Glycolipid Precursors for the Membrane Anchor of Trypanosoma brucei Variant Surface Glycoproteins CAN STRUCTURE OF THE PHOSPHATIDYLINOSITOL-SPECIFIC PHOSPHOLIPASE C SENSITIVE AND

A number of eukaryotic surface glycoproteins, including the variant surface glycoproteins of Trypanosoma brucei, are synthesized with a carboxyl-terminal hydrophobic peptide extension that is cleaved and replaced by a complex glycosyl-phosphatidylinositol (GPI) membrane anchor within 1-5 min of the completion of polypeptide synthesis. The rapidity of this carboxyl-terminal modification suggests the existence of a prefabricated precursor glycolipid that can be transferred en bloc to the polypeptide. We have reported the purification and partial characterization of a candidate precursor glycolipid (P2) and of a compositionally similar glycolipid (P3) from T. brucei (Menon, A. K., Mayor, S., Ferguson, M. A. J., Duszenko, M., and Cross, G. A. M. (1988) J. Biol. Chem. 263, 1970-1977). The primary structure of the glycan portions of P2 and P3 have now been analyzed by a combination of selective chemical fragmentation and enzymatic glycan sequencing at the subnanomolar level. The glycans were generated by deamination, NaB3H4 reduction, and dephosphorylation of glycolipids purified from different trypanosome variants. Glycan fragments derived from biosynthetically labeled glycolipids were also analyzed. The cumulative data strongly suggest that P2 and P3 contain ethanolamine-phosphate-Man alpha 1-2Man alpha 1-6Man alpha 1-GlcN linked glycosidically to an inositol residue, as do all the GPI anchors that have been structurally characterized. The structural similarities suggest that GPI membrane anchors are derived from common precursor glycolipids that become variably modified during or after addition to newly synthesized proteins.

A number of eukaryotic surface glycoproteins, including the variant surface glycoproteins of Trypanosoma brucei, are synthesized with a carboxyl-terminal hydrophobic peptide extension that is cleaved and replaced by a complex glycosyl-phosphatidylinositol (GPI) membrane anchor within l-5 min of the completion of polypeptide synthesis.
The rapidity of this carboxyl-terminal modification suggests the existence of a prefabricated precursor glycolipid that can be transferred en bloc to the polypeptide.
We have reported the purification and partial characterization of a candidate precursor glycolipid (P2) and of a compositionally similar glycolipid (P3) from T. brucei (Menon, A. K., Mayor, S., Ferguson, M. A. J., Duszenko, M., and Cross, G. A. M. (1988) J. Biol. Chem. 263, 1970. The primary structure of the glycan portions of P2 and P3 have now been analyzed by a combination of selective chemical fragmentation and enzymatic glycan sequencing at the subnanomolar level. The glycans were generated by deamination, NaB3H4 reduction, and dephosphorylation of glycolipids purified from different trypanosome variants. Glycan fragments derived from biosynthetically labeled glycolipids were also analyzed.
The cumulative data strongly suggest that P2 and P3 contain ethanolamine-phosphate-Mana l -2ManLu l -6ManLu 1 -GlcN linked glycosidically to an inositol residue, as do all the GPI anchors that have been structurally characterized.
The structural similarities suggest that GPI membrane anchors are derived cram common precursor glycolipids that become variably modified during or after addition to newly synthesized proteins.
The covalent linkage of a glycosylinositol phospholipid to the carboxyl-terminal amino acid of many eukaryotic cell * This work was supported by National Institutes of Health Grant AI21531 (to A. K. M. and G. A. M. C.), the Lucille P. Markey Charitable Institute, Miami, FL, the Rockefeller University/Oxford University graduate student exchange program (to S. M.), and by the Wellcome Trust and Monsanto Co. (to M. A. J. F., R. A. D.. and T. W. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
surface glycoproteins provides the sole means of membrane attachment (Cross, 1987;Low, 1987;Low and Saltiel, 1988;Ferguson and Williams, 1988), and in many cases some fraction of these proteins can be released from the membranes by treatment with a phosphatidylinositol-specific phospholipase C (PI-PLC). ' Chemical structures of the glycosylinositol phospholipid (GPI) membrane anchors of a trypanosome variant surface glycoprotein (VSG) , rat brain Thy-l (Homans et al., 1988), and human erythrocyte acetylcholine esterase (Roberts et al., 1988a;Roberts et al., 198813) are known. These GPI anchors have a common backbone structure of ethanolamine-phosphate-6Mancrl-2Manal-GMancul-4GlcN linked al-6 to an inositol phospholipid. The carboxyl terminus of the mature protein is attached to the glycolipid via an amide linkage to ethanolamine. The GPI membrane anchors of decay accelerating factor from human erythrocytes (Medof et al., 1986), HeLa cell alkaline phosphatase (Jemmerson and Low, 1987), and human placental alkaline phosphate (Howard et al., 1987;Ogata et al., 1988) are not completely characterized, but current data suggest similar backbone structures.
The different anchors appear to have both protein and celltype specific modifications that are manifest as "decorations" branching from the core backbone structure and variations in the composition and linkage of fatty acids. For example, a variable a-mannose (attached to the mannose residue linked to ethanolamine phosphate) found in rat brain Thy-l (Homans et al., 1988) is absent from rat thymocyte Thy-l (Tse et al., 1985). Both the Thy-l anchors, however, have an extra ethanolamine phosphate attached to the mannose residue adjacent to glucosamine (Homans et al., 1988), a feature absent from the glycolipid anchors of trypanosome VSGs which have, instead, a variable number of a-galactose residues attached to the mannose adjacent to the glucosamine.
The extra ethanolamine phosphate may be an otherwise common feature of protein glycolipid anchors, since ethanolamine has been detected at a stoicbiometry of about 2 in human erythrocyte acetylcholine esterase, bovine erythrocyte acetylcholine esterase (Haas et al., 1986;Roberts et al., 1987), alkaline Glycan Structure of Glycolipid Anchor Precursors 6165 phosphatase , scrapie prion protein (Stahl et al., 1987), squid brain Sgp2  and decay accelerating factor (Medof et al., 1986).
cDNA sequence analyses of GPI-anchored proteins suggest that the proteins are synthesized as precursor polypeptides with an amino-terminal signal sequence and a hydrophobic carboxyl-terminal extension that is rapidly replaced by the glycolipid. The addition of the glycolipid anchor occurs within 1 min of completion of protein synthesis for the VSGs of Trypanosoma brucei (Bangs et al., 1985;Ferguson et al., 1986) and the neural cell adhesion molecule, N-CAM 120 (He et al., 1988), within 2 min for Thy-l (Conzelmann et al., 1987) and within 5 min for alkaline phosphatase . The rapidity of this modification suggested the existence of a prefabricated precursor glycolipid that could be transferred en bloc to the newly synthesized protein, probably in the endoplasmic reticulum. Recent studies using a yeast secretory mutant (secl8) support an endoplasmic reticulum location for the addition of the GPI anchor (Conzelmann et al., 1988). A glycolipid with properties consistent with a role as a VSG anchor precursor was subsequently identified in biosynthetic labeling experiments (Krakow et al., 1986;Menon et al., 1988a). The lipid, referred to as P2 (Menon et al., 1988a), presumably identical to lipid A described by Krakow et al. (1986) (Krakow et al., 1986;Menon et al., 1988a). The amino group of the ethanolamine residue in P2 was found to be unsubstituted and chemical analyses of subnanomolar quantities of P2 (purified by thin layer chromatography) confirmed the presence of ethanolamine and glucosamine in a 1:l molar ratio, as well as the presence of myristic acid and mannose (Menon et al., 1988a). Thin layer chromatography of polar lipid extracts containing P2 revealed another compositionally similar glycolipid, P3 (similar to lipid C, Krakow et al., 1986), apparently differing from P2 only by its somewhat greater hydrophobicity and its insusceptibility to bacterial PI-PLC (Menon et al., 1988a). The reason for the PI-PLC insensitivity of P3 is the subject of the accompanying paper (Mayor et al., 1990). In this paper we present a detailed analysis of the glycan portions of P2 and P3, purified from three T. brucei variants. The primary structure of the glycan portion of these glycolipids has been determined by a combination of selective chemical fragmentation and enzymatic glycan sequencing of subnanomolar amounts of material as well as by analyses of biosynthetically labeled material.  (Parekh et al., 1987), unless otherwise mentioned. All solvents were analytical or high performance liquid chromatography (HPLC) grade. A mixture of partially 0-methylated methylmannosides (2,3,4-triand 2,3,6-tri-0-methyl-o+methylmannosides) was a gift from Dr. S. J. Turco, University of Kentucky (Hull and Turco, 1985). The mixture was hydrolyzed as described below to generate 2,3,4-tri-and 2,3,6-tri-O-methyl mannose. rat blodd as previously described (Cross, 1975): Glvcoli~id Purification-Glvcolinids were uurified from -10" trvpanosomks as previously described (Menon kt al., 1988a) except that the lysed trypanosomes were centrifuged at 100,000 x g (instead of 45,000 x g). The 100,000 X g pellet was lyophilized and glycolipids extracted as before. Znositol AncZysis-Glycolipid samples purified by TLC and corresponding blank regions of the TLC plate were analyzed for total inositol content by gas chromatography-mass spectrometry (GC-MS) using selected ion monitoring (Smith et al., 1987). Samples were dissolved in methanol/pyridine/water (2:l:l) and aliquots (5 or 10%) were mixed with 20 pmol of scyllo-inositol internal standard, dried,  (Kobata and Amano, 1987), or treated with jack bean (Ymannosidase, or subjected to acetolysis, which selectively cleaves Mannl-6Man linkages (Rosenfeld and Ballou, 1974). The products of these reactions were analyzed by Bio-Gel P4 chromatography. Mannosidase digests were carried out in 25 11 of 0.1 M sodium acetate buffer, pH 5.0, under toluene withA. phoenicis oc-mannosidase I (20 *g/ml) for 2 h at room temperature or jack bean oc-mannosidase (30 units/ml) for 16 h at 37 "C. The reactions were stopped by heating 2 min at 100 "C, and the products were desalted by passage through 0.1 ml of AG50X12(H+) over 0.1 ml of AG3X4(OH-) and filtered through a 0.5-Frn Teflon filter prior to Bio-Gel P4 chromatography. Samples for acetolysis were per-acetylated in 40 11 of pyridine/ acetic anhydride (1:l) for 16 h at room temperature and dried. Acetolysis was performed in 30 ~1 of acetic anhydride/acetic acid/ sulfuric acid (1O:lO:l) for 6 h at 37 "C. The reaction was stopped by adding 10 ~1 of pyridine and the per-acetylated products were recovered in the chloroform phase following the addition of 400 ~1 of water and 200 ~1 of CHCl,.
The CHCla phase was washed three times with 400 ~1 of water.
Greater than 95% of the radioactivity was recovered in the organic phase.  (Cross, 1984) and treated with ice-cold aqueous HF (500 ~150%) for 60 h at 0 "C. HF was neutralized with saturated frozen LiOH and the samples centrifuged to remove the LiF precipitate and the insoluble protein residue. The supernatant was lyophilized, the residue resuspended in 500 11 of 10% trichloroacetic acid, kept on ice for 3 h to precipitate soluble protein, centrifuged, and the supernatant separated. The supernatant was passed over 3.0 ml of AG3X4 (OH-) and the eluate and washings pooled, filtered, and stored at -20 "C. Aliquots of this preparation were deaminated and reduced or N-acetylated as described below to generate the Gal,ManaAHM or Gal.Man3GlcNAcInos standards, respectively, where n = 0 for the glycan generated from sVSG 118 (Holder, 1985)' and n = 2-5 for the neutral glycans generated from sVSG 117 . Extraction and Purification of Biosynthetically Labeled Glyco-@ids-All labeling experiments were carried out at 37 "C in a shaking water bath for l-2 h as described in Mayor et al. (1989). In incubations involving tunicamycin, trypanosomes were preincubated with or without 400 rig/ml tunicamycin (Behring Diagnostics) for 30 min prior to the addition of the radiolabel. Biosynthetically labeled glycolipids were extracted and purified as follows.
At the end of an incubation the cells were centrifuged and the cell-pellet was first extracted with chloroform/methanol (2:l) and then with chloroform/methanol/ water (10:10:3).
The In each case the eluate and washings were pooled and dried, and the residue was flash evaporated with 3 x 50 ~1 of toluene, resuspended, filtered through a 0.2-pm filter, and stored at -20 "C until required for analysis. Samples were deaminated in 400 ~1 of deamination buffer (0.1 M NaAc, pH 3.5, 0.25 M NaNOz), and the reaction was terminated by adding 300 ~1 of 0.4 M HaBOa followed by 130 pl of 1 M NaOH. The samples were reduced at pH 10.0 for 12 h at ambient temperature by adding 100 ~1 of 2 M NaBH,, in 0.3 M NaOH.
After 5-8 h the reduction reaction was terminated by adding 20 ~1 of HAc and desalted on a lml column of AG50WX12(H+). The column eluate and washings were pooled and dried, the residue was flash evaporated with 5% HAc in methanol (2 x 200 11) and then with toluene, resuspended in 100 ~1 water, and passed over a mixed-bed ion-exchange column.
The pooled eluate and washings were filtered and stored at-20 "C until required. ethanolamine-labeled P2 and P3 were dried in 200-~1 glass microdispenser tubes. Samples were dansylated either prior to hydrolysis or after hydrolysis (6 M HCl, 8-10 h, 110 "C) and analyzed by thin layer chromatography as described earlier (Menon et al., 1988a). and washes were pooled. The chloroform phase was separated, washed with water (5 x 1 ml), and dried. The residue containing the permethylated glycans was hydrolyzed for 4 h at 120 "C in 200 ~1 of 2 N-trifluoroacetic acid (Pierce Chemical Co.). The hydrolysate was dried and the residual acid flash-evaporated with 3 X 100 ~1 of MeOH.

RESULTS
plates. Fig. 1 shows that [3H]Gl~N labeling of P2 and P3 is tunicamycin insensitiue, consistent with a role for these lipids in glycolipid anchor biosynthesis (Ferguson et al., 1986). Pl contains a glycan of the structure Man5GlcNAcz3 and [3H]GlcN-labeling of Pl is tunicamycin sensitive (Fig. l) and diacylglycerol . Glycolipid samples were incubated at 0 "C with aqueous HF for different periods of time to investigate the susceptibilities of the different phosphodiester linkages in the two molecules and to establish conditions for complete dephosphorylation. The ethanolamine-glycolipid linkage in P2 and P3 was quantitatively cleaved by aqueous HF within 24 h (Fig. 2, A and 0 HF cleavage of the phosphodiester bond linking diacylglycerol to the rest of the molecule was monitored by following the release of water-soluble radioactivity from [3H]Gl~Nlabeled lipids (Fig. 2). About 80% of the radioactivity in the [3H]GlcN-labe!ed P2 sample was rendered water-soluble after HF treatment for 60 h. Analysis of the material remaining in the butanol phase (~20% of the starting material) suggested that it was the partially dephosphorylated product (P2 minus ethanolamine phosphate), since it migrated ahead of the starting material on silica high performance-thin layer chromatography (compare Fig. 3 (Fig. 2B). TLC analysis of the butanol phase revealed two products, one migrating close to the starting material and another faster moving species (compare Fig. 3, A and B). The faster moving molecule (B) was shown in a separate experiment to be [3H]myristic acidlabeled (data not shown) and is likely to be the partially dephosphorylated product (P3 minus ethanolamine phosphate). The slower species, migrating only slightly ahead of the starting material, was not labeled with [3H]myristic acid (data not shown) and was assumed to be the completely dephosphorylated product (P3 minus ethanolamine phosphate and minus the phosphodiester-linked diglyceride moiety). These data suggest that the completely dephosphorylated P3 glycan contains a hydrophobic modification that cannot be removed by HF treatment. This hydrophobic modification can be removed by base treatment (Mayor et al., 1990).
Analysis of Glycans from Biosynthetically Labeled Glycolipids-After deamination and reduction, the dephosphorylated water-soluble material from a total butanol extract of [3H]GlcN-labeled trypanosomes was analyzed by anion exchange chromatography (Fig. 4). The major peak obtained cochromatographed with the Man3AHM neutral glycan standard. The tunicamycin sensitivity of the peaks with retention times > 20 min, which did not coelute with the variably galactosylated glycan standards, suggests that they are likely to be derived from lipid-linked oligosaccharides involved in N-linked glycan biosynthesis.
The analyses were repeated with TLC-purified lipids. sis of the deaminated-reduced material by Dionex HPLC produced a single peak, cochromatographing with Man3AHM (Fig. 5A). Identical results were obtained with the glycan from [3H]Man-labeled P3 (Fig. 9D). The N-acetylated material chromatographed as two peaks (Fig. 5B), one corresponding to the neutral glycan ManaGlcNAcInos (generated by Nacetylation of the glycan portion of the dephosphorylated glycolipid anchor of VSG 118) (Fig. 5, B and C), and a second peak at longer retention time (-36 min), which could be removed by passing the N-acetylated reaction mixture over an anion exchange column prior to HPLC analysis (compare Fig. 5, B and C). This second peak is probably derived from the partially dephosphorylated P3 molecule (P3 minus ethanolamine phosphate) identified in the TLC analysis of dephosphorylated [3H]GlcN-labeled P3 (Fig. 3B) and hence still carries a negative charge and the glycerol moiety.
Sequence Analysis of the Neutral Glycans Derived from P2 and P3-GC-MS analysis of TLC purified P2 and P3 samples identified myo-inositol in both glycolipids. 904 pmol of inositol were present in a sample of P2 purified from 10" trypanosomes (variant 11&J), and 150 pmol of inositol were present in a sample of P3 purified from a similar number of cells (variant 221). Assuming a stoichiometry of 1 mol of inositol/mol of glycolipid, the analyses represent recoveries corresponding to -lo4 and -lo3 molecules/trypanosome of P2 and P3, respectively. These figures are consistent with our previous estimates based on ethanolamine analysis and biosynthetic labeling experiments (Menon et al., 1988a). The analysis scheme used to determine the sequence of the glycans derived from the glycolipids (Fig. 6) was derived from the approach used to determine the structures of the glycan portions of the glycolipid membrane anchors of a trypanosome variant surface glycoprotein  and Thy-1 (Homans et al., 1988) glycans were generated by base treatment, nitrous acid deamination, NaB3H4 reduction and dephosphorylation of glycolipid samples, and chromatographed on high-resolution Bio-Gel P4 gel filtration columns. The radiochromatograms of the neutral glycans derived from the P2 glycolipid (variant clone 117) and from sVSG 118 showed a single neutral glycan species with a specific activity of -0.3 Ci/mmol, which chromatographed at 4.2 GU (Fig. 7). This is the same size as the authentic Manal-2Manal-GMancul-4(2,&anhydromannitol) species derived by a-galactosidase treatment of the neutral glycan from VSG 117 . The minor peaks observed at larger than 4.2 GU in the chromatogram of the glycan derived from sVSG 118 correspond to the trace amount of galactosylation observable in this variant.' Control experiments were performed to investigate the possible nonspecific destruction of glycosidic linkages during the cold aqueous HF dephosphorylation procedure. The deaminated and reduced glycan portion from the glycolipid membrane anchor of VSG 117 (dAR-gp,  contains two Galal-2Gal bonds, which are known to be relatively acid labile (Ferguson et al., 1985).* A sample of dAR-gp was treated with HF (50% aqueous HF, 65 h, 0 "C), neutralized, desalted, and reduced with NaBH4. The resulting reduced glycans were desalted, methanolyzed, converted to their Steps i-iii produce a labeled neutral glycan that can be further fragmented by enzymatic and chemical digestions: A. phoenicus ol-mannosidase I (Apdf) is an ol-mannosidase specific for ManLul-2Man linkages, acetolysis is specific for Manoll-6Man linkages, and jack bean a-mannosidase (J&M) is an exo-glycosidase that removes terminal oc-mannose residues.
The size of the expected products of the different treatments are given in glucose units (GU). Radioactivity (uertical axis) was measured in an aliquot (5-50 ~1) of 0.5 ml (2.5 min) column fractions and plotted against fraction number.
In each case the deaminated, NaB3H4-reduced and dephospborylated neutral glycan (4.2 GU peak material (filled circles, A)) was subjected to A. phoenicis a-mannosidase I digestion of acetolysis or jack bean a-mannosidase digestion, and the fragments generated were rechromatographed on Bio-Gel trimethylsilyl derivatives and analyzed by GC-MS as described earlier (Ferguson et al., 1985). Only a trace amount of galacitol and barely detectable amounts of mannitol were produced when compared with the untreated sample (data now shown), indicating that all of the glycosidic bonds in the mature GPI glycans were stable to prolonged HF treatment. Therefore, it appears that the HF dephosphorylation method preserves the integrity of the core glycans.
The column elution profiles (obtained by liquid scintillation counting of fractions collected) for the glycans derived from the glycolipids P2 (variant clone 118), P2 (variant clone 221), and P3 (pooled from variant clones 117, 118, and 221) are shown in Fig. 8A. All the samples yielded a single neutral glycan species with a specific activity of -0.3 Ci/mmol, which chromatographed at 4.2 GU. Peak fractions (4.2 GU, filled circles, Fig. 8A) were pooled for further analysis. Corresponding "blank" samples from each preparation, radiolabeled and analyzed as above, yielded no detectable peaks (data not shown). When an aliquot of the P2 (118) sample (A) was analyzed by Dionex HPLC (gradient program a), all the radioactivity chromatographed as a single peak coeluting with Man3AHM (data not shown). Neutral glycans (4.2 GU, filled circles, Fig. 8A) from P2 (118), P2 (221), and P3 (117, 118, and 221) were subjected to further analysis. On treatment with A. phoenicis cy-mannosi-dase I, the 4.2-GU material from each sample was completely digested to a 3.2-GU product (Fig. 8B), consistent with the removal of a single terminal mannose residue in a al-2 linkage to Man2AHM.
Acetolysis of the 4.2-GU neutral glycan from each lipid sample generated Bio-Gel P4 profiles (Fig. 8C) consistent with the presence of a Mancul-6Man linkage between the second and third mannose residues. The major product in each case (Fig. 8C) was a 2.4-GU molecule corresponding exactly to ManiAHM.
The two other species at 4.2 and 1.7 GU correspond to the starting material and AHM (see below), respectively.
Profiles identical to those shown in C were obtained on acetolysis of the defined Man3AHM derived from sVSG 118 (data not shown).
The observation of under-and over-digestion products in acetolysis studies is by no means unusual (see Natsuka et al., 1987Natsuka et al., , 1988, and the partial acetolysis conditions employed in these reactions were chosen to exploit the relative selectivity of acetolysis for the Mancul-6Man bond (Rosenfeld and Ballou, 1974). The conditions were optimized using the defined M,AHM glycan structure as follows. The reaction period was chosen so that a small portion of the starting material (Man3AHM, 4.2 GU) remained behind to show that no partial product at 3.2 GU (i.e. minus one mannose residue) was present. Also, conditions where all the 4.2-GU material disappeared were not used since, under these conditions, 2,5anhydromannitol (1.7 GU) was the major product. The appearance of 2,5anhydromannitol suggests that the  glycosidic bond is relatively labile to acetolysis compared with other glycosidic bonds. This is not surprising since, the Mancul-4(2,5-anhydromannitol) linkage in Man3AHM is an artificial and novel linkage whose susceptibility to acetolysis was previously unknown.
Digestion with jack bean a-mannosidase produced a characteristic 1.7-GU species (Fig. 8D) corresponding to the elution position of authentic 2,5-anhydromannitol. The 1.7 GU peak, in all cases, quantitatively cochromatographed with authentic 2,5-anhydromannitol in TLC and Dionex HPLC (using elution program a and b) analyses (data not shown).
These results strongly suggest that the sequence of the neutral glycan from P2 and P3 is  regardless of the trypanosome variant from which the glycolipids were obtained.
Methylation Analysis-TLC analyses of the 0-methylated mannose derivatives obtained after methylation and hydrolysis of [3H]Man-labeled, dephosphorylated, deaminated, and reduced glycans from P2 and P3 (Fig. 9, B and D) showed identical profiles consisting of three distinct partially methylated mannose species (Fig. 9, A and C). The three products in each case, 2,3,4-tri, 3,4,6-tri-, and 2,3,4,6-tetra-O-methyl mannose, were derived from 6-O-substituted, 2-O-substituted, and terminal mannose residues, respectively, in agreement with the MancYl-BManal-6Man sequence determined by enzymatic and chemical sequencing as described above. derived from P2 and P3, in agreement with the sequence determined as described above. The limiting amounts of available material prohibit further confirmation of these results by conventional methylation/GC-MS analysis on purified glycolipids.
Site of Ethanolamine Phosphate Linkage-Jack bean 01mannosidase requires an unsubstituted terminal mannose residue for activity but has little or no specificity for the aglycon to which the mannose residue is linked (Li and Li, 1972). The substrate specificity of jack bean a-mannosidase was exploited in experiments designed to determine the site of attachment of ethanolamine phosphate to the P2 and P3 glycans Homans et al., 1988).5 ["H]GlcN-labeled P2 and P3 were deaminated with nitrous acid, and the resulting water soluble, ["H]GlcN-labeled glycans were reduced with NaBH,. The HPLC profile in each case (Fig. 10, A and D) showed a single peak at -23.5 min, ahead of the Man3AHM standard (-14 min). The deaminated-reduced glycans (Fig. 10, A and D) were subjected to jack bean cu-mannosidase digestion and HF treatment and analyzed by Dionex HPLC. Jack bean o-mannosidase digestion after HF treatment resulted in the removal of all mannose residues to give 2,5-anhydromannitol (Fig. 10, C and F). Jack bean a-mannosidase digestion prior to HF treatment resulted in a product that cochromatographed with Man3AHM (Fig.  10, B and E). Linkage of ethanolamine phosphate to other than the terminal mannose residue in a linear structure or to a branched arrangement of mannose residues as proposed for the glycolipid anchor of VSG 121 (Schmitz et al., 1987), would have resulted in the removal of 1 or more mannose residues '  by jack bean cw-mannosidase digestion from the deaminatedreduced glycans prior to HF treatment, and the appearance of products (Man*AHM, ManlAHM, or 2,5anhydromannitol) other than ManzAHM.
These data strongly suggest that ethanolamine phosphate is linked to the terminal mannose in a linear arrangement of mannose residues in both P2 and P3, analogous to the position of ethanolamine phosphate in the glycolipid membrane anchor of membrane from VSG , Thy-l (Homans et al., 1988), and human erythrocyte acetylcholine esterase (Roberts et al., 1988b

DISCUSSION
The results presented in this paper show that the P2 and P3 glycolipids from different trypanosome variants have the same carbohydrate structure. The structure suggested by the data is ethanolamine-phosphate-Manal-2Manal-GMl-GlcN, identical to the core backbone structure found in the glycolipid anchors of VSG and Thy-l Homans et al., 1988 glycan standard when analyzed by Dionex HPLC. Even though cochromatography on the highly "structure sensitive" anion exchange Dionex HPLC cannot be taken as total proof of structure, it suggests that the dephosphorylated glycans from P2 and P3 have structures identical to the dephosphorylated glycan core of the VSG glycolipid anchor. To obtain additional support for these data, the dephosphorylated, deaminated, and reduced glycans from [3H]Man-labeled P2 and P3 were methylated and hydrolyzed, and the hydrolysate was analyzed by thin layer chromatography.
Three distinct methylated derivatives were resolved, consistent with the glycan sequence determined by the enzymatic/chemical sequence analyses. The stoichiometry of ethanolamine and glucosamine in P3 is 1:l (Menon et al., 1988a) and dansylation analyses of [3H] ethanolamine-labeled P2 and P3 showed that the ethanolamine amino group was unsubstituted in both lipids. The data from jack bean a-mannosidase digestion of the deaminatedreduced glycans from both lipids suggest that the ethanolamine-phosphate group is attached to the terminal mannose residue of a linear arrangement of the 3 mannose residues and, given the striking identity between the glycolipid structure and the VSG anchor, it is quite likely that the ethanolamine-phosphate group is attached to the 6-position of the terminal mannose residue. Methylation analysis of [3H]Man-labeled neutral glycans from P2 and P3 and the data from jack bean a-mannosidase digestion of the deaminated-reduced glycans from [3H]G1~Nlabeled P2 and P3 provide no evidence for an alternate arrangement of the mannose residues such as the branched mannose structure as proposed by Schmitz et al., (1987) for the glycolipid anchor of a VSG variant, VSG 121. Given the sensitivity of the analyses, alternate arrangements of the mannose residues would have been detected if they were present at 210% of the amount of the linear structure. The limiting amounts of available material prohibit further confirmation of these results by conventional methylation/GC-MS or NMR analyses.
The single structural difference between P2 and P3 has been identified as an ester-linked fatty acid(s) on the P3 inositol and is responsible for the PI-PLC insusceptibility of P3 (Mayor et al., 1990). Cold aqueous HF dephosphorylation followed by partitioning of the products between butanol and water was found to be a useful method of distinguishing between the structures (acyl-inositol uersus inositol) of the two glycolipids.
The identical glycan structure of the two glycolipids suggests a common biosynthetic route. Recently, in vitro systems that synthesize the glycolipids P2 and P3 have been described (Menon et al., 1988b;Masterson et al., 1989)6 and biosynthetic intermediates with a variable number of mannose residues (O-3) attached to "P2 like" (PI) and "P3 like" (acyl-PI) structures have been identified (Masterson et al., 1989).6 The two types of structures have been differentiated and characterized on the basis of their hydrophobicity after dephosphorylation.6 The glycolipid membrane anchors of the VSGs from the three subclasses of T. brucei, classified on the basis of carboxyl-terminal peptide sequence homology and the carboxylterminal amino acid of the mature protein (Holder, 1985), have a common backbone structure that is variably galactosylated depending on the subclass to which they belong 'Menon, A. K., Schwarz, R. T., Mayor, S., and Cross, G. A. M. (1990) J. Biol. Chem., in press. (Holder, 1985;.' The galactose residues are arranged in a variably branched structure attached to the mannose residue immediately adjacent to the glucosamine . In general, class I VSGs (e.g. VSG 117) contain 2-4 mol of galactose/mol of glycolipid, class II VSGs (e.g. VSG 221) contain 8 mol of galactose/mol of glycolipid, and the only known example of class III VSGs, VSG 118, contains no galactose in its glycolipid (Holder, 1985). Galactose addition could conceivably occur in at least two steps, with the core galactose residues ) added just before, or soon after glycolipid addition, and the terminal galactose residues added later in the secretory pathway. Recent data on the processing kinetics of the carboxyl-terminal glycopeptide from a class I VSG (ILTat 1.3) (Bangs et al. 1988) are consistent with this two-step hypothesis.
The common, variant-independent structure for P2 and P3 that we have presented in this paper strongly supports our previous proposal (Menon et al., 1988a) that the VSG glycolipid membrane anchor is prefabricated as a non-galactosylated glycolipid. However, minor amounts of tunicamycininsensitive [3H]GlcN-labeled glycolipids with dephosphorylated, deaminated, and reduced glycan moieties larger than Man3AHM and with retention times corresponding to galactosylated (GalZm6) Man3AHM have been observed,4 consistent with the possibility that a small fraction of the P2/P3 pool is galactosylated.
Thus, it is possible that a pre-guluctosylated glycolipid is transferred to VSG and further processed to generate the heterogeneous branched structure observed on the mature protein. Galactosylation may enhance the efficiency of transfer of the glycolipid to VSGs, similar to the effect of glucosylation on the transfer of the high mannose oligosaccharide, ManG_BGlcNAcz from the dolichol-linked precursor to proteins in mammalian systems .