugates were GPI-anchored. Triplicate analysis of the fraction for GlcN content suggested a yield of (3.0 Ϯ 0.4) ϫ 10 4 molecules/parasite. This is a minimum figure because it does not take into account losses during purification.
GPI Lipid Structure-A sample of purified NETNES (2.5 nmol based on GlcN content) was subjected to nitrous acid deamination and extracted with butan-1-ol. The extracts, which contained the released PI moieties, were analyzed by negative ion ES-MS (Fig. 4A) and parent ion scanning ES-MS/MS for parents of m/z 241, a PI-specific [inositol-1,2-cyclic-P] Ϫ daughter ion (Fig. 4B). The latter identified the ion at m/z 795.3 as the only one belonging to a PI species. The collisioninduced dissociation daughter ion spectrum of the ion at m/z 795 (Fig. 4C) Fig. 4D.
GPI Glycan Structure-To determine the structure of the GPI glycan core, the aqueous phase remaining after butan-1-ol extraction of deaminated NETNES was reduced with NaB 3 H 4 and dephosphorylated with aqueous HF. The resulting GPI neutral glycans, which terminated in AHM*, were purified by Dionex high performance anion exchange chromatography  (data not shown) and eluted as one component with a chromatographic value of 3.0 Dionex units. This value corresponds to that of authentic Man 4 -AHM* (25). This sequence was confirmed by exoglycosidase digestion and partial acetolysis, followed by HPTLC (Fig. 5). Digestion with Man␣1-2Man-specific A. saitoi ␣-mannosidase yielded a major product that comigrated with Man 2 -AHM* (Fig. 5A, lane 2). Partial acetolysis, which is selective for Man␣1-6Man linkages, yielded a major product that comigrated with Man 1 -AHM* (Fig. 5A, lane 3). Digestion with jack bean ␣-mannosidase, which cleaves all nonreducing terminal ␣-mannose residues, yielded a major product that comigrated with AHM* (Fig. 5A, lane 4). The microsequencing results are summarized in Fig. 5B.
NETNES Contains One or Two N-Linked Glycans-When NETNES was dephosphorylated with aqueous HF for 48 h,

TABLE I Amino acid sequence of NETNES deduced by ES-MS/MS
and Edman sequencing The amino acid sequences were obtained by ES-MS/MS of the NET-NES peptide prepared by aqueous HF dephosphorylation and PNGase F digestion, Edman sequencing of the native NETNES peptide, Edman sequencing of NETNES after PNGase F digestion, and Edman sequencing of NETNES after aqueous HF dephosphorylation. X represents the absence of an amino acid in the Edman sequence due to substitution. Lower yield residues are underlined. The asterisk indicates the detection of equal amounts of Asn and Asp due to partial N-glycosylation of this residue.

Sequencing method
Amino acid sequence AQE NETXESGSID Deduced native sequence AQENETNESGSID MALDI-TOF analysis revealed a change from two polydisperse peaks with principal components of 5.6 and 6.8 kDa to two principal components of 2.7 and 3.9 kDa (Fig. 2B). This reduction in mass and heterogeneity was due to the removal of the GPI anchor and two phosphate-linked glycans (described below). The continuity of a mass difference of ϳ1.2 kDa between the principal components of NETNES before and after dephosphorylation (Fig. 2, A and B) led to the hypothesis that aqueous HF-stable N-linked glycans might be present in the glycoprotein. This was tested by digestion with the N-linked glycanspecific enzyme endoglycosidase H, followed by SDS-PAGE and periodate-Schiff staining. Digestion with endoglycosidase H, which cleaves between the two GlcNAc residues of oligomannose N-linked glycans, caused the two bands to collapse into one lower mass band of ϳ9 kDa (Fig. 1B, lanes 1 and 2), aliquot (400 pmol) of NETNES was exhaustively digested with PNGase F and the products were permethylated. Analysis by ES-MS revealed a major ion at m/z 801.4 (Fig. 6B), which was assigned as the permethylated pseudomolecular ion [Hex 5 HexNAc 2 ϩ 2Na] 2ϩ following collision-induced fragmentation (Fig. 6C). Interpretation of the daughter ions (Fig. 6D) supports the branched Man 5 GlcNAc 2 structure suggested by the specificity of endoglycosidase H, i.e. Man␣1-6(Man␣1-3)Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc. This structural assignment was also supported by endoglycosidase H digestion and permethylation, ES-MS, and ES-MS/MS analyses. In this case, an ion at m/z 678.8 was observed, which was shown to correlate with the permethylated pseudomolecular ion [Hex 5 HexNAc 1 ϩ 2Na] 2ϩ following collision-induced fragmentation (data not shown). We conclude that NETNES can therefore contain one or two conventional Man 5 GlcNAc 2 Nlinked glycans. In addition to the predominant Man 5 GlcNAc 2 structure, minor ions at m/z 699, 903, and 1005 (Fig. 6B) suggest that there are also traces of Man 4 GlcNAc 2 , Man 6 GlcNAc 2 , and Man 7 GlcNAc 2 structures in NETNES that will contribute to the glycosyl heterogeneity of the native molecules ( Fig. 2A).
Sequencing of the Peptide Chain-To deduce the amino acid sequence of the glycoprotein, an aliquot (100 pmol) of NETNES was dephosphorylated with aqueous HF before treatment with PNGase F to remove all phosphate-linked glycans, N-glycans, and the GPI anchor. The remaining peptide was purified by C18 reverse-phase chromatography and analyzed by ES-MS, which revealed a major doubly charged [M ϩ 2H] 2ϩ ion at m/z 719.8, its [M ϩ H ϩ Na] 2ϩ sodium adduct at m/z 730.8, and an [M ϩ 2H Ϫ H 2 O] 2ϩ dehydration product at m/z 710.8 (Fig. 7A). Collision-induced fragmentation of the [M ϩ 2H] 2ϩ ion gave a daughter ion spectrum that was readily interpreted as belonging to the amino acid sequence AQEDETDESGSID-ethanolamine (Fig. 7B). N-Glycosylation can occur at asparagine resi-dues within Asn-X-(Ser/Thr) sequences (where X is any amino acid except proline). Since the action of PNGase F converts asparagine to aspartic acid, two possible sites of N-glycosylation are revealed by the sequences DET and DES. We therefore deduce the original amino acid sequence to have been AQENETNESGSID-ethanolamine. The sequence also has three possible sites of phosphorylation (one threonine and two serine residues), as discussed below. GPI anchors are invariably attached to proteins through either EtNP or AEP. This spectrum indicates the presence of an amide linkage to ethanolamine and therefore shows that the protein is linked to the GPI anchor through EtNP.
To confirm that the aspartic acids in DET and DES were the result of cleavage of N-glycans from asparagine residues, Edman sequencing was performed on NETNES before and after PNGase F digestion (Table I). Sequencing of the native sample gave the sequence AQENETXEXGXID, where the abundance of the underlined asparagine was significantly lower compared with the other residues. The Xs indicate substituted residues that could not be detected by Edman sequencing. Upon PNGase F treatment, the first X residue was detected as Asp and was therefore originally N-glycosylated Asn. Moreover, after PNGase F digestion, the first (weak) asparagine detected in the native sample was detected as a mixture of Asn and Asp, indicating that it is this residue that is unoccupied in the singly N-glycosylated form of NETNES. Following aqueous HF dephosphorylation, the second and third X residues appeared as Ser residues by Edman sequencing (Table I), indicating that both serines (but not the threonine) are substituted with phosphate or sugar phosphate.
NETNES Contains Two P-Linked Glycans-Since Edman sequencing suggested that the two serine residues of NETNES contain phosphoryl substituents (Table I), we investigated the nature of these modifications. A sample (10 nmol) of NETNES was hydrolyzed with mild acid, and MALDI-TOF analysis re-  The P-linked glycans were released by mild acid, reduced, permethylated, and separated by reverse-phase HPLC. The resolved di-, tri-, and tetrasaccharitols were processed for GC-MS methylation linkage analysis. The presence (ϩ) or absence (Ϫ) of particular residues is indicated. PMAA, partial methylated alditol acetate. vealed a change from two polydisperse peaks with principal components of 5.6 and 6.8 kDa ( Fig. 2A) to two principal components of 4.6 and 5.8 kDa (Fig. 2C). Mild acid hydrolysis cleaves sugar 1-phosphate bonds and so the mass loss for each form of NETNES of ϳ970 Da suggests the loss of six hexoses (972 Da), originally attached to the serine residues via phosphodiester linkages. To investigate this further, the saccharide chains released through the hydrolysis reaction were purified from the remaining glycoprotein by octyl-Sepharose chromatography. Monosaccharide analysis of the released glycans (recovered in the flow-through fractions) revealed that the saccharide chains consisted of only mannose. 20% of the released glycans were radiolabeled by NaB 3 H 4 reduction and microsequenced by HPTLC with jack bean ␣-mannosidase and A. saitoi ␣-mannosidase digests alongside a Man␣1-3-[1-3 H]mannitol standard (Fig. 8). The untreated sample contained di-, tri-, and tetrasaccharitols, with the trisaccharitol being the most abundant, followed by the disaccharitol and then the tetrasaccharitol (Fig. 8, lane 1). Jack bean ␣-mannosidase digestion yielded a single band that comigrated with [1-3 H]mannitol (Fig. 8, lane 2), suggesting that all the anomeric linkages in the saccharitol chains are ␣. Digestion with A. saitoi ␣-mannosidase (specific to Man␣1-2Man linkages) yielded a single band that migrated close to, but not coincident with, Man␣1-3-[1-3 H]mannitol (Fig. 8, compare  lanes 3 and 4).
The remaining 80% of the saccharide chains were deuteroreduced, added to NaB 3 H 4 -reduced saccharide chains (to act as a radiochemical tracer), and permethylated. Analysis of a small sample by positive ion ES-MS revealed four major ions at m/z 290.2, 494.2, 698.4, and 902.5 (Fig. 9A), which were shown to represent the [M ϩ Na] ϩ pseudomolecular ions of [1-2 H]hexitol and of deuteroreduced linear di-, tri-, and tetrasaccharitols, respectively (Fig. 9, B-D). The permethylated deuteroreduced linear di-, tri-, and tetrasaccharitols were then separated by HPLC using a C18 column and processed for methylation linkage analysis by GC-MS. The analyses revealed that the interresidue linkages are exclusively Man1-2Man and Man1-2mannitol (Table II), in agreement with the result of the Man␣1-2Man-specific A. saitoi ␣-mannosidase digestion (Fig.  8, lane 3).
Taken together, these data support the conclusion that the P-linked glycans attached to Ser are a mixture of Man, Man␣1-2Man, Man␣1-2Man␣1-2Man, and Man␣1-2Man␣1-2Man␣1-2Man. The presence of four possible lengths of mannose chain at two sites accounts for much of the glycosyl heterogeneity of the native NETNES molecules observed by MALDI-TOF mass spectrometry ( Fig. 2A). Whether there is an even or skewed distribution of chain lengths between the two Ser sites is not known.
The GlcN Residue of the GPI Glycan Is Substituted with AEP-Glucosamine analysis of NETNES from 2.8 ϫ 10 9 cell eq yielded a significant amount of the deamination product AHM (Fig. 3A). However, without an aqueous HF dephosphorylation step, the AHM produced was negligible (Fig. 3B), indicating the presence of a phosphoryl or phosphonyl substituent on the glucosamine residue of the GPI anchor glycan core.
The MALDI-TOF spectrum of NETNES after mild acid hydrolysis provides the average mass of the [M Ϫ H] Ϫ ion for the entire NETNES glycoprotein minus the heterogeneous P-linked glycans. Thus, the average (nominal) molecular mass of mono-N-glycosylated NETNES, after removal of the P-linked saccharides, is 4587 Ϯ 2 Da (Fig. 2C). Most of this can be accounted for by the combined average molecular masses (4482 Da) of the AQENETNESGSID-ethanolamine component (1436.3 Da), the phosphate bridge to the GPI anchor (63.0 Da), the glycan core (809.7 Da), the phosphatidylinositol (796.1 Da), the N-glycan (1217.1 Da), and the two serine-linked phosphate groups (160.0 Da). The remaining unexplained mass (105 Ϯ 2 Da) is consistent with the presence of an AEP substituent (107 Da). Taken together with the evidence for a phosphoryl or phosphonyl substituent on the GlcN residue, we interpret these data in terms of the presence of an AEP group at C-6 of the GPI anchor GlcN residue, as is commonly seen in T. cruzi GPI structures (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)33).
Identification of Putative NETNES Genes-A BLAST search of the T. cruzi genome data base with the experimentally determined AQENETNESGSID amino acid sequence revealed four putative NETNES genes, each encoding a 65-amino acid open reading frame with very similar sequences (Fig. 10). The post-translational modification site prediction algorithms Sig-nalP, DGPI, big-PI Predictor, NetNGlyc, and NetOGlyc predicted that all four open reading frames have a potential cleavable N-terminal signal sequences (with a predicted mature N terminus at Gln 27 ), that they are likely to be GPI-anchored (with a possible cleavage site after Ser 35 or Asp 39 (DGPI) or Ala 43 (big-PI Predictor)), that there are two potential N-glycosylation sites (at Asn 30 and Asn 33 ), and that there are no conventional (mucin-type) O-glycosylation sites. These predictions are in good agreement with our experimental data, except that the mature N terminus is Ala 26 rather than Gln 27 and that the mature C terminus (Asp 39 ) was not the preferred prediction of either GPI attachment algorithm. DISCUSSION Based on the data described in this study, we present a chemical structure for the NETNES glycoprotein from the epimastigote form of T. cruzi (Fig. 11). The structure is largely complete, but does contain four assumed structural assignments. The first assumption, that the AEP substituent is attached to the 6-position of the GPI anchor GlcN residue, is based on precedent from T. cruzi GIPLs and GPI-anchored mucins (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). The second and third assumptions, that the GlcN-to-myo-inositol glycosidic linkage in the GPI anchor is ␣1-6 and that the EtNP bridge from the C-terminal Asp residue is linked to the 6-position of the third ␣-Man residue of the GPI anchor, are based on precedent for all eukaryote GPI anchors analyzed in sufficient detail (reviewed in Refs. 34 -36). The fourth assumption, that the Man␣1-P-Ser phosphodiester linkages of the P-linked glycans are in the ␣-configuration, is based on precedent for Leishmania phosphoglycans, where Man␣1-P is transferred en bloc from GDP-␣-Man via mannosyl-phosphate transferase (37).
The co-extraction of NETNES with the mucins and GIPLs may explain why the pioneering structural work on the GIPLs reported the presence of amino acids (38,39). Indeed, the GIPLs were originally named lipopeptidophosphoglycan for this reason. Significantly, the reported amino acids in Y strain lipopeptidophosphoglycan are Gly, Ser, Glx, Ala, Asx, and Thr (38), whereas those of CL-Brener strain NETNES are Gly, Ser, Glx, Ala, Asx, Thr, and Ile.
The NETNES glycoprotein contains an unusually small mature peptide component (13 amino acids). Prior to the discovery of NETNES, the smallest T. cruzi cell-surface glycoproteins known were the trypomastigote-specific surface antigen, a 42amino acid mucin (40), and the 55-60-amino acid epimastigote TcMUG S mucins (16 -18), all of which are GPI-anchored and galactose-rich due to being multiply O-glycosylated.
NETNES is unique in having a combination of N-linked glycans, P-linked glycans, and a GPI membrane anchor. With up to five post-translational modifications on 13 amino acids, it is arguably one of the most complex glycoconjugates yet described. Nevertheless, it has features in common with certain other atypical glycoproteins. The most similar in terms of size is human CD52, a 12-amino acid GPI-anchored glycoprotein with a single N-glycosylation site occupied by a complex Nlinked glycan (41,42). On the other hand, the presence of P-linked glycans, a feature thus far described only in protozoa and Dictyostelium, clearly distinguishes NETNES from all mammalian glycoproteins. In this regard, NETNES, which contains two (Man␣1-2) 0 -3 Man1-P-Ser glycans, has similarity to FIG. 11. Proposed primary structure of the NETNES glycoprotein isolated from the epimastigote form of the CL-Brener strain of T. cruzi. EtN, ethanolamine. other trypanosomatid glycoproteins. The secreted acid phosphatase SAP1 and the promastigote proteophosphoglycan pPPG2 of Leishmania mexicana are large glycoproteins that contain (Man␣1-2) 0 -5 Man1-P-Ser structures (43,44) as well as larger repeating Gal␤1-4Man1-P-Ser-linked phosphosaccharide structures that are common to a number of other Leishmania proteophosphoglycans (reviewed in Ref. 45). Other examples of P-linked glycans include T. cruzi epimastigote gp72, a transmembrane glycoprotein with glycans containing Gal, Galf, Rha, Fuc, and Xyl P-linked to Ser and Thr residues (20,23,46), and the glutamate-and alanine-rich protein GARP from the procyclic and epimastigote forms of the African trypanosome Trypanosoma congolense. GARP is a 207-amino acid GPI-anchored protein with two large Man 11 Gal 7 -P-Ser glycans of unknown structure (47). Other, non-trypanosomatid examples of P-linked glycans are the GPI-anchored proteophosphoglycan of Entamoeba histolytica, which contains (Glc␣1-6) 0 -23 Gal␤1-6Gal1-P-Ser structures (48), and the GlcNAc␣1-P-Ser and Fuc␤1-P-Ser structures in Dictyostelium glycoproteins (49 -51).
Our current image of the cell-surface molecular architecture of the epimastigote form of T. cruzi is that of a carpet of densely packed GIPLs with numerous multiply O-glycosylated short mucins extending beyond this carpet (15,29,52), interspersed with less abundant cell-surface molecules like members of the trans-sialidase superfamily and gp72. To this latter group of less abundant molecules we must now add NETNES, which is present only at ϳ30,000 copies/cell. In comparison with the other surface molecules, NETNES is notably devoid of Gal (both Galp and Galf) and, consequently, sialic acid. Unlike the abundant mucins, NETNES is noticeably rich in terminal ␣-Man residues.
Although we have no functional data on NETNES, its potential for interacting with possible insect vector ␣-Man-specific lectins may be significant. Gene knockout studies to assess the function of NETNES are feasible, but will be complicated by the presence of multiple NETNES genes (although two are in tandem array). Other issues that need to be addressed are its life cycle stage specificity, an assessment of its immunogenicity, and whether it is recognized by human chagasic sera. The distribution of NETNES between different strains of T. cruzi also needs to be addressed. Thus far, preliminary periodate-Schiff staining of SDS-polyacrylamide gels of mucin/GIPL extracts suggests that a similar molecule is also present in the Y strain, although it is probably less abundant than in the CL-Brener strain.