A (cid:1) - N -Acetylglucosaminyl Phosphate Diester Residue Is Attached to the Glycosylphosphatidylinositol Anchor of Human Placental Alkaline Phosphatase A TARGET OF THE CHANNEL-FORMING TOXIN AEROLYSIN*

Glycosylphosphatidylinositol (GPI)-anchored proteins are ubiquitous in eukaryotes. The minimum con-served GPI core structure of all GPI-anchored glycans has been determined as EtN-PO -inositol-PO 3 Human placental al- kaline phosphatase (AP) has been reported to be a GPI-anchored in our data. Although (cid:2) -linked GlcNAc phosphate is replaced at C-6 of the mannose, and mild acid-phosphatase digestion abolished the interaction between sAP and proaerolysin, the binding could not be inhibited by 10 (cid:5) 5 M mannose 6-phosphate, inositol phosphate, GlcNAc phosphate, or ethanolamine phosphate (data not shown). These results suggest that the minimum structure of the GPI-anchored glycan which is required for recognition by aerolysin is a larger sized com-ponent than mannose 6-phosphate.

Alkaline phosphatase (orthophosphoric monoester phosphohydrolase, EC 3.1.3.1) is a glycoprotein that is widely distributed in the plasma membrane. The isozymes derived from various mammalian organs show different enzymatic and immunochemical properties (1). Human placental alkaline phosphatase (AP) 1 was the first protein that was identified as being released from the plasma membrane by bacterial phosphatidylinositol (PI)-specific phospholipase C (2,3). The structures of the N-linked glycan and GPI anchor of AP have been already reported. Although the amino acid sequence indicates that AP has two potential N-glycosylation sites (4 -6), only Asn 249 was suggested to be actually glycosylated (4). Endo et al. (7) reported that the structure of the N-linked glycan of AP is NeuAc␣233Gal␤134GlcNAc␤132Man␣133(NeuAc␣23 3Gal␤134GlcNAc␤132Man␣136)Man␤134GlcNAc␤13 4(ϮFuc␣136)GlcNAc, as derived by a combination of sequential exoglycosidase digestion and methylation analysis. On the contrary, the structure of the GPI-anchored glycan of AP was investigated by Redman et al. (8) and reported to be Thr-Asp-EtN-PO 4 -6Man␣1-2Man␣1-6Man␣1-4GlcN-(sn-1-O-alkyl-2-O-acylglycerol-3-PO 4 -1-myo-D-inositol), with an additional ethanolamine phosphate group. However, the ethanolamine phosphate had been analyzed quantitatively by amino acid analysis and metabolic labeling method (19). In this study using a combination of ␤-GlcNAc-specific PVL-Sepharose column chromatography, LC/ESI-MS analysis, chromatofocusing, nitrous acid deamination treatment, and periodate oxidation, we found that not only ethanolamine phosphate, but also a ␤-N-acetylglucosaminyl phosphate diester (GlcNAc-P) residue are present in AP and are positioned as side chains of its GPI-anchored glycan. GlcNAc-P has not been reported previously in GPI-anchored glycans, probably because it is easily hydrolyzed by mild acid treatment.
We also investigated whether GlcNAc-P residues were asso-ciated with GPI-anchored glycans of human carcinoembryonic antigen (CEA), cholinesterase, and Tamm-Horsfall (T-H) glycoprotein using PVL-Sepharose column chromatography. All of these glycoproteins carry both N-glycans (9 -12) and GPI-anchored glycans, and most of them effectively bound to a PVL-Sepharose column. Because nonreducing terminal ␤GlcNAc residues were not present in their N-glycans, the GlcNAc-P residues appear to be linked to GPI-anchored glycans as a side chain. These results suggest that the GlcNAc-P residue is widely distributed on GPI-anchored glycoproteins of the human cell membrane.
Aerolysin is a toxin released by Aeromonas hydrophila (13). Binding of aerolysin to receptors on target cells promotes receptor oligomerization, and this phenomenon is followed by membrane insertion and channel formation. Although GPIanchored glycans of cell surface components have been reported to function as aerolysin receptors (14 -17), the precise carbohydrate binding specificity of aerolysin remained unclear. Because proaerolysin, a nontoxic precursor of aerolysin, also has carbohydrate binding ability, we investigated the binding of AP to proaerolysin using surface plasmon resonance (SPR) and enzyme-linked immunosorbent assays (ELISAs).

EXPERIMENTAL PROCEDURES
Materials-Pronase K, trypsin, wheat germ acid phosphatase, and NaNO 2 were purchased from Sigma. NaB 3 H 4 (490 mCi/mmol) was purchased from PerkinElmer Life Sciences. Diplococcal ␤-N-acetylhexosaminidase was purified from a fluid culture of Diplococcus pneumoniae according to the method of Glasgow et al. (18). N-Acetylglucosamine and Arthrobacter ureafaciens sialidase were purchased from Nacalai Tesque (Kyoto, Japan).
Preparation of Soluble Form of AP and Its Derivatives-sAP, which was converted from the membrane form (1), was purified from human term placentas that were obtained within 2 h of delivery as reported previously (19). In brief, sAP extracted from the membranes with 1-butanol at pH 5.5 was purified by sequential chromatography through ConA-Sepharose, Sephacryl S-300, and hydroxylapatite columns. Mild acid-treated sAP was prepared as follows. 10 g of sAP was incubated in 50 l of 0.01 M HCl at 100°C for 30 min, to hydrolyze GlcNAcphosphodiester linkages (20), and the pH was adjusted to 7. sAP treated with acid hydrolysis and phosphatase was prepared from acid-treated sAP by incubation with a 20-l slurry of bovine intestine alkaline phosphatase-immobilized beads (Sigma) after adjusting the pH to 8.
PVL-Sepharose Column Chromatography-PVL was purified from the fruiting bodies of Psathyrella velutina as reported previously (21). PVL-Sepharose (3 mg/ml) was prepared using the CNBr method (22). PVL-Sepharose columns were equilibrated with PBS containing 0.02% NaN 3 . Samples dissolved in 100 l of PBS were applied to the column and kept at 4°C for 15 min. Elution was started with 5 ml of PBS at 4°C followed by 5 ml of PBS containing 0.3 M GlcNAc at room temperature.
Enzyme Activity-sAP enzyme activity was assayed using the enzyme substrate, 6.7 mol of p-nitrophenylphosphoric acid disodium salt, in 0.1 M carbonate buffer, pH 9.6. The reaction mixture was incubated at 25°C for 30 min. After stopping the reaction with 100 l of 1 M NaOH, the released chromogen was measured with a spectrophotometer (EIA Reader, Bio-Rad model 3550).
Following the application of 20 l of human serum to PVL-Sepharose columns, cholinesterase activity in the fractions was measured using the ChE B-test Wako (Wako Pure Chemical Co., Osaka, Japan). One IU was designated as the amount of enzyme which hydrolyzed 1 mol of benzoylcholine chloride in 1 min at 37°C.
PVL-Sepharose Column Chromatography of CEA-The human colon carcinoma cell line CCK81 was cultured in RPMI 1640 medium supplemented with 10% heat-inactivated calf serum (Invitrogen) and 100 units/ml penicillin at 37°C. Cells cultured on 60-mm dishes were treated with 5 milliunits of PI-phospholipase C in 1 ml of PBS at 37°C for 1 h. 100 l of the supernatant was applied to a PVL-Sepharose column, as described above. The CEA content was measured by ELISA using anti-CEA (23) as follows. ELISA plates (Corning, Inc., Corning, NY) were coated with 20 g/ml rabbit anti-CEA polyclonal antibody in 20 mM carbonate buffer, pH 9.6, at 4°C overnight. The plates were washed with 0.05% Tween 20 in PBS, pH 7.3, blocked with PBS containing 0.05% Tween 20 and 1% bovine serum albumin, and then treated with sample. After washing with 0.05% Tween 20 in PBS, 5 g/ml goat anti-CEA polyclonal antibody was added and incubated for 1 h at 37°C. After further washing, AP-conjugated rabbit anti-goat IgG antibody was added and incubated for 1 h at 37°C. After washing, 6.7 mol p-nitrophenylphosphoric acid disodium salt in 0.1 M carbonate buffer, pH 9.6, was added, and absorbance at 405 nm was measured with a spectrophotometer.
PVL Column Chromatography of T-H Glycoprotein-T-H glycoprotein was purified from 1 liter of pooled urine obtained from normal adults as follows (24). 0.58 mol of NaCl was added to the urine, and it was stirred at 4°C for 18 h. The precipitate formed was collected by centrifugation, solubilized with distilled water, and dialyzed against distilled water. After centrifugation, the supernatant was freeze dried and dissolved in PBS. An aliquot of T-H glycoprotein was subjected to SDS-PAGE using 10% polyacrylamide gels to check the purity. It showed a single band of 80 kDa corresponding to the molecular mass of T-H glycoprotein. One mg of the purified T-H glycoprotein was applied to a PVL column, and each fractionated T-H glycoprotein was measured using a protein assay (Bio-Rad).
Chromatofocusing of sAP and Its Derivatives-Nontreated and treated sAPs were separated on a Mono P HP5/20 chromatofocusing column (0.5 cm inner diameter ϫ 9.6 cm long) at 4°C. After the column was equilibrated with 0.025 M histidine HCl, pH 6.2, 1 g of sAP was injected onto the column. Elution was carried out isocratically with 10 ϫ diluted polybuffer 74 (Amersham Biosciences), pH 4.0, at a flow rate of 0.5 ml/min. The effluent was monitored using a pH sensor, and sAP activities were assayed. The sAP isoforms, collected in 1 ml of the eluent, were digested with 10 l each of Arthrobacter sialidase (5 milliunits), diplococcal ␤-N-acetylhexosaminidase (10 milliunits), or wheat germ acid phosphatase (10 g) at 37°C for 2 h.

Preparation of 3 H-labeled GPI-anchored Glycan from sAP by Nitrous Acid Deamination or Direct Labeling with NaB 3 H 4 Reduction-One mg
of sAP was digested thoroughly with 1 mg of Pronase K in 100 l of Tris-HCl buffer, pH 8.0, at 37°C for 2 h. The glycopeptides were dried and dissolved in 0.1 ml of 50 mM acetate buffer, pH 3.5. 100 l of 0.33 M NaNO 2 was added to the mixture, and this was incubated at room temperature for 2 h. Subsequently 0.2 ml of 0.2 M boric acid was added, and the GPI-anchored glycans were reduced with NaB[ 3 H] 4 . 3 H-Labeled GPI-anchored glycans were purified by Dowex AG-50 column chromatography. In direct labeling with NaB 3 H 4 reduction, 100 g of sAP was reduced with NaB 3 H 4 in 100 l of 0.05 N NaOH at 30°C for 3 h and separated from NaB 3 H 4 by PD-10 column chromatography. 3 H-Labeled sAP was digested thoroughly with 0.1 mg of Pronase K in 50 l of Tris-HCl buffer, pH 8.0, at 37°C for 2 h. 3 H-Labeled GPI-anchored glycans were purified by Dowex AG-50 column chromatography.
LC/ESI-MS of Trypsin-digested sAP-One mg of sAP was digested with 100 g of trypsin (sequence grade, Sigma) in 100 l of Tris-HCl buffer, pH 8.0, at 37°C for 2 h. LC/ESI-MS was carried out using a CapLC system (Waters Corp.) with an Xterra MS-18 column (1.0 ϫ 50 mm, Waters Corp.) and a ZQ mass spectrometer with an electrospray ion source (Waters Corp.). The eluents were 0.1% trifluoroacetic acid in H 2 O (A pump), and 0.1% trifluoroacetic acid in acetonitrile (B pump). The flow rate was 20 l/min, and the effluent was monitored at 210 -400 nm. The gradient conditions were 5% of B for 5 min, 5-25% of B at 1%/min, and 25-60% of B at 0.5%/min. The ESI voltage was set at 3.5 kV, the desolvation gas (nitrogen) temperature was 150°C, and the ion source block temperature was 80°C. Two scan functions were recorded simultaneously to detect molecular masses of sAP trypsin digests and fragment ions from sugars. The first scan function was from 70 to 2,500 atomic mass units in 2 s with the low cone voltage at 25 eV, and the second scan function was from 70 to 1,200 atomic mass units in 0.5 s with the high cone voltage at 70 eV.
Solid Phase Binding Assay-Proaerolysin was purchased from Protox Biotech (Victoria, Canada) and was biotinylated by EZ-link Sulfo-NHS-LC-biotin (Pierce) according to the manufacturer's instructions, following which it was dialyzed against PBS. The binding of proaerolysin to AP was measured using a solid phase binding assay. ELISA plates were coated with AP at 10 mg/ml in 20 mM carbonate buffer, pH 9.6, at 4°C for 18 h. The plates were washed with 0.05% Tween 20 in PBS, pH 7.3, blocked with PBS containing 0.05% Tween 20 and 3% human serum albumin, and then treated with biotin-proaerolysin in PBS at 37°C for 2 h. After washing with 0.05% Tween 20 in PBS, the plates were treated with avidin-alkaline phosphatase (Zymed) at 37°C for 1 h. After further washing, AP activities were measured, as described above.
SPR Assay-Binding kinetics and affinities were determined by SPR assay using a Biacore 2000 biosensor system (Biacore, Inc., Piscataway, NJ). All analyses were performed on research-grade CM5 sensor chips (Biacore, Inc.). To assess proaerolysin binding to nontreated sAP, sialidase-treated sAP, acid-treated sAP, and sAP treated with acid and phosphatase, they were immobilized on the chips in 10 mM sodium acetate buffer, pH 4.5, at protein concentrations of 10 g/ml, using the amine coupling kit supplied by the manufacturer (Biacore, Inc.). The surface density was 405 RU for nontreated sAP, 520 RU for sialidasetreated sAP, 620 RU for acid-treated sAP, and 510 RU for sAP treated with acid and phosphatase. Measurements were carried out in 10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, and 0.005% Tween 20 (Biacore, Inc.) at 25°C and at a flow rate of 10 l/min. After analyte binding, surfaces were regenerated with 0.1 M HCl and a contact time of 30 s. Sensorgram data were analyzed using BIA evaluation version 3 software (Biacore, Inc.).
Monosaccharide Composition Analysis-To determine the reducing terminal residue released from 3 H-labeled sAP, 1 ϫ 10 5 dpm of direct 3 H-labeled GPI-anchored glycan was hydrolyzed in 100 l of 4 N HCl at 100°C for 3 h, N-acetylated, and applied to a SP0810 column (10 mm ϫ 300 mm, Shodex). The eluent was 20% ethanol, and the flow rate was 0.5 ml/min.
Periodate Oxidation-1.1 mg of sAP was radiolabeled with NaB 3 H 4 to monitor the recovery of the following steps. sAP was digested thoroughly with 1 mg of Pronase K in 100 l of PBS at 37°C for overnight. The glycopeptides were dried and dissolved in 100 l of 0.05 M acetate buffer, pH 4.5, containing 0.6 mol of sodium metaperiodate, and the mixture was kept at 4°C for 40 h in the dark. Excess oxidant was destroyed by adding 10 l of ethylene glycol. After standing at room temperature for 1 h, 10 mg of NaBH 4 in 500 l of 0.1 M sodium borate buffer, pH 9.5, was added, and the mixture was left at 30°C for 3 h. After adding acetic acid, the reduced sample flowed through a column of Dowex AG-50 (H ϩ form) (2-ml bed volume). The eluates were evaporated, and the sample was freed from boric acid by repeated evaporation with methanol. The sample was then hydrolyzed in 1 ml of 2 N HCl at 100°C for 2 h and thoroughly dried up. The sample was radiolabeled with NaB 3 H 4 in 100 l of 0.1 M sodium borate buffer, pH 9.5, at 30°C for 3 h, and the excess NaB 3 H 4 was consumed by 10 l of benzaldehyde. After passing through with C18 Sep-Pak column (Waters Co.), the sample was evaporated and freed from boric acid by repeated evaporation with methanol. The oxidized oligosaccharide alcohols were then separated with paper electrophoresis. High voltage paper electrophoresis was performed by using pyridine/acetate buffer, pH 5.4 (pyridine/ acetic acid/water, 3:1:387), at a potential of 73 V/cm for 40 min. The acidic components were extracted and digested with 50 l of 1 mg/ml acid phosphatase in 0.2 M citrate buffer, pH 5.0, at 37°C overnight, then purified using ion exchange gels AG-50 and AG-3 (Bio-Rad). Then, the ascending paper chromatography was performed using the solvent butanol/ethanol/water (4:1:1) for 5 h. NaB 3 H 4 -reduced glyceraldehyde, erythrose, and mannose were used as standards. The developed paper was cut into 1-cm pieces and measured with a scintillation counter.
In preliminary experiments, we observed that asialo-sAP bound to a PVL-Sepharose column, which interacts with ␤-Gl-cNAc residues (21), and was eluted with 0.3 M GlcNAc (Fig. 1A). Because all the N-glycans of asialo-sAP are galactosylated, nonreducing terminal ␤-GlcNAc residues must be present in the GPI-anchored glycan of sAP.
To determine how the ␤-GlcNAc residue is attached as a side chain of the GPI-anchored glycan, several analytical methods were applied. After Pronase digestion of sAP, glucosamine residues in GPI-anchored glycopeptides were deaminated by nitrous acid treatment and reduced with NaB 3 H 4 . [ 3 H]Anhydromannitol was confirmed by monosaccharide composition analysis (data not shown). The 3 H-labeled GPI-anchored glycopeptides also bound to a PVL column and were eluted with 0.3 M GlcNAc (Fig. 2A). The PVL binding ability of 3 H-labeled GPI-anchored glycopeptides was abolished by digestion with diplococcal ␤-N-acetylhexosaminidase (Fig. 2B). Mild acid hy-drolysates of 3 H-labeled GPI-anchored glycopeptides also failed to bind the PVL column (Fig. 2C), suggesting that the ␤-Glc-NAc residue does not bind to GPI-anchored glycopeptides via O-glycosidic bonding. Because N-acetylglucosaminyl phosphate diesters can be easily hydrolyzed by mild acid treatment (20), the existence of a GlcNAc␤13phosphodiester residue (Gl-cNAc-P) as a GPI-anchored glycan side chain was suggested. Because 3 H-labeled GPI-anchored glycopeptides were easily adsorbed to paper or to a high performance liquid chromatography column, we investigated whether asialo-AP itself contained a GlcNAc-P residue.
Chromatofocusing of sAP-sAP used in this study flowed through a RCA-I-agarose column, indicating that all N-acetyllactosamine moieties within its glycan are replaced by sialic acid. Purified sAP was eluted from the chromatofocusing column at pH 4.6 (peak a) and pH 4.8 (peak b), as shown in Fig.  3A. sAP forms a homodimer and contains 4 mol of sialic acid/ molecule (7). Because sialidase-treated APs were eluted at pH 5.2 and 5.0 (Fig. 3, B and E), one negatively charged sialic acid corresponded to an approximate ⌬pH of 0.1. Asialo-sAPs were resistant to wheat germ acid phosphatase. However, after di-FIG. 1. PVL column chromatography of human GPI-anchored glycoproteins. Each sample was applied to a PVL-Sepharose column (1-ml bed volume) at 4°C. The column was washed with PBS and then with 0.3 M GlcNAc in PBS from the position indicated by the black arrows. A, 100 l of sAP (1 mg/ml) was fractionated, and enzyme activities were measured. B, human serum was fractionated, and cholinesterase activities were measured. C, PI-phospholipase C digests of CCK81 cells were fractionated, and CEA contents were measured by ELISA. D, purified T-H glycoprotein was fractionated and measured by protein assay. gestion with diplococcal ␤-N-acetylhexosaminidase, sAPs were eluted at pH 5.0 and 4.8 (Fig. 3, C and F). It was suggested that the negative charges exposed by ␤-N-acetylhexosaminidase digestion contribute to the delay in retention time. Subsequent to that digestion step, sAPs were digested thoroughly by wheat germ acid phosphatase and were eluted at pH 5.4 and 5.2 (Fig.  3, D and G), suggesting that they had lost their negatively charged phosphate residues. Mass spectroscopy indicated that the GPI-anchored glycan consisted of two isoforms, one with and one without an ethanolamine phosphate. Because ethanolamine phosphate diester has a single negative charge at pH 5.0, isoform b must contain both a GlcNAc-P residue and ethanolamine phosphate as side chains, whereas isoform a contains only the GlcNAc-P residue. These results suggested that ␤-N-acetylglucosaminyl phosphate binds to the GPI-anchored glycans of AP.
LC/ESI-MS of Trypsin-digested sAP-To confirm the structures of the sAP GPI-anchored glycans, LC/ESI-MS analysis was applied. Trypsin-digested sAP was analyzed by LC/ ESI-MS with an Xterra MS-18 column using trifluoroacetic acid/acetonitrile as an eluent. The total ion chromatogram, which is presented in Fig. 4A, showed the respective peptides and GPI-anchored glycans corresponding to trypsin-digested sAP. Peaks c and d, at retention times 25.9 and 61.3 min, were assumed to represent peptides containing GPI-anchored glycans from the mass chromatograms monitored at m/z 79, 162, and 447 (Fig. 4, B, C, and D), which are characteristic fragment ions, HPO 4 highϪ, GlcN ϩ , and (EtN-PO 4 highϪ)Man-GlcN ϩ , derived from GPI anchors. From the mass spectra of peak c (Fig. 4B), the observed molecular masses of peak c were deduced as 3,033 and 3,156 Da. When the cone voltage of peak c was increased to 70 eV, several fragment ions were obtained from the GPI-anchored glycan (Fig. 4C). These were assigned as GlcN (m/z 161.9), Man-PO 4 GlcN (m/z 646.3). The detection of a m/z 646.3 peak indicated that ␤-N-acetylglucosaminyl phosphate binds to the second mannosyl residue of the core structure. Accordingly, the peptide portion of peak c should correspond to a peptide comprising Ala 466 -Asp 484 (1,980 Da). Although sAP was digested with trypsin (sequence grade), peptide Ala 466 -Asp 484 was suggested to be digested by a trace amount of contaminating chymotrypsin. Because other sites, which can be theoretically digested by chymotrypsin, were not digested, it is possible that the Phe 465 -Ala 466 site may be easily hydrolyzed by a slight chymotrypsin contamination. In fact, several charged positive ions of peak d in Fig. 4 were detected, and these were calculated to be of molecular mass 5,433 and 5,556 Da. Thus, they were assumed to represent peptide Gly 443 -Asp 484 with GPIanchored glycans carrying either GlcNAc-P or both GlcNAc-P and EtN-P as side chains, which should be theoretically released by trypsin (Fig. 4D). Thus, LC/ESI-MS analysis of trypsin-digested sAP suggested that the GPI-anchored glycan of sAP is hydrolyzed between GlcN and myo-inositol. To determine whether the reducing terminal residue is glucosamine, asialo-sAP was reduced directly with NaB 3 H 4 . Asialo-sAP, purified using a PD-10 column, was reproducibly tritium-labeled and the tritiated asialo-sAP bound to a PVL column (Fig. 5A). The radioactivity of the sAP Pronase digest was recovered in the PVL-bound fraction (Fig. 5B), indicating that the tritiumlabeled glycopeptide must be derived from a GPI-anchored glycan. An aliquot of the tritium-labeled pronase digest was hydrolyzed with 4 M HCl at 100°C for 3 h. After N-acetylation, the monosaccharide composition was analyzed on an SP0810 column, and this indicated that the 3 H-labeled monosaccharide corresponded to authentic N-acetylglucosaminitol (Fig.  5C). These results suggested that the reducing terminal residue of sAP is a glucosamine moiety, which may be cleaved, between glucosamine and inositol of membrane-bound AP, by endo-␣-glucosaminidase.
Periodate Oxidation of the Tritium-labeled Pronase Digest-We determined which hydroxyl group at C-6, C-3, or C-4 of the mannose is replaced by the ␤-linked GlcNAc phosphate using the periodate oxidation method (25). The tritium-labeled GPI-anchored glycans were periodate oxidized, and the excess oxidant was consumed by ethylene glycol. After reducing with NaBH 4 , the products were hydrolyzed and then reduced with NaB 3 H 4 . When the [ 3 H]-labeled components were seperated by pH 5.4 paper electrophoresis phosphorylated mannitol was not detected (Fig. 5D). These results indicated that GlcNAc-P is replaced at C-6 position of the mannose residue. If GlcNAc-P is replaced at C-3 or C-4 of the mannose residue, the periodateoxidized phosphorylated monosaccharide should be identified as phosphorylated mannitol. On the basis of the combined results of LC/ESI-MS, nitrous acid deamination, periodate oxidation, PVL column chromatography, chromatofocusing, and tritium labeling with NaB 3 H 4 reduction, the structures of the GPI-anchored glycans derived from sAP were determined and are summarized in Fig. 6.
GlcNAc-P Residues of Several Human GPI-anchored Glycoproteins-To investigate whether GlcNAc-P residues are also contained within other GPI-anchored glycoproteins, human serum cholinesterase, CEA, and T-H glycoprotein were applied to a PVL column. Most of these glycoproteins effectively bound to the column (Fig. 1). Because their N-glycan structures have been determined to be mature, at least to the point of incorporating N-acetyllactosamine residues (9 -12), the nonreducing terminal N-acetylglucosamine residue of these glycoproteins should be derived from the GPI-anchored glycan. These results suggested that GlcNAc-P residues are widely distributed in human cell surface GPI-anchored glycoproteins.
Proaerolysin Binding to sAP-Proaerolysin is a 52-kDa protein secreted by Aeromonas spp. and is a precursor of aerolysin, which is required for the virulence of A. hydrophila (13). All of the proteins that are known to bind aerolysin with high affinity are attached to the cell surface with GPI anchors. These include T-lymphocyte Thy-1, brain contactin, and variant surface glycoprotein of Trypanosoma spp. (14 -17). However, some GPI-anchored proteins cannot bind aerolysin because of species differences in modifications to the core structure (17). To determine whether the GlcNAc-P residue of the GPI-anchored glycan is required for binding to proaerolysin, binding ability was measured using both plate method and SPR assays.  7. Proaerolysin binding to sAP and its derivatives. sAP and its derivatives were coated on the 96-well plate. Biotinylated proaerolysin was added to the well and incubated at 37°C for 2 h. Bound proaerolysin was detected by avidin-peroxidase and its substrate. f, intact sAP; Ⅺ, asialo-sAP; q, acid-treated sAP; OE, acid-and phosphatase-treated sAP.
In binding assays using plates coated with sAP or asialo-sAP, biotin-labeled proaerolysin bound to sAP and asialo-sAP in the same dose-dependent manner (Fig. 7, f and Ⅺ, respectively). GlcNAc-P residues were released from the GPI-anchored glycan of sAP by sequential mild acid hydrolysis and acid phosphatase digestion. Mild acid hydrolysis (0.01 M HCl at 100°C for 30 min) should release both ␤-GlcNAc of GlcNAc-P diester residues as well as sialic acids of the sialylated biantennary glycan. The binding ability of proaerolysin to mild acidtreated sAP was decreased by approximately one-third (Fig. 7,  q). Diplococcal ␤-N-acetylhexosaminidase digestion also affected the binding ability of proaerolysin in the same manner (data not shown). Moreover, the release of GlcNAc-P residues from sAP completely abolished its binding to proaerolysin (Fig.  7, OE), although mild acid-and phosphatase-treated sAP showed the same molecular mass as that of intact sAP on SDS-PAGE (data not shown). These results suggested that the ␤GlcNAc-P residue in the GPI-anchored glycan is required for its high affinity binding to proaerolysin.
The binding of proaerolysin to sAP was also assayed using the SPR method. To compare the binding of proaerolysin with intact sAP and treated sAPs, the surface densities of the immobilized sAPs were kept at ϳ500 RU. Proaerolysin binding profiles to sAP and its derivatives are shown at various concentrations up to 300 nM (Fig. 8). Although the binding of proaerolysin to intact sAP and asialo-sAP increased in similar dose-dependent manners (Fig. 8, A and B), the binding to mild acid-treated sAP was decreased substantially (Fig. 8C), and the binding to GlcNAc-P-released sAP was extinguished (Fig. 8D). These results corresponded to those obtained using the plate method, and the K D value for binding of proaerolysin to sAP (5.6 ϫ 10 Ϫ8 M) was similar to the K D values of Thy-1, variant surface glycoprotein, and contactin reported by MacKenzie et al. (17) (Fig. 8). These results indicated that GlcNAc-P residues of GPI-anchored glycans are targets of proaerolysin. DISCUSSION A variety of membrane proteins are anchored to the cell surface via GPI (26). All GPI anchors, from yeast to mammals, have the same core structure consisting of EtN-PO 4 -6Man␣1-2Man␣1-6Man␣1-4GlcNH 2 ␣1-6myo-D-inositol-PO 4 linked to a lipid moiety (27). Using a hydrogen fluoride treatment method, which cleaves both GlcNAc-phosphodiester and ethanolamine phosphodiester, Redman et al. (8) reported that AP also has the same core structure. However, we observed that asialo-AP bound to a PVL column that recognizes ␤-GlcNAc, even though the N-glycan structures of AP are (NeuAc␣2Ϫ3) 2 Gal 2 GlcNAc 2 Man 3 GlcNAc(ϮFuc)GlcNAc (7). Accordingly, we investigated whether the GPI glycan of AP has ␤-GlcNAc residues. After sequential pronase digestion and nitrous acid deamination of sAP, the GPI-anchored glycan was labeled with NaB 3 H 4 . The reducing terminal residue was identified as 2,5-[ 3 H]anhydromannitol (data not shown). Because the tritium-labeled GPI-anchored glycan of sAP bound to a PVL column, a ␤GlcNAc residue must be attached to this glycan. Moreover, its binding diminished after mild acid hydrolysis or ␤-N-acetylhexosaminidase digestion. If a ␤GlcNAc residue is attached directly to mannosyl residues of the core structure, it should not be hydrolyzed under mild acid conditions. It was suggested, therefore, that the ␤GlcNAc residue is linked to mannosyl via phosphate, as has been reported for lysosomal enzymes (20), although the anomeric configuration of Glc-NAc-P was ␣. The results of chromatofocusing and LC/ESI-MS analysis in this study confirmed the existence of a ␤GlcNAc-P residue as a side chain of the GPI-anchored glycan.
The presence of GlcNAc-P residues in GPI-anchored glycans may be a common feature not only of human proteins but also of various mammalian, yeast, and fish proteins because various GPI-anchored glycoproteins bound to a PVL-Sepharose column and were eluted with 0.3 M GlcNAc (data not shown).
Moreover, it was demonstrated by LC/ESI-MS analysis that sAP is cleaved between glucosamine and inositol. These results suggest that GPI-anchored AP in human placenta is not hydrolyzed by PI-phospholipase C or D but by an endo-␣-glucosaminidase, although this enzyme remains to be identified. We are currently investigating whether this enzyme exists on the plasma membrane of human placental cells.
Although GPI-anchored glycans of surface glycoproteins had been reported to function as aerolysin receptors (28), the precise carbohydrate binding specificity remained unclear. We have demonstrated in this paper that the GlcNAc-P residue in the GPI-anchored glycan is a high affinity determinant for aerolysin binding. Hong et al. (29) also reported that N-glycans on GPI-anchored proteins influence aerolysin binding (29), but at least biantennary sugar chains of sAP were not bound to proaerolysin in our data. Although ␤-linked GlcNAc phosphate is replaced at C-6 of the mannose, and mild acid-phosphatase digestion abolished the interaction between sAP and proaerolysin, the binding could not be inhibited by 10 Ϫ5 M mannose 6-phosphate, inositol phosphate, GlcNAc phosphate, or ethanolamine phosphate (data not shown). These results suggest that the minimum structure of the GPI-anchored glycan which is required for recognition by aerolysin is a larger sized component than mannose 6-phosphate.