Characterization of a Glycophospholipid Intermediate in the Biosynthesis of Glycophosphatidylinositol Anchors Accumulating in the Thy- 1-negative Lymphoma Line SIA-b*

Several mammalian mutant cell lines are deficient in the biosynthesis of glycophosphatidylinositol anchors for membrane proteins. When metabolically labeled with [3H]rnyo-inositol or [SH]mannose, two out of five mutant lines (SIA-b and EL4-f) accumulated abnormal lipids which remained undetectable in the corresponding parental cell lines. The most abundant glycolipid of SIA-b cells (named lipid X) was isolated and par-tially characterized using hydrofluoric acid, nitrous acid deamination, acetolysis, and exoglycosidase treat- ments alone or in combination. The partial structure for the carbohydrate moiety of lipid X is Mana-(X+)Mana-GlcN-inositol, X being a charged, HF- sensitive substituent (possibly phosphoethanolamine). Lipid rendered the by with methanolic NH3, which suggests the of an on the The a glycophosphatidylinositol-anchored

Numerous membrane glycoproteins of eukaryotic organisms are anchored in the lipid bilayer by a glycophosphatidylinositol (GPI)' anchor and the potential functional roles of these anchors have been discussed in recent reviews (1)(2)(3)(4)(5). The structure of the GPI anchors of a variant surface glycoprotein from Trypanosoma brucei, of the mammalian Thy-l glycoprotein, of the human red blood cell acetylcholinesterase *This work was supported by grants from the Swiss National Foundation for Scientific Research No. 31-9069.87 and 31-9418.88. 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.
Several mammalian mutant cell lines which are deficient in the biosynthesis of GPI anchors have been obtained through selection for cells which are deficient in the surface expression of GPI-anchored plasma membrane glycoproteins (23-28). The lymphoma mutants studied in our lab fall into six complementation classes, five of which make normal amounts of the Thy-1 precursor intracellularly but fail to remove the hydrophobic C-terminal peptide and rapidly degrade the Thy-l precursor in the endoplasmic reticulum. Class E mutants are deficient in dolicholphosphomannose synthase and cannot make GPI anchors since dolicholphosphomannose is required for the synthesis of the anchor glycolipid (22,24,26,(29)(30)(31). The exact defect in the other mutant lines is at present unknown. Here we provide evidence that some of these mutants are deficient in the biosynthesis of the anchor glycolipid.

MATERIALS AND METHODS
Reagents were obtained from the sources described recently (18). Pure PI-PLC from B. cereus was from Boehringer Mannheim. PLD from Streptomyces chromofuscus was obtained from Sigma. Jack bean a-mannosidase was a gift from Dr. S. Kornfeld, Washington University, St Louis, MO. a-Mannosidase from Aspergillus phoenicis was purchased from the Oxford GlycoSystems, Oxford, United Kingdom. Methanolic NH, was from Janssen Chimica, Geel, Belgium.
Cells, Culture Conditions, and Radiolabeling-The murine Thy-lnegative mutant lymphoma lines and corresponding parental lines were the ones described previously (26,27). Cells were grown in Dulbecco's modified Eagle's medium plus 5% heat-inactivated horse serum and 20 pg/ml of gentamicin at 37 "C in a humidified atmosphere containing 5% COZ. For labeling with [2-3H]myo-inositol, the cells were washed and resuspended in inositol-free medium at IO6 cells/ml (with serum dialyzed against Hanks' balanced salt solution); [3H]myo-inositol was added as an aqueous solution and cells were incubated at 37 "C. After labeling cells were washed, and labeled lipids were extracted twice for 45 min using CHC13/CH30H/Hz0 (1010:3). The lipid extracts were dried under nitrogen and were stored in butanol or CHC13/CH30H/Hz0 (10103) at -20 "C.
Analysis of Lipids-Lipids were desalted by partitioning between butanol and water as described (11). Ascending TLC was performed 21051 on 0.2-mm-thick Silica Gel 60 plates (Merck) using CHCl3/CH3OH/ 0.25% KC1 in water (55:45:10) as a solvent. The developed TLC plates were sprayed with EN3HANCE (Du Pont-New England Nuclear) and fluorograms were obtained using X-OMAT film (Kodak) exposed for 3-21 days at -80 "C. For preparative purposes, lipids were scraped from TLC plates and eluted from the silica with CHC13/CH30H/Hz0 (10:10:3). HPLC separation of lipids was performed as previously described (32), except that we used a Radial Pak cartridge containing silica gel particles (10 pm in diameter) (Waters Associates, Inc.), and that the column was run at room temperature. The fractions were neutralized by the addition of acetic acid. Fractions containing radioactivity were pooled, the solvent was evaporated under Nz gas, and the lipids were desalted by phase separation in n-butanol and water (11). hydrolysis using 0.1 M NaOH in CHC13/CH30H/Hz0 (1010:3) for 30 Lipid mixtures or isolated lipids were deacylated by mild alkaline min at 37 "C as described (33). For treatments of lipids with phospholipases, lipids were dried in a Speed-Vac evaporator and then incubated as follows: PI-PLC from B. cereus, 20 mM Tris-HC1, pH 7.4, 0.1% TX-100, 0.2-1 mM EDTA (50 pl/tube) for 1-2 h at 30 "C with 0.05 unit of enzyme; PLD from S. chromofuscus, 20 mM Tris-HCI, pH 8.0, 0.2 mM CaClZ, 0.05% TX-100, 0.02% NaN3, 20 mM NaCl, (50 plltube) for 16 h at 30 "C with 50 units of enzyme. Treatments with the PIG-PLD contained in serum were done as described (34) using 1 p1 of either fresh human serum, thawed horse serum, or fetal calf serum in a final volume of 50 pl for 1 h at 37 "C. Removal of acyl chains linked to the inositol with methanolic NH3 was performed as described (8).
For analysis of the hydrophobicity of the fragments generated by phospholipase and base treatments, samples were dried in the Speed-Vac evaporator and resuspended by bath sonication in 0.5 ml of 1% TX-114 in 150 mM NaCl, 10 mM Tris-HC1, pH 7.4. After warming to 37 "C, phases were separated by centrifugation (10,000 X g X 20 s), the upper, aqueous phase was transferred to a new vial and radioactivity determined in both phases using HiSafe I1 (Pharmacia LKB Biotechnology Inc.) for scintillation counting. Phase separations with either [3H]myo-inositol or [3H]PI were performed for control in most experiments and showed that 1% of PI and 90% of [3H]myo-inositol partitioned into the aqueous phase. Thus, all values were corrected using the following formula: fraction of hydrophilic fragments (in %) Nitrous acid deamination of lipid extracts was performed as described (8). Treated lipid extracts were then extracted with watersaturated butanol and chromatographed on TLC plates.
Analysis of Headgroups-Headgroups were prepared from purified lipids either using phospholipase D or phospholipase C or alternatively by treatment with hydrofluoric acid as described (6). Before or after treatments, acyl chains attached to the inositol were removed by treatment with NH3 (8) or NaOH and lipids were extracted by butanol extraction. The hydrophilic headgroups were desalted either by gel filtration on a P2 Bio-Gel (Bio-Rad) column or by passing them over mixed ion-exchange minicolumns containing Chelex 100, Dowex 50, AG3, and QAE-Sephadex A50 (6). Headgroups were Nacetylated in 50 pl of methanol, 5 pl of pyridine, and 5 pl of acetic anhydride for 15 min at room temperature. N-Acetylated headgroups were treated with jack bean a-mannosidase (0.2 unit) in 50 pl of 5 mM sodium acetate, pH 4.5, 1 mM ZnClz, 0.01% bovine serum albumin, 3 mM NaN3 at 37 "C for 16 h or with a-mannosidase from A. phoenicis (10 microunits) in 10 pl of 100 mM sodium acetate, pH 5.0, at 37 "C for 20 h. If [3H]mannose-labeled headgroups were analyzed, the headgroups were deaminated with nitrous acid as described (8) and were then reduced with NaBH, as described (6). Acetolysis of headgroups was done as described (10).
The Gl~NAc-[~H]inositol standard was obtained through similar procedures from a GPI intermediate accumulating at 37 "C in the sec53 mutant of S. cereuisiae (29).
Purification and Radioiodinution of Thy-1-Thymocytes obtained from 6-week-old female Balb/c mice or lymphoma cells obtained through tissue culture were washed and detergent lysed at 5 X 10' cells/ml (lymphomas) or lo8 cells/ml (thymocytes) in prepurified TX-114 (36) as described (26) except that the lysis buffer additionally contained 2 pg/ml of 2,6-di-tert-butyl-p-cresol. After removal of cytosolic proteins by phase separation and one additional re-extraction, the detergent phase was diluted to the original lysis volume and Triton X-100 was added to a final concentration of 0.5%. Thy-1 was isolated from this membrane protein preparation by immunoabsorption using a monoclonal rat anti-mouse Thy-1 antibody (antibody 111-5 (37)) coupled directly onto cyanogen bromide-activated Sepharose beads. After incubation for 16 h at 7 "C on a wheel, the beads were washed as described previously (38), and Thy-1 was eluted by a 3-min incubation at 0 "C in 0.1 M glycine, pH 2.6, 0.05% TX-114 followed by a similar incubation at pH 2.3, 0 "C, followed by a final 3-min incubation at pH 2.3 and room temperature. Desorbed Thy-1 was neutralized immediately with NaHC03 and Tris buffer. Aliquots of 250 pl of the pooled eluates corresponding to the Thy-1 from 250-450 X lo6 of cells were radioiodinated with Na"'I using 10 pg of iodogen (Pierce) as a catalyst (39). (No prior dialysis of the immunoprecipitate was required prior to the labeling procedure). After transfer to fresh tubes, 1 ml of buffer A (1% TX-114, 20 pg/ml leupeptin, 10 pg/ml pepstatin, 10 pg/ml antipain, 1 mM EDTA, 0.1 mM cysteine, and 10 mM KI) was added. Phases were separated and the aqueous phase (containing free lZ5I) was discarded. lZ5I was further removed by reextracting the detergent pellet several times with buffer B (buffer A without TX-114). Radioiodinated Thy-1 was then immunoprecipitated once more using a rabbit anti-mouse Thy-1 antiserum and protein A-Sepharose and reeluted from the affinity sorbent at low pH as described above.
Analysis of Lipid Components of Radioiodinated Thy-1-For mild base hydrolysis, aliquots of 1z51-Thy-l were resuspended in 0.1 ml of 0.1 M NaOH in HzO plus 0.05% TX-100 or, for control, in 0.05% TX-100 plus 10 mM Tris-HC1, pH 7.4 and left at 37 "C for 1 h. Thereafter, samples were neutralized and diluted to 1 ml to have final concentrations of 17 mM of Tris-HC1 and 105 mM of NaC1. Tubes were left for 1 h at room temperature to allow for renaturation of lZ51-Thy-l. TX-114 was then added to a final concentration of 0.7%, and tubes were left on ice for 1 h to allow for equilibration of detergent between micelles. Phases were then separated by warming to 37 "C and supernatants and pellets were counted separately in a y-counter. Values were corrected based on controls containing either [3H]myo-inositol or [3H]PI as described above. For PI-PLC treatments, aliquots of lz5I-Thy-1 were resuspended in 30 pl of 250 mM a-methylmannoside, 20 mM Tris-HC1, pH 7.5, 10 mM NaCl, 0.2 mM EDTA, and 0.1% TX-100. After incubation with enzyme, tubes received 500 p1 of 1% TX-114 in 150 mM NaC1, 10 mM Tris-HC1, pH 7.4. After standing on ice for 1 h, phases were separated and counted.

Metabolic Labeling of Lipids in Parental and Mutant Lymphoma Lines with [3H]myo-Inositol-Mutant cell lines and corresponding parental cells were metabolically labeled with
[3H]myo-inositol, a tracer which is expected to label all potential intermediates of the GPI anchor biosynthesis. Analysis of the lipid extract from labeled cells by TLC showed that the major spot comigrated with PI but several less abundant, more polar lipids could be observed as well (Fig. 1, lipids a,b,c). Based on similar metabolic studies in T. brucei, the complete anchor glycolipid is expected to run very close to the origin in this TLC system (11). However, only minor and relatively diffuse bands could be observed in this region, and no significant difference between parental and mutant cells was observed. One reason for the absence of distinct and reproducible bands in this region might derive from the fact that mam-  Rate of Accumulation of Lipid X-The incorporation of [3H] myo-inositol into lipids was linear over 24 h and amounted usually to 5-10% of radioactivity added to the cells after 24 h. As illustrated in Fig. 3, lipid X from SIA-b increased in intensity in parallel with other lipid species over extended labeling periods, suggesting that lipid X was not turning over more rapidly than the bulk of [3H]myo-inositol-labeled lipids. When labeling for longer than 24 h we also observed faint bands in SIA-b which were more polar than lipid X and which could not be seen in the parental SIA line (Fig. 3). myo-Inositol was added to the preculture medium in an attempt to down-regulate endogenous myo-inositol production by the cells. This however did not increase subsequent incorporation of [3H]myo-inositol (not shown).
Isolation of Labeled Lipids from SIA-b-HPLC of [3H]myoinositol-labeled lipids from SIA-b widely separated PI, lipid a and lipid b but the fractions eluting between lipids a and b contained a mixture of several lipids rather than pure lipid X. Therefore, radiochemically pure lipid X was prepared by scraping the relevant zone from a preparative TLC plate. The eluted lipid X ran to the same position upon TLC analysis as in lipid extracts (see Fig. 7B).   The isolated lipid X migrated to its original position as shown in Fig.  7, lane 6. An acyl chain bound to the inositol residue of lipid X (see below) was removed with methanolic NH3 and headgroups were then cleaved off by HF treatment. One aliquot of the thus liberated headgroups was N-acetylated and directly analyzed by paper chromatography ( B ) . The remainder was passed over a diethyl(2-hydrox-ypropy1)aminoethyl-Sephadex column to remove remaining fluoride ions and treated with nitrous acid, reduced with NaBH4, and desalted by passing over an AG-50 column followed by repeated acetic acid/ methanol evaporation (6). After a final desalting step the material was analyzed by paper chromatography (system 12:5:4) either directly (C) or after digestion with a-mannosidase from A. phoenicis (APAM, E ) or from jack bean (JBAM, F ) . All  treated with HNOz and NaBH4, a procedure which cleaves the GlcNal,6myo-inosito1 link of classical GPI anchors and transforms the GlcN into 2,5-anhydromannitol (6). The labeled fragments thus generated comigrated with a Manal,6Manal,4anhydromannitol standard from GPI anchors in two different solvent systems (Fig. 4C, data not shown). Treatment of the HNOZ/NaBH4 fragment with jack bean a-mannosidase reduced it to [3H]mannose and a peak comigrating with Mana1,4anhydromannitol (AM1) in two different solvent systems (Fig. 4F, data not shown). The latter peak probably indicates that the a-mannosidase treatment was incomplete. a-Mannosidase from A. phoenicis, an exoglycosidase specific for al,2-linked mannoses, did not cleave the HF-generated headgroup X (Fig. 4E), although it readily removed one mannose residue from the anhydromannitol-Man3 standard (not shown).
Analysis of [3H]myo-Zmsitol-lled Headgroups of Lipids a, 6, and X from SZA-b-Headgroups from several PLDgenerated, [3H]myo-inositol-labeled lipids were analyzed by paper chromatography (summarized in Table I). Headgroups of lipid a migrated as myo-inositol. Since lipid a comigrates  (lipids a and b) or PIG-PLD (lipids X and c) followed by treatment with NaOH to remove remaining acyl chains (see "Results"). The headgroups were treated with nitrous acid or control incubated without nitrous acid. After desalting on mixed bed ion exchange minicolumns products were analyzed by paper chromatography and found to comigrate with the standards given in this A different result was obtained when phospholipase-generated headgroup X was treated with HF, since the HFtreated, N-acetylated headgroup X was uncharged and could be analyzed by paper chromatography, as shown in Fig. 5B (similar migration as the HF-generated, [3H]mannose-labeled headgroup X (Fig. 4B)). Thus, the charged substituent of headgroup X was removed by HF treatment. A second peak comigrating with myo-inositol (Fig. 5B) and seen in several independent experiments might have been generated by HF cleavage of the probable GlcN-myo-inositol bond of lipid X. When the N-acetylated headgroup X was treated with jack bean a-mannosidase, the product comigrated with the GlcNAcal,6inositol standard, confirming the presence of alinked mannoses in headgroup X (Fig. 5C). If headgroups X were treated with jack bean a-mannosidase before the HF treatment, yet another species was generated which migrated, after N-acetylation, intermediate between the untreated headgroup and the GlcNAcal,6inositol standard (Fig. 50). Since lipid X contains two mannoses only, this peak must be ManlGlcN-inositol. These results indicate that one of the two mannose residues of headgroup X can be removed by the exomannosidase, whereas the second mannose carries an HFsensitive group which blocks its removal by the a-mannosidase. This same group might carry the charges found on PLDgenerated headgroup X although formally, other groups might contribute to the charge as well. In the context of GPI anchors, we are tempted to assume that an internal mannose is substituted by a phosphoethanolamine group since ethanolamine has been found to be linked via a phosphodiester to the C2 position of the al,4-linked mannose of the only so far characterized mammalian GPI anchor, namely the GPI anchor of rat brain Thy-1 (7). The mature anchor of Thy-1 also contains a @1,4-linked GalNAc (7) which, based on our data, is certainly not present on lipid X. [3H]myo-inositol-labeled lipid X was deacylated with NHs and treated with HF and the free headgroups were N-acetylated. One-third of the material was used for paper chromatography directly (A, 400 cpm spotted), whereas the remainder was subjected to acetolysis followed by N-acetylation ( E , 1000 cpm spotted). Solvent system, 12:5:4; standards, 2 = ManzGlcNAc-inositol; 1 = ManlGlcNAc-inositol; 0 = GlcNAc-inositol; My0 = myo-inositol. Standards 2 and 1 are the ones generated in the experiment shown in Fig. 5, B and D

, respectively.
To obtain further information, the HF-generated, [ 3H] myo-inositol-labeled headgroup X was subjected to acetolysis, a procedure which preferentially cleaves al,6-linked mannoses. As expected, the starting material containing two mannoses migrated only slowly on paper (Fig. 6A). Acetolysis generated a new peak (Fig. 6B) which comigrated with the fragment generated by jack bean a-mannosidase and subsequent HF treatment of the headgroup X and which was identified as Man,GlcN-inositol (Fig. 50). This suggests that one of the two mannoses of headgroup X is unsubstituted and is linked a1,6, whereas the other mannose is linked differently. Thus, assuming that lipid X is a normal intermediate of the GPI anchor biosynthesis pathway, all the data are compatible with its headgroup having the structure Mana1,6(ethanolamine-P~)Manal,4GlcNa1,6myo-inositol.
Sensitivity of Lipids a,b,c, and X from SIA-b to Phospholipases and Mild Alkaline Hydrolysis-Purified lipids were subjected to mild alkaline hydrolysis and phospholipase treat-ments, and the resulting fragments were partitioned in the TX-114 phase separation system, in which hydrophobic components partition into the detergent phase, whereas hydrophilic ones stay in the aqueous phase (Table 11). Lipids X, b, and c were largely resistant to PI-PLC and phosphatidylinositolglycan-specific PLD (PIG-PLD). Interestingly however, partial deacylation with NH3 followed by PI-PLC treatment resulted in almost complete release of the label from lipids X and c into the aqueous phase, although NaOH on its own released significantly less of the counts. Near to complete release of label was also achieved when PLD or PIG-PLD treatments of lipids X or c, although inefficient on their own, were followed by NaOH or NH3 treatments. The behavior of lipids X and c was consistent with the presence of a NH3sensitive acyl chain on the inositol which either prevented the cleavage of the phosphodiester bond by PI-PLC (but not by PIG-PLD) or which retained the headgroup in the TX-114 detergent phase after removal of the diradylglycerol by PI-PLC. To distinguish between these two possibilities, experiments as depicted in Table I11 were performed. We consistently found that NH3 treatment released a higher fraction of counts from PI-PLC-treated lipid X than from untreated lipid X but that even more counts were released if the PI-PLC treatment was performed after the NH3 treatment. This was independent of the concentration of PI-PLC (Table 111). On the other hand, TLC analysis of complete lipid mixtures after PI-PLC treatment confirmed that a fraction of lipid X resisted PI-PLC treatment (Fig. 7, lane 4 ) . These findings suggest to us that acyls are attached in more than one position on the inositol of lipid X and that acyls block PI-PLC only if attached in some but not in other positions. The presence of an acyl on the inositol was confirmed also by the finding that HF-generated headgroups of lipid X partitioned into the TX-114 detergent phase as long as they were not treated with NHs (not shown). PI and lipids a (lyso-PI) and b (PIP) were largely sensitive to deacylation by mild base treatment but only a fraction of lipids X and c were sensitive to mild base (Table 11). TLC analysis of complete lipid mixtures showed that NaOH treatment removed all of lipid X from its original position to a position close to the origin (Fig. 7, lane 3). This indicated that most of lipid X molecules contained base labile acyl chains. The partial retention of NaOH-treated lipids X and c in the TX-114 detergent phase can not be explained by the acyl chain on the inositol which is mild base-sensitive (8,9) but indicates the presence of a base-resistant component on a fraction of these lipids which most likely consists of a 1-alkyl-2-acylglycerol.
Analysis of Lipid Moieties of GPI Anchors of Thy-1-The partial structure of the carbohydrate moiety of lipid X being compatible with the idea of its being a biosynthetic intermediate in anchor biosynthesis, we were amazed to find a base-resistant lipid component on part of lipid X since abaseresistant lipid component had not been recognized previously during a gas chromatography-mass spectrometry analysis of Thy-1 from rat brain and rat thymocytes (41). Therefore, we were interested to see whether the parental SIA cells from which the SIA-b mutant was derived, contained any baseresistant GPI anchors. To this end, microgram quantities of Thy-1 were purified from SIA and EL4 lymphomas and from normal mouse thymocytes. Murine Thy-1 contains tyrosine residues in positions 57, 68, 85, and 106 while the anchor is attached to cysteine 112. After radioiodinating tyrosines and a further round of immunoprecipitation of Thy-1, the final product was analyzed by SDS-PAGE/autoradiography and found to consist of a broad 24-30-kDa band which was shifted to 21-27 kDa after the single high-mannose type N-glycan Glycophosphatidylinositol Anchor Intermediate in SIA-b

3H]rnyo-inositol-labeled lipids to mild alkaline hydrolysis and phospholipase treatments
Aliquots of 400-1000 cpm of [3H]myo-inositol-labeled lipids were subjected to a single hydrolytic or enzymatic treatment or were subjected first to one, then another procedure as indicated at the top of each column. Numbers indicate the percentage of water-soluble counts as determined by partitioning of products in TX-114. Corresponding figures for control incubations without NaOH, without methanolic NH3, or without phospholipases are given in brackets. Numbers represent means of two to seven independent determinations done in part on different lipid preparations. I J -~ values are indicated for treated samples.  (lanes 1,3,4 (38) was removed by endoglycosidase H (not shown). Radioiodinated Thy-1, which still was expected to be in its native form at this stage, was subjected to PI-PLC treatments or to mild alkaline hydrolysis and the hydrophobicity of labeled products was evaluated by partitioning in TX-114. The ma-jority of Thy-l molecules of all cell types were found to get released from the detergent into the aqueous phase by PI-PLC (Table IV, experiment 2). After mild base treatment, only a fraction of Thy-1 was found in the aqueous phase, this fraction being significantly higher for thymocyte Thy-1 than for Thy-1 from lymphomas (Table IV, experiment 1). It seemed conceivable that the Thy-1 remained in the detergent phase after mild base treatments not because of the persistence of a base-resistant lipid on its anchor but rather because the NaOH treatment denatured the molecule and led to irreversible exposure of detergent-binding sequences. To distinguish these two possibilities we first treated Thy-1 with PI-PLC, then with base (Table IV, experiment 3). If base treatment irreversibly denatured Thy-1, it was expected that part of the PI-PLC treated Thy-1 would be retained in the detergent phase after a subsequent treatment with NaOH. However, NaOH treatment of PI-PLC treated Thy-1 did not change its hydrophilicity. Another concern was that the release of counts into the aqueous phase resulted from protein degradation rather than removal of acyl chains. SDS-PAGE analysis of the products of Thy-1 after various periods of NaOH treatment indeed showed that during prolonged exposure to NaOH the protein part of Thy-1 gets degraded and that even material running at 24-30 kDa was slowly losing its detergent-binding properties (Fig. 8). Nevertheless, 60-min incubations at 37 "C were sufficient to release 95% of counts from ph~sphatidyl-[~H]myo-inositol without destroying Thy-1 to a significant extent. Thus, it appears that this approach is a reliable indicator of the base sensitivity of the lipid component of the GPI anchor of Thy-1, and we conclude that part of Thy-1 from thymocytes as well as from lymphoma cells contains base-resistant anchor lipids.

DISCUSSION
A scheme of the reactions involved in the synthesis of the GPI glycophospholipid implying the stepwise addition of monosaccharides to PI has been established on the bases of biosynthetic intermediates observed in T. brucei (19)(20)(21). It is likely that the same reactions are also operating in mammalian cells although it seems that, due to the comparatively low abundance of GPI anchors in these cells, these reactions are more difficult to demonstrate. In spite of this, one might expect that some biosynthetic intermediates accumulate in mammalian mutants lacking enzymes involved in GPI biosynthesis. Indeed, our study shows that two out of five T-cell lymphoma mutants which are deficient in the biosynthesis of GPI anchors accumulate inositol-containing intermediates. The partial structure of lipid X, the most abundant lipid accumulating in the SIA-b line, is consistent with the idea of its being an intermediate in GPI biosynthesis containing 2 Aliquots of lZ5I-Thy-1 corresponding to 3000 cpm were either treated with NaOH or with PI-PLC. In experiment 3, some PI-PLC treated samples were subsequently treated with NaOH. Values indicate the fraction (in percentage) of radioactivity partitioning into the aqueous phase after each treatment. Brackets contain the same values for control incubations in the absence of NaOH or PI-PLC. All values have been corrected as described under "Materials and Methods." mannoses only, and it therefore is reasonable to assume that the accumulation of lipid X is related to the defect in GPI anchor biosynthesis of this line.
The simplest interpretation would be that lipid X accumu-  (Fig. 3) might point towards this conclusion. Unfortunately, the low abundance of these more polar lipids made it impossible to carry out structural analysis on them, and we therefore cannot decide whether they truely represent later intermediates in the GPI biosynthetic pathway or if they are degradation products of lipid X, e.g. lysoforms lacking the acyl on the inositol or the glycerol. (c) Lipid X also might be the product of an aberrant reaction, e.g. caused by the mislocalization or changed specificity of some biosynthetic enzyme. For example, the enzyme transferring ethanolamine phosphate onto the al,4-linked mannose (the intracellular localization of which is at present unknown) could be mistargeted to the endoplasmic reticulum instead of the Golgi and abort GPI-biosynthesis by prematurely adding phosphoethanolamine. It would appear that such a hypothesis is less likely in view of the fact that the absence of Thy-1 on the cell surface of SIA-b behaves as a recessive trait upon somatic cell hybridization inasmuch as Thy-1 is expressed at the surface of hybrids between SIA-b and wild type cells or mutants belonging to different complementation classes (42).
(Previous biochemical analysis of SIA-b has shown that, in contrast to mutants of other complementation classes, the intracellular form of Thy-1 is smaller, has lost its hydrophobic C-terminal peptide and that substantial amounts of Thy clear that the mere knowledge of the structure of lipid X does not allow to discriminate between several possible molecular defects. The suggested presence of an acyl chain on the inositol of lipid X is noteworthy. Acyl chains on inositol have first been described on the GPI anchor of human red blood cell acetylcholinesterase (8,9) but they are equally present on part of the biosynthetic GPI anchor intermediates generated by T.
brucei in vivo or in vitro (11,12,14,19). In fact, biosynthetic intermediates of GPI anchors of T. brucei exist with or without this acyl chain although no acyls are found on proteinbound GPI anchors in this organism. Thus, it is unclear at the moment whether acylated or nonacylated anchors are transferred onto proteins and whether the stepwise addition of monosaccharides takes place on acylated or nonacylated intermediates or on both. PI-PLC was reported to be hindered by the presence of an acyl on the inositol in all cases except for the GPI anchor of folate binding protein (43). At present it is not known to which hydroxyl of myo-inositol the acyl residue is attached, but it has been proposed that the attachment site in folate binding protein is different from the one of other GPIs (43). It seems conceivable that acyls would migrate on the myo-inositol ring much in the same way the acetyl groups migrate from C7 to C9 of sialic acids. Our data show that lipid X is heterogeneous with respect to PI-PLC sensitivity, a finding which could well be due to a heterogeneity with respect to the attachment site of the acyl on the inositol ring. Although lipid X clearly behaves as if it had some acyl chain on the inositol, the bulk of PI-PLC treated Thy-1 glycoprotein from SIA partitions into the aqueous phase of the TX-114 phase separation system, suggesting that most of the mature anchors do not contain an acyl on the inositol. Thus, if lipid X truly represents an intermediate in the GPI anchor biosynthesis, the acyl must get lost before the protein reaches the mature state. Alternatively, the inositol of lipid X might be acylated as a consequence of an abnormal side reaction which takes place only when the normal biosynthetic pathway is blocked.
Our investigation strongly suggests that part of the Thy-1 anchors of SIA-b contain a base-resistant lipid moiety. Lipid X also contains both base-sensitive and base-resistant anchors in similar proportions as found on the Thy-1 of the corresponding parental cells. This further supports the idea that lipid X is an intermediate of the anchor biosynthesis and suggests that GPI anchors are synthesized on both base-labile and base-resistant phosphoinositides in parallel. Recent data on the EL4-f mutant have shown that this line is unable to synthesize the base-resistant phosphoinositides (most likely 1-alkyl-2-acylphosphoglycerides) which are made by the parental lymphoma cell line (40). If this defect is responsible for the deficiency in GPI anchor biosynthesis, one might expect that the parental EL4 cell line only uses base-resistant phosphoinositides for GPI anchor biosynthesis. Our data suggest that this is not the case. However, it cannot be excluded that the EL4-f mutant indeed was derived from a variant cell which only uses base-resistant lipids for GPI anchor biosynthesis. In fact, our data do not exclude the possibility that base-sensitive and base-resistant GPI anchors are made by different subpopulations of lymphoma cells or thymocytes.
If we consider lipid X to be a normal intermediate, its structure suggests that the phosphoethanolamine side chain is added early during GPI biosynthesis, whereas the GalNAc/31,4 group is added later on.
We hope that the further studies of mammalian mutant cells will shed some light on the pathways and regulation of GPI anchor biosynthesis as well as on the functional aspects of the various types of these anchors.

Note Added in
Proof-Labeling of SIA and SIA-b cells with [3H]ethanolamine followed by analysis of extracted lipids by TLC and fluorography shows that SIA-b contains a heavily labeled lipid comigrating with lipid X, whereas no such lipid is detectable in SIA. This confirms the idea that the charged group present on the PLDgenerated headgroup of lipid X consists of ethanolamine-phosphate. While this paper was under review, another report addressing the defect in SIA-b has appeared (44).