Glycan Components in the Glycoinositol Phospholipid Anchor of Human Erythrocyte Acetylcholinesterase NOVEL FRAGMENTS PRODUCED BY TRIFLUOROACETIC ACID*

Inositol glycans were prepared from reductively radiomethylated human erythrocyte acetylcholinester- ase by sequential treatment with Proteinase K, methanolic KOH, and phosphatidylinositol-specific phospholipase C. Four glycans denoted a-6 were resolved by anion exchange high performance liquid chromatography (HPLC). Each glycan was subjected to hydrolysis in 4 M trifluoroacetic acid, and their hexose and hexose phosphate compositions were determined by anion exchange HPLC. The predominant glycan (Y showed a relative stoichiometry of 2 mannoses, 1 mannose 6-phosphate, 1 radiomethylated glu- cosamine, 1 radiomethylated ethanolamine, and 1 inositol. In contrast, the stoichiometry of glycan /3 was 1 mannose, 2 mannose 6-phosphates, 1 radiomethylated glucosamine, 2 radiomethylated ethanolamines, and 1 inositol. Glycans a and /3 were analyzed by electrospray ionization-mass spectrometry, and respective parent ions of m/x 1266 and 1417 were observed. The frag- mentation pattern produced by collision-induced dis-sociation mass spectrometry of these

Inositol glycans were prepared from reductively radiomethylated human erythrocyte acetylcholinesterase by sequential treatment with Proteinase K, methanolic KOH, and phosphatidylinositol-specific phospholipase C. Four glycans denoted a-6 were resolved by anion exchange high performance liquid chromatography (HPLC). Each glycan was subjected to hydrolysis in 4 M trifluoroacetic acid, and their hexose and hexose phosphate compositions were determined by anion exchange HPLC. The predominant glycan (Y showed a relative stoichiometry of 2 mannoses, 1 mannose 6-phosphate, 1 radiomethylated glucosamine, 1 radiomethylated ethanolamine, and 1 inositol. In contrast, the stoichiometry of glycan / 3 was 1 mannose, 2 mannose 6-phosphates, 1 radiomethylated glucosamine, 2 radiomethylated ethanolamines, and 1 inositol. Glycans a and /3 were analyzed by electrospray ionization-mass spectrometry, and respective parent ions of m/x 1266 and 1417 were observed. The fragmentation pattern produced by collision-induced dissociation mass spectrometry of these parent ions was consistent with a common linear core glycan sequence prior to radiomethylation of ethanolamine-phosphatemannosemannosemannoseglucosamineinositol.
Glycan (Y contained a single additional radiomethylated phosphoethanolamine branching from the mannose adjacent to glucosamine, whereas glycan /3 contained two additional radiomethylated phosphoethanolamines, one branching from each of the mannoses nearest to glucosamine. Trifluoroacetic acid hydrolysis did not cleave within the N,N-dimethylglucosamine-inositolphosphate moiety in these glycans, and this component was resolved by anion exchange HPLC and structurally confirmed by mass spectrometry. Dephosphorylation of this component by treatment with 50% HF produced N,N-dimethylglucosamine-inositol, and this conjugate was shown to have a characteristic elution time on cation exchange chromatography in an amino acid analyzer. Both of these fragments involving an intact radiomethylated glucosamine-inositol bond are proposed as new diagnostic indicators in the search for minor glycoinositol phospholipids in cells and tissues. * This work was supported by Grants NS16577 and DK38181 (to T. L. R.) and National Research Service Award DK08441 (to M. A. D.) from the National Institutes of Health and by grants from the Muscular Dystrophy Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
5 To whom correspondence should be addressed.
The membrane association of many integral membrane proteins is mediated at least in part by covalently attached lipid groups. Among intracellular proteins, the attachments can involve acylation of protein residues by the fatty acids myristate or palmitate (Sefton and Buss, 1987) or the prenylation of cysteine residues by farnesyl or geranylgeranyl groups (Maltese, 1990). In contrast, more than 50 identified extracellular membrane proteins are anchored exclusively by a glycoinositol phospholipid (GPI)' linked covalently to the protein C terminus (see reviews by Ferguson and Williams (1988), Rosenberry et al. (1989), and Cross (1990)). GPI anchors may be considered as a reasonably well defined class of structures because important elements are conserved over the wide range of phylogeny from protozoan parasites to mammals. Complete GPI structural determination involves a variety of analytical approaches to deduce the core oligosaccharide and inositol sequence and linkage positions, the lipid composition and structure, and the location and nature of all branching groups. Detailed information is available on the GPI anchors of four proteins, of which three are outlined in Fig. 1. The first complete structure was obtained for variant surface glycoprotein (VSG), an abundant class of surface proteins on TPypanosoma brucei (Ferguson et al., 1988). This anchor contains a linear core glycan sequence of ethanolamine-P04-6Manal -2Manal-6Mancrl-4GlcNH2a1-6-myoinositol-l-P04. This core sequence is completely conserved in the GPIs of rat Thy-1 (Homans et al., 1988) and Leishmania major promastigote surface protease (Schneider et al., 1990), and previously reported structural information on the GPI of human erythrocyte (Eh") acetylcholinesterase (AChE) is entirely consistent with it (Roberts et al., 1988b). A complicating factor in these analyses is the heterogeneity apparent with several GPI anchors. One source of GPI heterogeneity is the lipid composition. The VSG anchor contains only dimyristoylphosphatidylinositol (Ferguson et al., 1985), whereas GPIs of Leishmania promastigote surface protease and many mammalian GPI-anchored proteins, including those on erythrocyte The abbreviations used are: GPI, glycoinositol phospholipid; VSG, trypanosome variant surface glycoprotein; Ehu, human erythrocyte; AChE, acetylcholinesterase; PIPLC, phosphatidylinositolspecific phospholipase C; HPLC, high performance liquid chromatography; GC, gas chromatography; MS, mass spectrometry; FAB, fast atom bombardment; ESI, electrospray ionization; CID, collision-  AChE and decay-accelerating factor (Walter et al., 1990) contain almost exclusively 1-alkyl-2-acylinositol phospholipids. The AChE anchor lipid compositions have been analyzed in detail. In Eh" AChE, 83% of the l-alkyl-2-acylglycer-01s contain an 18:O or 181 alkyl group and a 22:4, 225, or 22:6 acyl group (Roberts et al., 1988a), whereas in bovine erythrocyte AChE, 69% of the 1-alkyl-2-acylglycerols contain an 18:O alkyl and an 18:O acyl group (Roberts et al., 1988~). Furthermore, the Eh" AChE anchor contains an additional fatty acid, primarily palmitate, that acylates an inositol hydroxyl group and renders the anchor resistant to cleavage by phosphatidylinositol-specific phospholipase C (PIPLC) (Roberts et al., 1988a(Roberts et al., , 1988b. In contrast to the mammalian AChEs, the inositol phospholipid moiety of torpedo AChE consisted almost exclusively of diacyl species, over 85% of which contained 16:0,180, or 18:l acyl chains in the sn-1 and sn-2 positions (Butikofer et al., 1990).
Additional contributors to GPI heterogeneity are groups that branch from the linear core glycan sequence. Many mammalian GPI anchors have a phosphoethanolamine moiety with a free amino group, and in Thy-1 this moiety is in phosphodiester linkage to the 2-hydroxyl on M a d 2 (Fig.  1). The VSG anchor lacks this substituent but contains a branched galactose oligosaccharide of variable size linked to the 3-hydroxyl of Manl. The rat brain Thy-1 GPI anchor has a single GalNAc linked to Manl and a fourth Man linked to Man3 (Fig. 1). The latter branching substituents are also found in some of the GPI anchor structures on human brain scrapie prion protein (Baldwin et al., 1990)  structures yet known to deviate from the core glycan sequence noted above are found in Leishmania, where both the GPI anchor of a lipophosphoglycan (Turco et al., 1989) and free GPI glycolipids (McConville et al., 1990) retain only the mannose-glucosamine-inositol-phosphate portion of the core but replace other residues toward the nonreducing terminus of the core with sugars that include galactose. Although the functional significance of this GPI heterogeneity is unclear, it may reflect subtle differences in interactions with other membrane components or in targeting to specific cell organelles.
The glycan structures of VSG and Thy-1 in Fig. 1 were established by a combination of two-dimensional proton NMR, GC-MS analysis of partially permethylated alditol acetates, and size analysis on Bio-Gel P4 in conjunction with HF and exoglycosidase cleavage. Smaller amounts of available Eh" AChE anchor have precluded NMR analysis, and the structure in Fig. 1 was deduced by lipid composition analyses and fast atom bombardment-mass spectrometry (FAB-MS). These procedures do not identify the hexose residues or the linkage positions in the AChE anchor glycan, and this report provides new information on these points. In addition, we continue to apply reductive radiomethylation to label free amine groups in the GlcN and ethanolamine residues of the AChE GPI anchor. Novel radiolabeled fragments produced by trifluoroacetic acid and HF cleavage of the radiomethylated GPI anchor are characterized. These procedures offer new criteria for the identification of GPIs in tissue extracts.

EXPERIMENTAL PROCEDURES
Materials-Bacillus thuringiensis PIPLC was purified3 from the culture medium of a Bacillus subtilis strain that had been transformed with the gene for B. thuringiensis PIPLC (Henner et al., 1988 kindly provided by Dr. Martin Low, Columbia University, New York).
Protein Purification and Radiolabeling-AChE was extracted from outdated human erythrocytes with Triton X-I00 and purified by affinity chromatography on acridinium resin as described in Rosenberry and Scoggin (19841, except that a more extensive resin wash was employed before enzyme elution (Gnagey et al., 1987). The purified Eh" AChE (5-200 nmol) in 0.1% Triton X-100 was concentrated 10-25-fold in a SpeedVac concentrator (Savant Instruments), reductively radiomethylated with 10-20 mM ['4C]HCH0 (ICN, either stock 56 mCi/mmol or 0.1-10 mCi/mmol by dilution with unlabeled HCHO) and 50-65 mM NaCNBH3 (Haas et al., 19861, and dialyzed extensively against 10 mM HEPES, 0.05% Triton X-100 (pH 7.0-7.5). The methyl group specific activity deriving from the stock ["C] HCHO was estimated to be 96 cpm/pmol (Haas and Rosenberry, 1985;Haas et al., 1986), and methyl group specific activities for GPI anchor fragments radiomethylated with diluted stocks were calculated from the measured ratio of incorporated label to inositol relative to the ratio for the corresponding fragments radiomethylated with undiluted stock.

"
Isolation of Intact Inositol Glvcans from Ehu AChE-To the dialyzed concentrate~bf radiomethylated Eh"' AChE and Triton X-100 in 10 mM HEPES was added Proteinase K (to 5 mg/ml) and CaCh (to 2 mM), and the mixture was incubated at 50 "C for 10-22 h. After addition of sodium dodecyl sulfate (to 1%) and more Proteinase K (additional 1 mg/ml), the digestion was continued at 50 'C for 5-17 h. The digestion mixture was applied to a Sephacryl s-200 column (1.5 X 80 cm) equilibrated in 10 mM HEPES (pH 7.5), 1 mM NaN3, and 0.05% Triton X-100. The fragment containing the C-terminal glycine residue linked to the GPI is the only fragment to associate with detergent micelles and elutes in the first peak of radioactivity at an elution volume of 0.55 relative to the solvent marker (Roberts et  al., 198813). Pooled fractions containing this peak were analyzed for Triton X-100 = 21.0), dried on the SpeedVac, suspended in methanol (2-3 ml) containing KOH (a 6-fold molar excess of KOH over HEPES) at 25 "C for 30 min to cleave ester-linked fatty acids but not alkylglycerol, and neutralized with glacial acetic acid. The dried sample was suspended in 20 mM sodium phosphate (pH 7.0), adsorbed to phenyl-Sepharose (0.1-0.5 ml resin/mg of Triton X-100 at 25 "C for 1 h), and washed twice with 1 resin volume of 20 mM sodium phosphate. The resin was suspended in an equal volume of 20 mM sodium phosphate (pH 7) and rocked with PIPLC (2-10 pg/ ml) at 37 "C for 1 h. The resin supernatant was removed, the packed resin was washed twice with 1 resin volume of 50 mM acetic acid, and the combined supernatant and washings containing the released radioactive GPI glycans were concentrated on the SpeedVac to <1 ml and desalted on Bio-Gel P2 equilibrated in water. The overall recovery of inositol at this point relative to that in the initial enzyme averaged 52 f 5% (S.D., for five preparations). The radioactive fractions were pooled, dried, and dissolved in water for anion exchange HPLC on a Dionex BioLC system (Hardy et al., 1988) that included a column (4.6 X 250 mm) of Dionex CarboPac PA-1 pellicular anion exchange resin and an AG-6 guard column operated at 1 ml/min at ambient temperature. Except where noted, collected fractions (0.5 ml) were neutralized by addition of 30 pl of glacial acetic acid.
Gel Exclusion Chromatography on Bio-Gel P2-Radiomethylated glycan fragments were desalted by chromatography on a Bio-Gel P2 column (1 X 90 cm). In columns equilibrated in water, the larger fragments a, p , 7 , and b were recovered in good yield (>95%), but recoveries of radiomethylated smaller fragments produced by trifluoroacetic acid or HF cleavage generally were low (<20%). Recoveries of these small fragments were improved (>95%) by equilibration of the column in 50 mM acetic acid.
Determination of Hexose and Hexose Phosphate Composition-Radiomethylated glycans isolated by anion exchange HPLC were desalted by chromatography on Bio-Gel P2 in water (Baxter, HPLC grade), and pooled dried fractions were hydrolyzed in sealed tubes (Kimble borosilicate glass prewashed with HNO3) containing 100 pl of trifluoroacetic acid at the indicated concentration at 100 "C for 4 h. Hydrolyzed samples were dried on the SpeedVac and dissolved in water for anion exchange Dionex PA-1 HPLC as described above. The column was pre-equilibrated with 15 mM NaOH (solution A) for 15 min prior to injection of the 100-pl sample, and the following elution protocol with the indicated linear gradient steps was found to give optimal resolution of neutral hexose and hexose phosphate standards: 0-15 min, isocratic solution A; 15-17 min, solution A to 100 mM NaOH plus 50 p~ sodium acetate (solution B); 17-25 min, solution B to 100 mM NaOH plus 100 mM sodium acetate (solution c ) ; 25-35 min, solution c to 100 mM NaOH plus 150 mM sodium acetate (solution D); 35-45 min, solution D to 100 mM NaOH plus 300 mM sodium acetate (solution E); 45-50 min, solution E to 100 mM NaOH plus 1.0 M sodium acetate (solution F); 50-55 min, isocratic 100 mM NaOH; 55-65 min, isocratic solution A with three cycles of injector valve closing/openings to clean the injection valve. The effluent from the HPLC column was combined with postcolumn addition of 300 mM NaOH (Hardy et al., 1988) and connected on-line to a pulsed amperometric detector (PAD) with a detector sensitivity of 1 pA full scale (pulse potentials E, = 0.05 V (t, = 480 ms), E2 = 0.60 V ( t 2 = 120 ms), E3 = -0.60 V (t3 = 60 ms)), and PAD data were integrated and stored for microcomputer analysis with Dionex software. After passing the PAD, the effluent was neutralized with 0.05 N H2S0, provided by a Dionex anion micromembrane suppressor.
Hydrolysis of standards at 1-4 M trifluoroacetic acid indicated no significant degradation of neutral hexoses but 13% loss of Man-6-PO, with 4 M trifluoroacetic acid. Correction factors were applied to compensate for hydrolysis blanks6 and for the degradation of Man-6-PO,.
Inositol Determination-To dried samples (10-250 pmol of inositol expected) were added scyllo-inositol as an internal standard, Tris-HC1 (1-2 pmol), and HCI (to 6 N, 200-500 al), and hydrolysis was conducted in sealed tubes at 115 "C for 12-16 h. Dried hydrolysates Throughout this report, values are listed with the standard error of the mean except where errors are designated as the standard deviation (S.D.).
were derivatized with 5% trimethylsilylchloride, 45% N,Obis(trimethylsi1yl)trifluoroacetamide in anhydrous pyridine (2-40 pl) at 65 "C for 1 h. Aliquots were analyzed by GC-MS on a Hewlett Packard system composed of a model 5890A gas chromatograph equipped with a splitless injector and a 15-m 0.20-mm inner diameter SPB-1 column (Supelco), a 5970B mass selective detector set for the characteristic ions m/z 305 and 318 (Smith et al., 1987), and a microcomputer work station.
Isolation of GPI Fragments Produced by Cleavage with HF-GPI samples were desalted on Bio-Gel P2, dried, and suspended in 50% HF (50 pl) at 0 "C for 60 h. Reactions were terminated by mixing with saturated LiOH followed by neutralization with NaHC03 (Ferguson et al., 1988). EthN(Me)2 was removed by drying samples at this point prior to desalting on Bio-Gel P2 (equilibrated in 50 mM acetic acid).
Moss Spectrometry-ESI-MS spectra were obtained on a triple quadrupole model TSQ-700 (Finnigan-MAT Corp, San Jose, CA) equipped with an electrospray ion source. Samples in water/methanol solutions were injected directly into the ESI chamber through a stainless steel hypodermic needle at a rate of 5-20 pl/min. A -3.5 kV difference between the needle tip and source electrode results in the expulsion of charged droplets, and ions with one or more charges may be generated from a single molecular species. Collision spectra (MS-CID-MS) on glycans a and p were obtained by selection of the doubly charged parent ion in the first quadrupole, CID in the second quadrupole, and product ion scanning in the third quadrupole. For collisions, the energy was offset at -15 eV using argon as the collision gas. Product ion abundance was enhanced with the doubly charged parent.

RESULTS
The structure of the Eh" AChE anchor in Fig. 1 was deduced from previous work in our laboratory that involved FAB-MS analysis of the radiomethylated GPI fragment obtained after complete pronase digestion and deacylation with base (Roberts et al., 1988b). Individual hexose residues in the glycan and their linkage positions could not be determined by this procedure. To isolate homogeneous Eh" AChE anchor glycans for further analysis, the proteolyzed and deacylated GPI fragments were digested with PIPLC to remove the 1-alkylglycerol residue and chromatographed on anion exchange HPLC as outlined in Fig. 2. Four peaks identified by PAD and containing radiolabel were obtained and designated CY, @, y, and 6. Individual hexose components of the predominant glycan CY were determined by hydrolysis with trifluoroacetic acid and anion exchange HPLC as shown in Fig. 3. The PAD trace in Fig. 3 0 shows three components, Glc, Man, and Man-6-P, that align with the standards in Fig. 3A. However, a similar amount of Glc was also present in the trifluoroacetic acid hydrolysis blank in Fig. 3C.' In addition to these three components, four other peaks denoted a-d are apparent in the PAD trace in Fig. 30. Fractions corresponding to peaks a, c, and d contained components labeled by radiomethylation as shown in Fig. 3E. Furthermore, GC-MS analysis indicated that virtually all of the inositol was in peaks c and d (Table  I). Peak b was not identified.
We have shown previously that the reductive radiomethylation procedure employed here converts free amine groups in AChE to their dimethylamine derivatives, and two components containing free amines, ethanolamine and GlcN, were identified in 6 N HC1 hydrolysates of the isolated GPI anchor by cation exchange chromatography on an amino acid analyzer (Haas et al., 1986). This is illustrated in Fig. 4A, where hydrolysates of Eh" AChE GPI fragments isolated after Proteinase K digestion are compared at various hydrolysis times. After 16 h, hydrolysates contained about equal amounts of radiomethylated ethanolamine and GlcN components, as observed previously (Haas et al., 1986) and as predicted from GlcN(Me)n and its hydrolysis product Xs decreased. This decrease was accompanied by an increase in two early peaks, one at 8 min and the second at 27 min (denoted X, in Haas et al., 1986). These observations suggested that GlcN(Me)z remained partially linked to other components under less rigorous acid hydrolysis conditions. This suggestion was confirmed by further analysis of the radiolabeled peaks in Fig.   3E by cation exchange chromatography. The label in peak a consisted exclusively of EthN(MeI2 with or without further acid hydrolysis, whereas the label in peaks c and d corresponded only to GlcN(Me), components after further hydrolysis in 6 N HC1 (Table I). Without further hydrolysis, peaks c and d were not retained on cation exchange chromatography (data not shown), as expected from their elution in the region of phosphate-containing standards in Fig. 3. Given the presence of inositol in peaks c and d and the structures in Fig. 1, it appeared likely that these peaks contained a trifluoroacetic acid-resistant core consisting of GlcN(Me)2-Inos-P.
To establish this trifluoroacetic acid-resistant core structure, peak c in Fig. 3E was treated with 50% HF to cleave phosphate ester bonds and subjected to cation exchange chromatography. A peak comigrating with X, was observed (Fig.   Fig. 3E, and the percentage of the total recovered inositol or amine in each fragment was determined. GC-MS analysis of 10-fraction pools from the chromatographic run in Fig. 3E revealed significant amounts of inositol only in fractions containing a, c, and d. GlcN(Me)* and EthN(Me)2 recoveries were calculated from the distribution of total radioactivity among the three labeled fragment pools in Fig. 3E and the amine contents of each pool determined by 16-h hydrolysis in 6 N HCl and cation exchange chromatography (Fig. 4).
4B) that contained equal amounts of labeled GlcN and inositol. Cleavage of glycan a first with HF and then with trifluoroacetic acid gave the same labeled product X, (Fig.  4C). Further analysis of peak c by ESI-MS revealed two prominent phosphorylated species, a major m/z 450.2 ion that corresponded precisely to Gl~N(Me)~-1nos-P and a minor m/ z 532.3 ion consistent with Man-Gl~N(Me)~-Inos-P. Peak c in Fig. 30 appears to be a fused peak, consistent with a mixture of these two components.
However, the MS and cation exchange chromatography data confirm that the predominant component in peak c is trifluoroacetic acid-resistant Gl~N(Me)~-1nos-P and indicate that its dephosphorylated derivative X, is GlcN(Me)*-Inos. These conclusions are summarized in Table 11.
Quantitation of glycan a components by trifluoroacetic acid hydrolysis and anion exchange HPLC as in Fig. 3 depended on the hydrolysis conditions. As the trifluoroacetic acid concentration was increased from 1.0 to 4.0 M, recoveries of Man and Man-6-P increased progressively and the ratio of peak c t o peak d also increased (Table 111). The late elution position of peak d suggests that phosphate groups are retained in addition to that in the Gl~N(Me)~-1nos-P core, and the structures in Fig. 1 indicate that this can occur only if these fragments also retain Man groups. Thus, it is likely that initial trifluoroacetic acid fragments in peak d are broken down to Man, Man-6-P, and peak c at higher trifluoroacetic acid concentrations. Since peak d is still detected even after 4 M trifluoroacetic acid hydrolysis (Fig. 3E), the stoichiometries of Man and Man-6-P in Table I11 are slight underestimates. Despite this limitation, the observed stoichiometries i n glycan a of nearly 2 residues of Man, 1 residue of Man-6-P, 1 residue of GlcN(Me)*, and 1 residue of EthN(Me)2 for each residue of inositol are in excellent agreement with the GPI anchor structures in Fig. 1.
Composition data for the other GPI fragments obtained from Eh" AChE in Fig. 2 are compared with that for glycan Q in Table IV. Glycan / 3 was eluted later than a during the HPLC run in Fig. 2, suggesting that it was more negatively charged, and the composition data support this point. Glycan p differs from a in containing 2 residues of Man-6-P, 1 residue of Man, and 2 residues of EthN(Me)z for each residue of inositol. These observations are consistent with an extra EthN(Me)2 phosphate linked to a Man 6-hydroxyl group, and t h e only 6-hydroxyl group available in the core glycan in Fig.  1 is on Man2.2 To confirm these structures for glycans a and /3, samples were analyzed by ESI-MS. Predominant ions with m/z of 634.1 and 709.7 were observed for a and /3, sates and other products from the Eh" AChE anchor on an amino acid analyzer. E"" AChE was radiomethylated with ["C] HCHO, and GPI fragments obtained at various stages of purification were treated as indicated and chromatographed as described previously (Haas and Rosenberry, 1985). Panel A, Proteinase K-digested samples from the first labeled peak off the Sephacryl S-200 column were hydrolyzed with 6 N HCl at 115 "C for the indicated times. Labeled GlcN(Me)2 and EthN(Me):, peaks were identified by elution relative to unlabeled standards, and the X, peak corresponded to a hydrolysis degradation product of GlcN(Me), reported previously (Haas et al., 1986). In the 16-h hydrolysate, labeled ethanolamine accounted for 52% and labeled glucosamine and X, accounted for 46% of the recovered radioactivity. Panel B, a sample corresponding to peak c in Fig. 3E was desalted by chromatography on Bio-Gel P2 equilibrated with 50 mM acetic acid, and an aliquot that contained 1.2 nmol of inositol was treated with 50% HF as outlined under "Experimental Procedures" and chromatographed. Based on the estimated methyl group specific activity, peak X,, contained 0.7 residues of GlcN(Me)2/residue of inositol. Panel C, a sample corresponding to peak a in Fig. 2 was completely dephosphorylated by treatment with 50% HF. The desalted sample was then run either directly (upper trace) or after hydrolysis in 4 M trifluoroacetic acid (lower trace).
TFA, trifluoroacetic acid. respectively, and the CID-MS spectra derived from these molecular ions are shown in Fig. 5. The major CID fragments are accounted for by rupture of Man glycosidic bonds or phosphodiester linkages as indicated. The mass difference between glycans a and /3 (1417 -1266 = 151) is accounted for by the additional EthN(MeI2 phosphate linked to Man2. Glycan y had a composition similar to a (data not shown), but it was such a minor component that additional character-  Dependence of glycan a fragmentation on trifluuroacetie acid eoneentration Glycan a was isolated by anion exchange chromatography as in Fig.2 and cleaved with trifluoroacetic acid. Components produced by the hydrolysis were quantitated by anion exchange chromatography as in Fig. 30. The data are means of two analyses from radiomethylated glycan a preparations of different specific activity. Each sample contained 1.5-2.0 nmol of inositol. A constant ratio of GlcN(MeL radioactivity to inositol in peaks c and d was observed within each a preparation, and the total inositol recovery was calculated from the recovery of radioactivity in peaks c and d. Methyl group specific activity estimates gave 0.8 & 0.1 residues of MezGlcN/residue of inositol, in reasonable agreement with the value of 1 expected from Fig. 1. The residues of EthN(Me)z were estimated as the ratio of radioactivity in peak a to that in the sum of peaks c and d (see Table  I), and the residues of Man and Man-6-P were determined by calibration with known standard amounts (Fig. 3

Composition of glycans released from EL AChE by PIPLC
Glycans were isolated by anion exchange chromatography as in Fig. 2 and cleaved with 4 M trifluoroacetic acid. Components produced by the hydrolysis were quantitated by anion exchange chromatography as in Table 111. The data are means of duplicate analyses. Data for glycan a are from Table 111 ization was not pursued. Peak 6 was eluted during the 1 M sodium acetate wash of the HPLC column in Fig. 2, and the basis of the apparent additional negative charge was not identified. This peak appears heterogeneous on Bio-Gel P2 (data not shown), an observation that could account for substoichiometric amounts of Gal and GalN in the 4 M trifluoroacetic acid hydrolysates (Table IV).

DISCUSSION
New structural information in this report permits us to account for some of the heterogeneity of the EhU AChE GPI anchor and to update the structure as shown in Fig. 6. The 3 hexose residues in the linear core glycan previously observed by FAB-MS (Roberts et al., 1988b)  Man3.' These features are consistent with those previously reported for VSG and Thy-1 in Fig. 1. The AChE glycans also all contain phosphoethanolamines with amine groups available for radiomethylation, and ESI-MS analysis of the radiomethylated CY and (3 glycans showed EthN(Me), phosphate located on Manl. A similar phosphoethanolamine is linked to the 2-position of the first Man in Thy-1. Glycan (3, involved in 10-20% of the AChE GPI anchors, includes a modification not previously documented in any GPI anchor: a second phosphoethanolamine with a phosphodiester linkage to the 6position of Man2. The linkage here is unequivocal, as trifluoroacetic acid hydrolysates demonstrated an increase of one Man-6-P and a decrease of one Man in (3 relative to CY. A glycan such as @ with two methylated ethanolamine phosphates was inferred from the initial FAB-MS studies (Roberts et al., 1988b), but the residue to which the additional phosphoethanolamine group was attached could not be deduced.
Although the position of phosphate linkage to Manl in the AChE glycans cannot be established from the available data, several observations are pertinent. First, composition data on glycan (3 following 4 M trifluoroacetic acid hydrolysis reveal Man residues (Table IV). Since control hydrolyses indicated that only about 6% of Man-6-P is degraded to Man during trifluoroacetic acid hydrolysis,6 this Man must have derived from a Man phosphate residue in (3 with a linkage more labile to trifluoroacetic acid than Man-6-P. Second, ESI-MS analysis of trifluoroacetic acid fragments in peak c detected a Man-GlcN(Me)z-Inos-P fragment apparently devoid of a Man phosphate linkage, suggesting that the phosphate linkage to Manl in the AChE glycans is more labile to trifluoroacetic acid than Man-6-P. Finally, unidentified peak b in Fig. 3 0 is eluted in the hexose phosphate region but contains no inositol or radiomethylated amine. This peak has the characteristics of a trifluoroacetic acid hydrolysis intermediate, as it is greater in 1 M trifluoroacetic acid than in 4 M trifluoroacetic acid. These are characteristics expected of a partially labile mannose phosphate that could represent Manl in the AChE glycans. Unfortunately, appropriate mannose phosphate standards are not available to test this assignment.
Completion of the Eh" AChE GPI anchor structure involves assignment of the core glycan linkages by permethylation and two-dimensional NMR analyses, techniques that require more purified glycan than that used in these studies. However, a major objective of this report is to introduce a new set of GPI fragments that can be used in identifying and characterizing GPI structures. The elegant studies of Ferguson and his colleagues (Ferguson et al., 1988;Ferguson, 1992) have provided several procedures that are diagnostic of GPI fragments, but most of the analyses involve complete dephosphorylation in 50% HF and characterization of the resulting neutral glycans. Direct fragmentation of GPIs by trifluoroacetic acid is shown in this report to provide a complementary approach that allows simultaneous Dionex anion exchange analysis of neutral hexoses, hexose monophosphates, and the glucosamine-inositol-phosphate core that many investigators consider to be the definitive feature of a GPI.
These trifluoroacetic acid fragmentation and anion exchange procedures have been used to quantitate the Man to Man-6-P ratios in GPIs biosynthetically labeled with [3H] Man. In addition, they are well suited to the detection and analysis of GPIs radiolabeled exogenously by radiomethylation (Deeg et al., 1992, accompanying paper). Since this procedure places a stable radiolabel on amine components, putative GPI candidates can be fragmented for informative anion exchange and cation exchange chromatography. For example, radiomethylated trifluoroacetic acid fragments that chromatograph at the position of GlcN(Me)2-Inos-P on anion exchange HPLC can subsequently be dephosphorylated with 50% HF and shown to chromatograph at the position of GlcN(Me)z-Inos on cation exchange columns (Fig. 4B). This two-step sequence provides compelling evidence for the identification of a GlcN(Me)z-Inos-P-containing GPI. The resolving power of cation exchange toward GPI components is high. In contrast to GlcN(Me)z-Inos, for example, the initial HFdephosphorylated product of a (predicted to be Man-Man-Man-GlcN(Me)z-Inos) was not retained on the cation exchange column (Fig. 4C). The absence of negatively charged groups in this product was confirmed by its lack of retention on the Dionex anion exchange column (data not shown).
Future technical refinements may permit even more structural information to be obtained by Dionex anion exchange analysis of radiomethylated GPI fragments obtained under limited trifluoroacetic acid hydrolysis conditions. For example, fragments in peak d were noted under "Results" to be consistent with species that contain phosphorylated Man residues still linked to GlcN(Me)z-Inos-P. The fact that peak d was eluted from the Dionex anion exchange column later than glycan CY (compare Figs. 3B and 3E) indicated that the peak d fragments were more negatively charged than glycan a, an observation consistent with the absence of radiomethylated ethanolamine in peak d. Initial cleavage of the two phosphodiester linkages between ethanolamine and phosphate in glycan a by trifluoroacetic acid would generate two mannose monophosphate moieties and increase the net negative charge by 2 in the residual glycan. Peak d elutes in a late, steep region of the acetate gradient that has not been optimized for maximal resolution, and it should be possible to resolve and identify these more highly phosphorylated radiomethylated trifluoroacetic acid fragments. Standards could then be established to deduce the number and location of ethanolamine phosphate groups in uncharacterized GPI species.