Unique Asn-linked Oligosaccharides of the Human Pathogen Entamoeba histolytica*

N-Glycans of Entamoeba histolytica, the protist that causes amebic dysentery and liver abscess, are of great interest for multiple reasons. E. histolytica makes an unusual truncated N-glycan precursor (Man5GlcNAc2), has few nucleotide sugar transporters, and has a surface that is capped by the lectin concanavalin A. Here, biochemical and mass spectrometric methods were used to examine N-glycan biosynthesis and the final N-glycans of E. histolytica with the following conclusions. Unprocessed Man5GlcNAc2, which is the most abundant E. histolytica N-glycan, is aggregated into caps on the surface of E. histolytica by the N-glycan-specific, anti-retroviral lectin cyanovirin-N. Glc1Man5GlcNAc2, which is made by a UDP-Glc: glycoprotein glucosyltransferase that is part of a conserved N-glycan-dependent endoplasmic reticulum quality control system for protein folding, is also present in mature N-glycans. A swainsonine-sensitive α-mannosidase trims some N-glycans to biantennary Man3GlcNAc2. Complex N-glycans of E. histolytica are made by the addition of α1,2-linked Gal to both arms of small oligomannose glycans, and Gal residues are capped by one or more Glc. In summary, E. histolytica N-glycans include unprocessed Man5GlcNAc2, which is a target for cyanovirin-N, as well as unique, complex N-glycans containing Gal and Glc.

Entamoeba histolytica is the protist (single cell eukaryote) that causes millions of cases of amebic dysentery and liver abscess in regions where its fecal-oral spread cannot be prevented (1,2). Asn-linked glycans (N-glycans) of E. histolytica are of great interest for seven reasons.
First, E. histolytica is missing many of the glycosyltransferases that make lipid-linked precursors to N-glycans and so makes a Man 5 GlcNAc 2 -PP-dolichol rather than Glc 3 Man 9 GlcNAc 2 -PPdolichol, which is present in most animals, plants, and fungi (3,4). Second, E. histolytica N-glycans contribute to the quality con-trol of protein folding in the endoplasmic reticulum (ER) 2 (4 -6). In particular, E. histolytica has UDP-Glc:glycoprotein glucosyltransferase, calreticulin, glucosidase II, and ERGIC-53, which are the essential components of an N-glycan-dependent quality control system (4 -6).
Third, the E. histolytica oligosaccharyltransferase (OST), which transfers N-glycans from the dolichol pyrophosphatelinked precursor to Asn on the nascent peptide in the lumen of the ER is composed of four subunits rather than seven or eight subunits found in metazoan, fungi, and plants (7). As a result, the E. histolytica OST has different kinetics than the OSTs of higher eukaryotes (8). As well, the E. histolytica OST prefers to transfer N-glycans, which resemble its own (Man 5 GlcNAc 2 ), rather than the longer N-glycans of metazoa and fungi (Glc 3 Man 9 GlcNAc 2 ) (8).
Fourth, E. histolytica has a limited set of nucleotide sugar transporters (UDP-Gal and UDP-Glc), which transport activated sugars into the lumen of the ER and Golgi (9). UDP-Gal and UDP-Glc are used to make unique O-phosphodiesterlinked glycans of E. histolytica proteophosphoglycans (10) and may be used to make complex N-glycans.
Fifth, E. histolytica causes disease when the protist uses a Gal-and GalNAc-binding lectin on its surface to adhere to mucins on the surface of host colonic epithelial cells and then lyses host cells by means of secreted proteases and pore-forming peptides (1,2,11,12). Heavy subunits of the E. histolytica Gal/GalNAc lectins have 7-14 potential sites for N-linked glycosylation, and inhibition of N-glycan synthesis results in a Gal/ GalNAc lectin that is unable to bind its target (11).
Sixth, the plant lectin concanavalin A binds to glycoproteins on the surface of E. histolytica and aggregates them into caps that are shed into the medium (13). This result suggests the possibility that E. histolytica may have on its surface high mannose N-glycans. In a similar way, gp120 of HIV has on its surface N-glycans composed of Man 9 GlcNAc 2 (14). These high mannose N-glycans on gp120 are the target of an anti-HIV human monoclonal antibody (2G12) and of bacterial anti-retroviral lectins (cyanovirin-N and scytovirin) (15,16).
Seventh, although N-linked glycans of Giardia lamblia and kinetoplastids (e.g. Trypanosoma cruzi, Trypanosoma brucei, and Leishmania mexicana) have been characterized, N-glycans of the vast majority of protists (e.g. E. histolytica, Trichomonas vaginalis, Plasmodium falciparum, Toxoplasma gondii, and Cryptosporidium parvum) remain uncharacterized (3,(17)(18)(19). Characterization of E. histolytica N-glycans then may lead to a better understanding of the diversity of N-linked glycosylation in protists and may identify novel sugar linkages that have not previously been identified in higher eukaryotes.
Here, biochemical methods including in vivo and in vitro labeling, as well as mass spectrometry, were used to characterize E. histolytica N-glycans. E. histolytica N-glycans include unprocessed Man 5 GlcNAc 2 , which is a target for cyanovirin-N, as well as unique, complex N-glycans containing Gal and Glc.

EXPERIMENTAL PROCEDURES
Bioinformatic Predictions-The predicted proteins of the E. histolytica genome, which has been extensively sequenced, were searched with BLASTP and representatives (e.g. Saccharomyces, Homo, or Escherichia) of each of 79 glycosyltrans-ferase families and 104 glycosylhydrolase families present in the data base of carbohydrate-active enzymes (CAZy) (20 -22). Putative E. histolytica glycosyltransferases and glycosylhydrolases were compared with proteins in the NR Database and to the conserved domain data base at the NCBI using PSI-BLAST. Signal peptides and transmembrane helices of E. histolytica proteins were predicted using SignalP and TMHMM, respectively (23,24).
Reagents  ). Structures were identified by multiple techniques including in vivo and in vitro labeling, size separation on Bio-Gel P-4 and HPAEC, glycosylhydrolase digestions, monosaccharide analysis, and mass spectrometry. An asterisk marks the linkage between Glc and Gal, which could not be determined experimentally.
scytovirin were generous gifts of Barry O'Keefe (NCI-Frederick, National Institutes of Health).
In Vivo Labeling Conditions-E. histolytica strain HM1:IMSS was grown axenically in TYI-S-33 medium containing 0.1% (w/v) Glc to minimize the storage of glycogen, in which fragments otherwise contaminate N-glycan extracts (26). E. histolytica cultures from five 32 cm 2 flasks were combined, centrifuged, and transferred to a microcentrifuge tube with 0.5 ml of radiolabeling medium (the same as described above, excluding Glc and containing 200 Ci of [2-3 H]Man or 2 mCi of [ 14 C]Glc) (3). Amebas were incubated for the indicated time at 37°C and then washed in PBS several times before processing.
Extraction of N-Glycans-Non-incorporated radiolabel was removed by washing the cell pellet with 50% methanol. Cells were lysed with a Dounce homogenizer in 4 ml of Tris-HCl buffer, pH 8. To reprecipitate proteins, the pH was adjusted to 5 with acetic acid, and 4 ml of methanol was added. Tubes were chilled at Ϫ20°C for 4 h and centrifuged at 14,000 rpm for 10 min. This step was repeated twice to wash out soluble glycans such as glycogen fragments. The methanol-denaturated protein pellet was dispersed and treated with 5,000 units of PNGase F in acetate buffer, pH 5, at 37°C for 16 h. The suspension containing the released N-glycans in solution was adjusted to 50% methanol to precipitate large carbohydrates and proteins. After 4 h at Ϫ20°C, the supernatant was cleared by centrifugation and dried. Oligosaccharides were further purified on a porous graphitic carbon Hypercarb column (Thermo Keystone) and washed with 3 ml of distilled water. The N-glycans, which were eluted with 30% acetonitrile, 0.1% trifluoroacetic acid, were dried and resuspended in 300 l of water.
In Vitro Labeling of N-Glycans-Intact E. histolytica vesicles from freshly harvested cultures were prepared as described for nucleotide sugar transport assays (9) Fig. 4). frct., fraction.
described (9). The vesicle preparation still carried a significant amount of cytosolic epimerase activity, which converts UDP-Gal to UDP-Glc, in a reversible reaction. Because of this activity, [ 3 H]Glc was incorporated into E. histolytica glycoproteins when labeling with UDP-[ 3 H]Gal and vice-versa. Therefore, specific labeling was not possible. The labeling of glycans was started by mixing 500 l of vesicle suspension and 500 l of reaction buffer (10 mM MnCl 2 , 10 mM MgCl 2 , 10 mM CaCl 2 , 0.5 M sucrose, 30 mM triethanolamine, pH 7.2, 5 M UDP-Gal, and 5 M UDP-Glc and 30 Ci of each radiolabeled precursor). After 45 min at 37°C, the reaction was diluted with 1 volume of 100 M UDP-Gal and UDP-Glc and separated into 100-l aliquots, each of which was precipitated with 900 l of 4% perchloric acid. After chilling on ice for 30 min, the preparation was centrifuged at 16,000 ϫ g for 30 min, and the pellets were combined and neutralized. The pellets were resuspended with 2 ml of 10 mM Tris-HCl buffer, pH 8, in 50% methanol and homogenized in a 3-ml Dounce. After 2 h at Ϫ4°C, the suspension was centrifuged at 14,000 ϫ g for 10 min, and the pellet was washed twice with Tris-HCl buffer, pH 8, to remove traces of methanol prior to release of N-glycans by digestion with PNGase F (see conditions above).
Sample Preparation for Mass Spectrometry-The starting material to isolate N-glycans for mass spectrometry was a cell pellet obtained from 60 flasks of E. histolytica culture. First, lipids were extracted three times with 30 ml of chloroform: methanol (1:1) and three times with 30 ml of chloroform:methanol:water (10:10:3). Soluble carbohydrates (mostly glycogen fragments) were removed by extracting three times with 20 ml of 10 mM Tris-HCl buffer, pH 8, and reprecipitating with 1 volume of chilled methanol for 4 h at Ϫ20°C. After a final wash in buffer, the pellet was digested with 1 g of TPCK-treated trypsin in 1 ml of 10 mM Tris-HCl buffer, pH 8, for 16 h at 37°C. After boiling to inactivate the protease, the suspension was digested with 20 ϫ 10 3 units of PNGase F in the presence of phenylmethylsulfonyl fluoride for 16 h at 37°C. The mixture was precipitated with 50% methanol, and the supernatant was purified through a porous graphite column, followed by passage through a 3-ml Amberlite mixed bed ion exchange column (H ϩ /acetate form). Finally, the N-glycan extract was dried and resuspended in 300 l of water.
Chromatography-A Bio-Gel P-4 superfine mesh column of 1 ϫ 120 cm was equilibrated in 0.1 M acetic acid, 1% n-butyl alcohol. 300 l of glycan sample was applied and run at a constant flow rate of 5 ml/h, and 1.3-ml fractions were collected. Radioactivity was determined by liquid scintillation counting. The distribution coefficient (K d ϭ (Ve Ϫ Vo)/(Vt Ϫ Vo)) was determined using the glycan elution volume (Ve) relative to the elution of cytochrome C (marker for the exclusion volume; Vo) and glucose (marker for the total volume; Vt). GlcNAc, diacetylchitobiose, and radiolabeled Man 5 GlcNAc 2 (prepared from the dolichol-linked precursors of ⌬alg3 Saccharomyces cerevisiae (4)) were used to calibrate the system. For non-radiolabeled samples, peaks were collected according to the K d of equivalent radiolabeled specimens.
To separate isomers with the same molecular mass, fractions obtained by gel filtration were separated by high performance anion exchange chromatography (HPAEC) with a pulse amper-ometric detector in a Dionex LC20 instrument through a PA100 column (250 ϫ 4 mm) equilibrated in 150 mM NaOH. The flow rate was 0.6 ml/min with a sodium acetate gradient from 12.5 mM (0 -3 min) to 50 mM (at 31 min) and finally to 175 mM (at 70 min). A desalter membrane was installed to remove sodium ions, neutralizing the carbohydrate-containing eluent. Fractions of 0.3 ml were collected, and aliquots were taken for scintillation counting. The retention time of each isomer was determined. The system was calibrated with known N-glycans (Man 3 GlcNAc 2 , Man 1 GlcNAc 2 , and Man 5 GlcNAc 2 prepared from ⌬alg3 S. cerevisiae cells) and with the series laminaribiose, laminaritriose, and laminaritetraose (routinely used as internal markers). Non-radiolabeled samples for mass spectrometry were isolated on a 250 ϫ 9 mm semi-preparative PA100 column (flow rate of 1.5 ml/min, collecting 0.7-ml fractions) calibrated and operated (excluding the internal markers) as described above. Carbohydrate-containing peaks were detected by a pulse amperometric detector, and the corresponding fractions were pooled and dried.
Monosaccharide Composition-Radiolabeled samples were dried, resuspended in 300 l of 2 N HCl in microcentrifuge tubes flushed with N 2 to minimize oxidation, and hydrolyzed for 2 h at 90°C. Acid was removed by several rounds of evaporation under N 2 . The neutral sugars were separated by HPAEC on a MA1 250 ϫ 4 mm column (flow rate of 0.4 ml/min, collecting 0.2-ml fractions) with an NaOH gradient from 100 mM (0 -3 min) to 850 mM (to 45 min). GlcN, Glc, Man, and Gal were used as internal standards.
Mass Spectrometry-Oligosaccharides were permethylated as described previously (27). MALDI-TOF MS was performed on a Bruker Reflex IV mass spectrometer in positive reflectron mode. Between 20 and 50 pmol of sample dissolved in 20% acetonitrile was applied to the MALDI target with an equal volume of 2,5-dihydroxybenzoic acid (20 mg/ml) in 20% acetonitrile with 10 mM sodium acetate added as a cation source. The spectra resulting from 150 and 200 shots from a 337-nm nitrogen laser were summed. The laser pulse was 3 ns. Collision-induced dissociation fragmentation data were collected using a QStar Pulsar i quadruple orthogonal time-of-flight mass spectrometer (Applied Biosystems Inc., Framingham, MA) equipped with an electrospray ionization source. Capillaries were pulled to a 1-micron orifice diameter. Argon (3 p.s.i.) was used as the collision gas for MS/MS experiments. The range of operator-con- In Vitro OST Assay-Total cellular membranes of E. histolytica were incubated for 2-90 min at 37°C with the membrane-permeable tripeptide acceptor 5 M N ␣ -acetyl-N-125 I-Tyr-Thr-NH 2 (NYT) in the presence of deoxynojirimycin to ensure that the glycopeptide products were not degraded by glucosidases I and II (29,30). In some experiments, swainsonine was added to determine the effect of endogenous E. histolytica mannosidases on glycopeptides. Glycopeptide products were collected by binding to immobilized concanavalin A and separated by HPLC using an aminopropyl silica column. Glycopeptide standards (Man 5 GlcNAc 2 -NYT and Man 9 GlcNAc 2 -NYT) were prepared using the Saccharomyces OST and purified dolichol-linked oligosaccharides (31).
Cyanovirin-N Labeling and Capping of E. histolytica-Freshly harvested E. histolytica were washed three times with PBS and incubated with 20 g/ml BODIPY-labeled cyanovirin-N for 30 min at 4°C (14,15). Unbound cyanovirin-N was removed by rinsing twice with PBS. For surface labeling, E. his- FIGURE 4. Characterization of the E. histolytica N-glycans by in vivo labeling and glycosylhydrolase digestions. E. histolytica N-glycans, which were prepared as in Fig. 2, were treated with glycosylhydrolases, and digestion products were identified according to retention times (r.t.) of known standards (e.g. Man 3 GlcNAc 2 ). For isomers H6.2, H6.5, and H7.1, the dashed line enclosing the Man␣1,6-arm indicates what corresponds to the residual fragment 11.5, which was fully characterized in a separate [U-14 C]Glc-labeling experiment. Note that jack bean ␣-mannosidase (JBAM) removes an exposed Man␣1,3-(as in isomer H4.3), but it does not remove the Man␣1,6-unless the Man␣1,3-arm is digested first. Hence, isomer H4.2 becomes susceptible to jack bean ␣-mannosidase digestion only after removal of Gal␣1,2-by ␣-galactosidase. The mode of action of jack bean ␣-mannosidase (37) supports our assumption that the different sensitivities of H4.2 and H4.3 are due to accessibility of ␣-linked Man residues. In accordance with this, the underlying structures of other isomers were inferred as well. ASAM, A. saitoi ␣1,2-mannosidase. tolytica were immediately fixed with 2% paraformaldehyde in PBS for 10 min at 4°C. To determine whether cyanovirin-N aggregates and forms caps on the surface of E. histolytica, we incubated cyanovirin-N-bound cells in PBS for 15 min at 37°C and subsequently fixed them in 2% paraformaldehyde. For labeling the internal membranes and surface of E. histolytica, amebas were fixed in 2% paraformaldehyde and 0.1% Triton X-100. Organisms were visualized with a DeltaVision deconvoluting microscope (Applied Precision, Issaquah, WA) with the help of Landon Moore of the Department of Genetics and Genomics at Boston University School of Medicine. Images were taken at ϫ100 primary magnification and deconvolved using Applied Precision softWoRx software.

RESULTS
Brief Summary of the Experimental Strategy-The structures of E. histolytica N-glycans, which are shown in Fig. 1, were determined by in vivo labeling with [2-3 H]Man (shown in Fig. 2) or [U-14 C]Glc (shown in Fig. 3) or by the separation of N-glycans released by PNGase F on Bio-Gel P-4 and HPAEC, followed by digestion with specific glycosylhydrolases (shown in Fig. 4). Alternatively, E. histolytica N-glycans were identified by in vitro labeling of E. histolytica membranes with nucleotide sugars (nucleotide sugar transfer assays shown in Fig. 6) or with a radiolabeled tripeptide (OST assays shown in Fig. 7). In addition, E. histolytica N-glycans were examined by mass spectrometry (shown in Figs. 8 and 9). These strategies allowed us to identify each E. histolytica N-glycan by multiple methods. Finally, we localized unprocessed N-glycans on the surface of E. histolytica using the anti-retroviral lectin cyanovirin-N (shown in Fig. 5).
The  Fig. 1), which was characterized by mannosidase digestions (Fig. 4). When E. histolytica were pulse-labeled for 10 min with [2-3 H]Man, washed, and chased for 3 h with unlabeled Man, biosynthetic Man 5 GlcNAc 2 was still the most abundant N-glycan (data not shown). This result shows that many E. histolytica N-glycoproteins are not modified by Golgi-type glycosylhydrolases and glycosyltransferases.

Man 5 GlcNAc 2 on the Surface of E. histolytica Trophozoites Is Bound and Capped by the Anti-retroviral Lectin
Cyanovirin-N-The presence of biosynthetic Man 5 GlcNAc 2 (H5.1) on the surface of cultured E. histolytica was demonstrated using the anti-retroviral lectin cyanovirin-N (Fig. 5) (15). Cyanovirin-N is specific for the Man␣1,2-Manon the single D1 arm of Man 5 GlcNAc 2 (E. histolytica N-glycans) or on three arms of Man 9 GlcNAc 2 (N-glycans of gp120) (Fig. 5A). Cyanovirin-N was bound to the surface and vesicular membranes of fixed and permeabilized E. histolytica (Fig. 5B). Cyanovirin-N was also bound evenly to glycoproteins on the surface of chilled but living E. histolytica (Fig. 5C). When E. histolytica were warmed to 37°C, cyanovirin-N aggregated into caps (Fig. 5D), which resemble those formed by the lectin concanavalin A on the E. histolytica surface (11,13). As a neg-ative control, scytovirin, another anti-retroviral lectin that is specific for Man on the upper (D3) arm of Man 9 GlcNAc 2 , which is absent from E. histolytica N-glycans, did not bind to the surface or vesicles of E. histolytica (data not shown) (15). Fig. 2A), E. histolytica N-glycans were composed of Man 5 GlcNAc 2 (H5.1) and Glc 1 Man 5 GlcNAc 2 (H6.1). Glc 1 Man 5 GlcNAc 2 was markedly increased, whereas Man 5 GlcNAc 2 was decreased, when E. histolytica trophozoites were labeled in the presence of castanospermine, which is a glucosidase II inhibitor (inset in Fig. 2A) (30). These results are consistent with the presence of all the components of N-glycandependent quality control of protein folding in E. histolytica (4 -6). See supplemental Table 1 for a list of the E. histolytica glycosylation-related genes.

GlcMan 5 GlcNAc 2 , the Product of UDP-Glc:Glycoprotein Glucosyltransferase Involved in N-Glycan-dependent Quality Control of Protein Folding, Is Also Present in Mature N-Glycans of E. histolytica-At the earliest time points of metabolic labeling with [2-3 H]Man (
When E. histolytica membranes were incubated in vitro with UDP-[ 3 H]Glc, the most abundant N-glycan was also Glc 1 Man 5 GlcNAc 2 (Fig. 6, B and E). Glc 1 Man 5 GlcNAc 2 was also present in relatively high amounts in mature glycoproteins, indicating that this N-glycan is not just a transient species in E. histolytica (Fig. 2, D and H). Some E. histolytica N-Glycans Are Trimmed Back to Man 3 GlcNAc 2 and Man 4 GlcNAc 2 by a Swainsonine-sensitive Mannosidase-The activity of the E. histolytica ␣-mannosidase was inferred by the accumulation of Man 3 GlcNAc 2 N-glycans (15, 16), binds to the surface E. histolytica trophozoites and forms caps. A, cyanovirin-N binds to terminal ␣1,2-linked Man, which is present on each of the three arms of Man 9 GlcNAc 2 and is also present on the single arm of Man 5 GlcNAc 2 . B, cyanovirin-N binds to the surface and to vesicular membranes of fixed and permeabilized E. histolytica. n, nucleus. C, cyanovirin-N is evenly distributed on the surface of a live E. histolytica incubated with the lectin in the cold, so it cannot cap. D, cyanovirin-N is capped on the surface of a live E. histolytica that is allowed to warm up to 37°C for 10 min. The negative control was scytovirin (a lectin that binds to an intact D3 arm, which is absent in E. histolytica N-glycans). E. histolytica incubated with scytovirin were not labeled (data not shown).
The largest oligosaccharide attached to the radiolabeled tripeptide (NYT) in vitro in the presence of UDP-Gal, UDP-Glc, and swainsonine is Glc 1 Man 5 GlcNAc 2 (H6.1) (6). These results suggest that the OST, the mannosidase, and the UDP-Glc:glycoprotein glucosyltransferase are all present in the ER, whereas the Gal-and Glc-transferases (see below) are in a distinct organelle (likely the Golgi apparatus).
GDP-Fuc (data not shown). Indeed, monosaccharide analysis of N-glycans of E. histolytica labeled in vivo with [ 14 C]Glc revealed Man, Gal, and Glc but did not show other hexoses or deoxyhexoses (Fig. 3B). Treatment with hexosaminidase, along with mass spectrometry results, showed that N-glycan extensions do not contain hexosamines. All of the N-glycans of E. histolytica were neutral oligosaccharides (data not shown) lacking charged sugars (e.g. sialic acids) or other modifications such as sulfation or phosphorylation.
In Many Complex N-Glycans of E. histolytica, Gal Is Capped by One or More Glc Residues-The evidence for complex E. histolytica N-glycans containing Gal and Glc included the following. With increasing time, Hex 7 to Hex 10 were present in N-glycans of E. histolytica labeled in vivo with [2-3 H]Man (Fig. 2) or [ 14 C]Glc (Fig. 3). As the size of N-glycans increased, the percentage of Glc in the oligosaccharides increased (Fig. 3B), and these larger N-glycans (e.g. H5.3, H6.2, H6.5, and H7.1) were resistant to ␣and ␤-galactosidases, consistent with a Glc cap (Fig. 4). With the exception of the product of the UDP-Glc: glycoprotein glucosyltransferase (Glc 1 Man 5 GlcNAc 2 ) (H6.1), Glc was only added to N-glycans of E. histolytica after the addition of Gal (Fig. 6).
Because the complex N-glycans of E. histolytica that have Glc caps were resistant to all the glycosylhydrolases that we tested, it was not possible to determine whether the Glc-extensions are ␣or ␤-linked; hence, the configuration of the linkage between Glc and Gal remains to be defined. A pseudo MS/MS/MS experiment performed on B 4 ions of Hex 8 (which originate from glycans with a linear series of four or more hexoses) suggests the possibility that some of the extending Glc residues are 1,6-linked (supplemental Fig. 4). A full discussion of the mass spectrometry experiments is presented in the supplemental material.

Properties of E. histolytica N-Glycans That Distinguish Them from N-Glycans of Fungi and
Metazoa-As shown by analysis of predicted proteins from whole genome sequences, E. histolytica is missing seven asparagine-linked glycosylation enzymes, and so N-glycans are built upon the truncated precursor Man 5 GlcNAc 2 (H5.1 in Fig. 1) (see supplemental Table 1) (3). Man 5 GlcNAc 2 is glucosylated by the UDP-Glc:glycoprotein glucosyltransferase of the N-glycan-dependent quality control system for protein folding to make Glc 1 Man 5 GlcNAc 2 (H6.1) (4 -6). For the same reason, N-gly-  Cross-ring fragments 0,4 A 2 , m/z 301.14 and 3,5 A 2 , m/z 329.16 (C) define the Man␣1,6-arm. A low abundance 1,3 A 2 fragment at m/z 315. 16 (C) indicates the presence of an isomer bearing a terminal 1,2-linked Gal (H3.2). A Y 3 / 0,2 A 4 fragment ion seen at m/z 375.17 (B) is consistent with the core region of a linear isomer. No signal is seen at m/z 505.2 or 533.2 (for a substituted Man␣1,6-). Therefore, the structure for H3.2 indicates a missing Man␣1,6-arm, with the remaining Man␣1,3-arm extended with Gal␣1,2-. These results are consistent with the glycosylhydrolase digestion data (Fig. 4).
In contrast to fungi and metazoa, a large proportion of E. histolytica N-glycans reach the plasma membrane as unmodified Man 5 GlcNAc 2 (H5.1), bypassing ER and Golgi mannosidases and glycosyltransferases. Similarly, a large fraction of mature N-glycans of E. histolytica are composed of Glc 1 Man 5 GlcNAc 2 , as has been described in Leishmania (33). The presence of Glc 1 Man 5 GlcNAc 2 in mature glycoproteins suggests that it may have other functions in addition to serving as an intermediate in the N-glycan-dependent quality control of folding in the ER lumen (4 -6).
Because we did not link N-glycans to sites on particular E. histolytica glycoproteins, there are three possible explanations for the abundance of unmodified Man 5 GlcNAc 2 in the plasma membrane. First, N-glycans on some E. histolytica glycoproteins are never processed. Second, N-glycans at multiple sites on the same glycoprotein are processed differently. Third, E. histolytica mannosidases and glycosyltransferases that make

Entamoeba N-Glycans
complex N-glycans are very inefficient. We speculate that the high Man N-glycans of E. histolytica may be harder to recognize by the host immune system than the unique complex N-glycans of E. histolytica. A similar argument has been made to explain the abundance of high Man N-glycans on gp120 of HIV (14).
The capping of E. histolytica plasma membrane glycoproteins containing Man 5 GlcNAc 2 by cyanovirin-N suggests the possibility that anti-viral lectins such as cyanovirin-N might be used to block binding of E. histolytica to the host epithelium. Cyanovirin-N inhibits phagocytosis of mucin-coated beads by E. histolytica. 3 As well, affinity columns with concanavalin A efficiently capture E. histolytica glycoproteins with N-glycans (so-called N-glycome) for mass spectrometric identification of peptides. 3 In addition to OST and UDP-Glc:glycoprotein glucosyltransferase activities (6,8), the ER of E. histolytica contains a swainsonine-inhibitable ␣-mannosidase, which trims a fraction of the N-glycans to Man 3 GlcNAc 2 (H3.1) and Man 4 GlcNAc 2 (H4.1). As in metazoa, the trimmed N-glycans of E. histolytica are building blocks for formation of complex N-glycans (34,35).
E. histolytica has a single predicted ␣-mannosidase (glycosylhydrolase family 92 or GH92), which is similar to those of some fungi (Aspergillus, Neurospora, and Magnaporta) and many Gram-positive bacteria (e.g. Streptococcus, Porphyromonas, Bacteroides, and Mycobacteria) (supplemental Table 1) (20 -22). Phylogenetic analyses strongly suggest that the E. histolytica GH92 mannosidase was obtained by lateral gene transfer because the E. histolytica mannosidase is present in a clade with bacteria, and this clade is supported with high bootstrap values (supplemental Fig. 5). In contrast, the fungal mannosidases were present in clades that did not include the E. histolytica mannosidase. Lateral gene transfer is an important mechanism by which E. histolytica has obtained many dozen bacterial genes encoding fermentation and other enzymes (20). E. histolytica is missing mannosidases similar to ER or Golgi mannosidases (GH47) or to lysosomal mannosidases (GH38) (20 -22, 32).
The N-glycans of E. histolytica, which have ␣1,2-linked Gal that is extended with Glc, are simpler than the complex N-glycans of metazoa. The N-glycans of E. histolytica are consistent with the experimental demonstration of two E. histolytica nucleotide sugar transporters (for UDP-Glc and UDP-Gal) (9) and with bioinformatic predictions of a limited number of Golgi glycosyltransferases in E. histolytica (see supplemental Table 1). O-phosphodiester-linked glycans of E. histolytica are also composed of Gal capped with Glc, but the linkages are different from those of N-glycans (10). Although ␣1,2-Gal linked to Man is also present in N-glycans of Schizosaccharomyces pombe (36), N-glycans with distal Glc capping residues are not present in the human host of E. histolytica. This suggests the possibility that the unique complex N-glycans of E. histolytica may be targets for future anti-amebic vaccines. In support of this idea, the O-phosphodiester-linked glycans, which are also rich in Gal and Glc, are one of a limited number of current anti-amebic vaccine targets (1,2,10).