Compositional analysis of Glucosaminyl(acyl)phosphatidylinositol Accumulated in HeLa S3 Cells

Eur. J. Biochem. 233, 384-394 (1995) 0 FEBS 1995

Compositional analysis of glucosaminyl(acy1)phosphatidylinositol accumulated in HeLa S3 cells Daniel SEVLEVER, Dawn R. HUMPHREY and Terrone L. ROSENBERRY

Department of Pharmacology, School of Medicine, Case Western Reserve University, Cleveland OH, USA

(Received 26 May 1995) - EJB 95 0837/5

GlcN(acyl)Ptdlns, a derivative of phosphatidylinositol (PtdIns) in which glucosamine and a fatty acid are linked to inositol hydroxyl groups, has been proposed to be an intermediate in the mammalian biosynthetic pathway for glycosylphosphatidylinositol (glycosyl-Ptdlns) anchors of membrane proteins. In this report, GlcN(acy1)Ptdlns metabolically labeled with [ 'Hlinositol is shown to accumulate in a HeLa S3 cell subline. The amount of GlcN(acy1)PtdIns in these HeLa S3 cells is about 10' molecules/cell, a level comparable to those of the most abundant glycosyl-Ptdlns-containing molecules reported to date. GlcN(acy1)PtdIns was purified by a two-step procedure involving octyl-Sepharose and thin-layer chromatography. Octyl-Sepharose separated phospholipids according to their number of hydrocarbon chains: one in 2-lysoPtdlns, two in PtdIns, and three in GlcN(acy1)Ptdlns. Purification also was aided by prior treatment of lipid extracts with bee venom phospholipase A2, an enzyme that did not cleave GlcN(acy1)PtdTns. The GlcN-inositol head group in purified GlcN(acy1)Ptdlns was confirmed by a number of procedures, including cation-exchange chromatography and mass spectrometry ; after radiomethylation, an equal molar ratio of GlcN(Me)Jinositol was measured. Fatty acid analysis indicated an overall stoichiometry of 2.3 mol fatty acidimol inositol with palmitic (16:0), stearic (18:O) and oleic ( 1 8 : l ) acids being predominant. Analysis of GlcN(acy1)inositol produced by HF fragmentation showed that palmitate was the acyl group attached to inositol and indicated that stearic and oleic acids were in the glycerolipid. Base methanolysis revealed that about 15 % of the purified GlcN(acy1)Ptdlns contained alkylglycerol. A substantial conversion of GlcN(acy1)PtdIns to a slightly more polar lipid occurred after overnight incubation in even mildly alkaline buffers. Although the current data do not allow proposal of a structure for this lipid, its formation from GlcN(acy1)Ptdlns may be important because the conversion appeared to occur in vivo.

Kejwords: glycoaylphosphatidylinositol ; fatty acids; alkylglycerols; octyl-Sepharose chromatography ; HeLa cells.

Glycosylphosphatidylinositol (glycosyl-PtdIns) provides a select group of cell-surface proteins with an alternative way of anchoring (reviewed by McConville and Ferguson, 1993 ; Eng- lund, 1993). The glycosyl-Ptdlns anchor is linked to the C-terminus of these proteins and contains a core glycan sequence, con- sisting of Etn-P -6Manul-2Mancxl-6Manrxl-4GlcN attached to an inositol phospholipid, that is conserved from protozoan to mammalian cells. All mammalian glycosyi-Ptdtns anchors characterized to date have additional substituents branching from the conserved core glycan. A common substituent is another Etn-P group on the first Man, adjacent to GlcN (Homans et al., 1988; Roberts et al., 1988a; Taguchi et al., 1994), and others occasion- ally branching from this Man residue are N-acetylhexosamine (Homans et al., 1988; Taguchi et al., 1994), and sialic acid (Stahl et al., 1992). Substituents on the other two Man residues also

Corrr.q,nndrrzc.e to D. Sevlever, Department of Pharmacology, Case

Fmt +1 216 368 3395. Ahhrc~vintions. Glycosyi-Ptdlns, glycosylphosphatidylinosito~; gly-

cosyl-Ptdlns-PLD, glycosyl-PtdIns-specific phospholipase D ; Ptdlns, phosphatidylinositol; PtdIns-PLC, Ptdlns-specific phospholipase C ; GlcN(acyl)Ptdlns, GlcNPtdIns with a fatty acyl group on inositol; Dol- P-Man, dolicholphoaphoryl-mannose; DMEM, Dulbecco's modified Eagle's medium; 2-IysoPtdIns. 1 -acyl-2-lyso-PtdIns.

Etizj/77c,r. Phuspholipase D (EC 3.1.4.4); phosphatidylinositol-~pe- cific phospholipase C (EC 3.1.4.10); phospholipase A2 (EC 3.1.1.4).

Western Reserve University, Cleveland, OH 44106-4965, USA

have been reported, including a third Etn-P attached to the second Man residue (Deeg et al., 1992a; Taguchi et al., 1994), and a fourth Man residue linked to the 2-position of the third Man residue (Homans et al., 1988; Taguchi et al., 1994).

In the past few years, free glycosyl-PtdIns species with structures resembling the glycosyl-PtdIns anchors of proteins have been reported. Although first identified in trypanosomes (Mas- terson et al., 1989; Menon et al., 1990), more polar mannosylated glycosyl-PtdIns species with the same core glycan found in glycosyl-PtdIns anchors (Puoti and Conzelmann, 1992, 1993 ; Hirose et al., 1992; Kamitani et al., 1992; Ueda et al., 1993) and less polar glycosyl-Ptdlns species devoid of mannose but with GlcNAc or GlcN linked to phosphatidylinositol (PtdIns) (Stevens and Raetz, 1991 ; Hirose et al., 1991 ; Sugiyama et al., 1991) were found in mammalian cells. Some of these glycosyl- PtdIns species accumulate in mutants that do not express glycosyl-Ptdlns-anchored proteins (Sugiyama et al., 1991 ; Hirose et al., 1992; Puoti et al., 1993; Mohney et al., 1994) and in cells in which glycosyl-PtdIns anchoring was inhibited by mannosamine (Sevlever and Rosenberry, 1993). These observations suggest that free glycosyl-PtdIns species are intermediates on the glycosyl-Ptdlns anchor biosynthetic pathway. In a recent proposal (Hirose et al., 1992), the initial steps in this pathway in mammalian cells involve the attachment of GlcNAc to PtdIns, deacetylation of GlcNAcPtdIns, and fatty acid acylation of GlcNPtdIns

Sevlever et al. (Euc J . Biochenz. 233) 385

on an inositol hydroxyl group to give GlcN(acy1)PtdIns. After the first Man residue is added, the pathway continues primarily through the attachment of Etn-P to this Man and sequential addition of two Man residues. This Man,GlcN(acyl)PtdIns (designated H6) is then modified by additions of Etn-P to the third Man residue (to give H7) and to the second Man residue to produce H8, a Man,GlcN(acyl)PtdIns with three Etn-P substituents.

One elusive intermediate on the proposed mammalian glycosyl-PtdIns biosynthetic pathway, GlcN(acyl)PtdIns, was recently identified in yeast (Costello and Orlean, 1992) and mammalian mutant cells (Urakaze et al., 1992) defective in the synthesis of dolicholphosphory-mannose (Dol-P-Man). This glycosyl-PtdIns had not been detected either by metabolic labeling of wild-type cells (Urakaze et al., 1992; Puoti and Conzelmann, 1992; Moh- ney et al., 1994) or by in v i m labeling of membrane fractions (Hirose et al., 1991 ; Stevens and Raetz, 1991), presumably because of its low abundance. However, it was recently found that addition of GTP to cell lysates stimulated the synthesis of GlcNPtdIns and that the synthesis and detection of GlcN- (acy1)PtdIns also was enhanced if the membranes were supplemented with acyl-CoA or CoA (Stevens, 1993; Stevens and Zhang, 1994). This glycosyl-PtdIns is of interest because it marks a point of apparent divergence in the glycosyl-Ptdh biosynthetic pathways in trypanosomes and mammalian cells. In contrast to mammalian cells which, as noted above, appear to require fatty acid acylation on the inositol before mannosylation can occur, bloodstream form trypanosomes are able to synthe- size mannosylated glycosyl-PtdIns species without acylation on the inositol (Guther et al., 1994), suggesting that inositol acylation may play different roles in mammalian and protozoan cells.

The simple notion that all free glycosyl-PtdIns species in cells are glycosyl-PtdIns anchor precursors is attractive, but it is possible that the precursors are only a small subset of a much larger population of free glycosyl-Ptdlns species with slightly divergent structures and other cellular roles. H7 and H8, the best candidates for the final precursors of glycosyl-PtdIns anchors, do differ structurally from most glycosyl-Ptdlns anchors on mammalian proteins because they are acylated on inositol and they contain a high percentage of diacylglycerol rather than alkylacylglycerol (see Rosenberry et al., 1994). Therefore, it is important to learn more about the detailed structures of free glycosyl-PtdTns species and the cellular roles they play. In this paper we report that a HeLa cell subline which expresses H6, H7 and HX as well as glycosyl-Ptdlns-anchored proteins (Sevlever et al., 1994) nevertheless accumulates high levels of GlcN(a- cy1)PtdIns. We describe the isolation of this GlcN(acy1)Ptdlns and its lipid compositional analysis. The acyl group on the inosi- to1 is palmitoyl, as we have previously reported for the glycosyl- PtdIns anchor of mammalian acetylcholinesterases (Roberts et al., 1988a,b) but the glycerolipid differs in containing primarily diacylglycerol with stearoyl and oleoyl acyl groups.

EXPERIMENTAL PROCEDURES

Materials. Dulbecco's modified Eagle's medium (DMEM), and inositol-free DMEM were obtained from Gibco. Heat-inacti- vated newborn calf serum, phospholipase A2 (PLA,) from bee venom and Crotalus nirox venom, jack bean a-mannosidase, alkaline phosphatase from bovine intestinal mucosa, and octyl- Sepharose were from Sigma. my0-[2-~HlInositol (20 Ci/mmol) was from American Radiolabeled Chemicals; [ 2-'H]mannose (20-30 Ci/mmol) and [ 'TIHCHO (50 Ci/mol) were purchased from ICN. Glycosyl-PtdIns-PLD was a generous gift from Dr K . 3 . Huang (Hoffman-LaRoche). PtdIns-specific phospholipase

C (Ptdlns-PLC) was purified from the medium of a Bacillus subtilis strain that overexpresses PtdIns-PLC of Bacillus thurin- gierzsis (Henner et al., 1988; Deeg et al., 1992a).

Cells and culture conditions. HeLa cells were maintained in DMEM supplemented with 10% newborn calf serum, 2 mM glutamine, 50 units/ml of penicillin and 50 yg/ml of streptomycin in an atmosphere of 5 % CO, at 37°C. These cells were obtained as the HeLa S3 line (a cloned HeLa cell line provided to the American Type Culture Collection after about 400 pas- sages) and were shown previously to contain mannosylated glycosyl-PtdIns species (Hirose et al., 1992; Ueda et al., 1993). During passaging leading up to the current studies, these cells were collected by scraping from culture flasks rather than by trypsinization. The passaging and culture procedures reproduci- bly generated cells with a more rounded morphology and a much higher level (3 -4-fold) of GlcN(acy1)PtdIns than the parental S3 cells. Karyotype analysis showed the same modal chromo- some count (65-67) and the same copy numbers of HeLa mark- ers (one of MI , one of M2, two of M3, one of M4) reported for HeLa S3.

Labeling and extraction of glycolipids. Cells (10" in 15 ml) were plated in 175-cm' flasks and incubated for 3 days with 50 pCi myo-['Hlinositol in inositol-free DMEM supplemented with 10% dialyzed newborn calf serum, 100 units penicillin, and 100 pg/ml streptomycin. Labeled cells were detached with a cell scraper, washed with phosphate-buffered saline and extracted twice with 4 ml chlorofordmethanoVwater (10: 10:3). The dried extract was partitioned between water-saturated n-butanol and n-butanol-saturated water, and labeled lipids in the dried rz-butanol phase were analyzed by TLC.

TLC. Dried samples were resuspended in water-saturated n-butanol, applied to silica gel 60 (Merck) and developed with chloroform/methanol/water (10 : 10:3) as previously described (Sevlever and Rosenberry, 1993) unless otherwise noted. The distribution of radioactivity on the plates was determined either with a Bioscan System 200 imaging scanner or by fluorography after spraying the plates with EN3HANCE (Dupont).

Enzymic cleavage of ['H]inositol-labeled lipids. Labeled lipids in the dried n-butanol phase were resuspended in the indicated buffer with or without enzyme and incubated overnight at 37°C. For treatment with glycosyl-PtdIns-specific phospholipase D (glycosyl-PtdIns-PLD ; Sevlever and Rosenberry, 1993), incubations were in 100 p1 40 mM Tris/HCI pH 7.0, 0.2% Noni- det P-40, and purified glycosyl-PtdIns-PLD (20 units/ml). PtdIns-PLC digestion was in 100 pl 20 mM sodium phosphate pH 6.9, 0.1 % Triton X-100, and I 0 pg/ml Ptdlns-PLC. Incuba- tion with alkaline phosphatase was i n 150 p1 0.1 M sodium car- bonate pH 9.0, 0.3% Triton X-100 and 100 units enzyme. Treat- ment with PLA, (bee venom) was in 50p1 5 mM Tris/HCl pH 8.0, 1 mM NaCI, 0.6 mM CaCI2, 0.2% Triton X-100 and 1 mg/ml (600- 1800 units/mg) of enzyme. Enzyme-treated samples and buffer controls were partitioned into n-butanol and analyzed by TLC.

Purification of [3H]GlcN(acyl)PtdIns. Radiolabeled lipids in the dried n-butanol phase were purified on an analytical scale (without combination with unlabeled lipids) or on a preparative scale (combined with an unlabeled lipid extract from 1 O8 cells). In some cases, samples were pretreated with bee venom PLA, by resuspending in 30 p1 n-butanol and incubating with 70 p1 5 mM Tris/HCl pH 8.0, 1 mM NaCI, 0.6 mM CaCL, and 1 mg/ ml of PLA, overnight at 37°C. In preparative-scale samples, 15-20% of the initial PtdIns was still intact after this incubation, and the digestion mixture was partitioned into n-butanol, dried, and resuspended for a second cycle of PLA, digestion. Digested samples were partitioned into n-butanol and dried. Samples to be purified were resuspended in 250 pl 100 mM am-

386 Sevlever et al. (ELK J . Biocl7enz. 233)

monium acetate pH 5.5 in 50% n-propanol, diluted to 600 p1 with 100 mM ammonium acetate, and loaded on a 10-ml octyl- Sepharose column connected to a Dionex HPLC system operat- ing at a maximum pressure of 1.38 MPa (200 psi). The elution program was basically as described by Ferguson (1992) with the following modifications: isocratic solvent A (100 mM ammonium acetate in 5 c/o n-propanol) at a tlow rate of 0.4 ml/min for 20 min; 20-26 min from 100% A to 6 0 % A and 4 0 % B (60% n-propanol); 26-273 rnin from 60% A and 40% B to 15% A and 85% B at a flow rate of 0.3 mllmin; and 273-293 min from 15% A and 85% B to 100% B at a flow rate of 0.3 ml/min. Radioactivity in 0.5-ml fractions was determined by scintillation counting. Peak fractions were combined and partitioned into n-butanol before drying for further analysis.

Chemical and enzymic treatment of lipids purified on octyl-Sepharose. Dried purified samples were deaminated by re- suspension in 200 p1 50 mM sodium acetate pH 3.5, 0.25 M NaNOZ, and 0.4% Tx-100. After incubation for 3 h at 37"C, another 100 pl 0.25 M NaNOZ was added for overnight incubation, and the samples were partitioned into n-butanol. In deamination controls, NaNO, was replaced with NaCl. For N-acetylation (Stevens and Raetz, 1991), a purified sample was resuspended in 1OO.pl methanol, 100 p1 of a saturated solution of NaHCO, was added, and the reaction was initiated with 10 pl acetic anhydride. After a 10-min incubation on ice, a second 10 p1 acetic anhydride was added and the reaction was allowed to proceed for SO min at room temperature. Reductive radiomethylation was performed with 10 mM [''CC]HCHO and 50 mM NaCNBH, in 50 p1 20 rnM sodium phosphate pH 7.5, 0.1 % Triton X-100 for 2 h at 37°C. After extensive dialysis against water, the product was partitioned into n-butanol and dried.

Lipids in dried purified samples were dephosphorylated by incubation in 50 pl 4 9 % HF at 4°C for 60 h. The sample was neutralized and partitioned into n-butanol as described previously (Sevlever and Rusenberry, 1993). For digestion with Crutalus PLA,, purified samples were resuspended in SO p1 5 mM Tris/HCI pH 8.0, 1 mM NaCI, 0.6 mM CaCI?, 0.1 c/o Tri- ton X-100 and 50 units enzyme (300-700 units/mg). After overnight incubation at 37"C, the products were partitioned into n- butanol. For base methanolysis, dried purified samples were resuspended in SO p1 0.2 M KOH in methanol for 30 min at room temperature. The samples were neutralized with 2 pl glacial acetic acid, dried and partitioned into n-butanol before further analysis.

Analytical procedures. Inositol in samples of silica beads scraped from TLC plates or of HPLC fractions was quantitated by GC/MS as described previously (Deeg et al., 1992a). Mea- surements of radiolabeled amines in acid hydrolyzates by cation- exchange chromatography were based on previous procedures (Haas and Rosenberry, 1985; Deeg et al., 1992a). Briefly, a [''CJmethylated or [lH]inositolLlabeled band was scraped from a TLC plate and the silica beads were hydrolyzed in 6 M HCI at 300°C for 10 h. The products were chromatographed on a Beckman 1 19CL amino acid analyzer in pH-2.2 buffer and identified by characteristic elution times (Haas et al., 1986; Deeg et al., 1992a). Analyses of fatty acids followed procedures described by Roberts et al. (1988b). A sample on scraped silica beads was combined with a 17:O fatty acid internal standard (5 nmol) and hydrolyzed in 100 pl 1 M anhydrous methanolic HCI for 16-20 h at 65°C. Chloroform (200 pl) and water (75 pl) were added, the partitioned aqueous phase was washed with chloroform ( I 00 PI) and the combined organic phases were dried. Samples for fatty acid methyl ester determination were resuspended in 20 pl isooctane. Aliquots were analyzed on a Hewlett-Packard model 5890 gas chromatograph containing a

SP 2380 column (15 m, 0.25 mm internal diameter, Supelco) and a flame ionization detector with response factors determined previously (Roberts et al., 1988b).

Electrospray ionization mass spectrometry (ESI-MS). The sample on silica scraped from a TLC plate was N-acetylated with acetic anhydride, dried, and methylated with Me1 in NaOH/ dimethylsulfoxide (Ciucanu and Kerek, 1984). Spectra were obtained as described previously (Deeg et al., 1992a).

RESULTS

Identification of [3H]inositol-labeled lipids from HeLa cells. By far the most abundant radiolabeled lipid in cells incubated with ['H]inositol is PtdIns, as shown in the TLC profile in Fig. 1A. Moreover, when HeLa cells derived from the S3 subline were labeled with ['H]inositol, a species more apolar than PtdTns was apparent (peak a in Fig. 1A) that was not evident after ['Hlinositol labeling of K562 cells (Mohney et al., 1994), lymphoma cells, or T cell hybridomas (Puoti and Conzelmann 1992; Urakaze et al., 1992). The PtdIns peak in Fig. 1 A accounted for about 90% and lipid a, for about 5 % of the total radiolabel in the lipid extract, and two other minor peaks also were observed. In initial steps to identify lipid a, enzymic cleavage studies were conducted. When the lipid extract was incubated with bee venom PLA2, an enzyme that releases the sn-2 fatty acid from PtdIns to give a labeled 2-lysoPtdIns, the major peak on TLC shifted to a more slowly migrating position in Fig. 1 B. This position corresponded to one of the minor peaks in Fig. 1A. The reduced recovery of radioactivity in 2-IysoPt- dIns was probably due to contamination of the PLA, preparation with a PLA, or 2-lysoPtdlns phospholipase activity, since TLC analysis of the aqueous phase after partitioning of PLA,-treated samples revealed a very polar labeled species migrating close to the origin (data not shown). The minor lipid that migrated more slowly than 2-1ysoPtdIns was also susceptible to PLA,, but its lyso form was not recovered in the n-butanol phase following partitioning. In contrast, lipid a was not cleaved by this PLA, (Fig. 1 B).

Treatment of the lipid extract with PtdIns-PLC, which cleaves the phosphodiester bond between phosphate and glycerol in PtdIns, resulted in the loss of peaks correponding to PtdIns and 2-IysoPtdTns but did not affect lipid a or the slowest migrating species (Fig. 1 C). This species was assigned as PtdInsP based on its relative TLC mobility and its removal by alkaline phosphatase (Fig. 1 D), an enzyme that cleaves phos- phomonoesters. Lipid a was the only labeled species cleaved by glycosyl-PtdIns-PLD, an enzyme specific for the phosphodiester bond between phosphate and inositol in glycosyl-PtdIns species (Fig. I E ) , and a new, relatively polar peak a' appeared to be a product of the cleavage that partitioned into n-butanol. Two conclusions about lipid a can be drawn from the data in Fig. 1 : (a) its minimum structure must be GlcNPtdlns, because glycosyl-PtdTns-PLD does not cleave PtdIns or GlcNAcPtdIns (Doer- in& et al., 1989); (b) it appears to have a fatty acyl group on inositol. This would account for its resistance to PtdIns-PLC hydrolysis in Fig. 1 C (Roberts et al., 1988b) and for the hydro- phobicity of the inositol-containing product a' produced by glycosyl-PtdIns-PLD as shown in Fig. 1 E. Further evidence that this hydrophobic substituent is an acyl group was obtained by demonstrating its susceptibility to the mild base hydrolysis treatment described i n the legend to Fig. 2A. Before treatment a' radioactivity partitioned to > 90% into n-butanol, but after treatment 9.5% of the radioactivity was recovered in the aqueous phase. The TLC mobility of lipid a was unchanged after treatment of the labeled lipid extract with jack bean a-mannosidase,

Sevlever et al. (EUK .I. Biochem. 2331 387

- c) I 0

X

.c

s 0

0 0 W

W 73

z > 0 m 0 -0 m

L

w

+-

- +- .- .- e

.-

a

Ptdlns

I Untreated 401 10 A

3 L , 1 ;lh, , , 0

2-lyso Ptdlns I

I : :

I i i PLAP - , 10

3 L A 1 , .. . . . . . . . . .

I

. . . .. . . . . . . :: . .... : !

... i __.._. i ; _......_... .: :.: ,...: '.. '.. ._..__.____..._.__.._._______ 0

Ptdlns-PLC 10 - 40J - 1

j ,j':,,j !.JL 1 , I I n .......... (,............. . .... . .- ...........................

- H I 1 I I I

40

10

3

0 I I 1

I GPI-PLD

10 B 3 L J K l . . , ,

. . ..........( ,___.. ..___. ... ... .-.__: : ........................... 0

0 a 12 16

Migration (crn)

Fig. 1. Enzymic analyses of [ZH]inositol-labeled lipids. The HeLa S3 subline was incubated for three days in the presence of50 pCi [ 'Hlinosi- tol; lipids were extracted, digested with enzymes. partitioned into n- butanol, and analyzed by TLC as outlined in Experimental Procedures. About 50% of the total label remained associated with the washed cells and 40% with the lipid extract. Sample3 of the lipid extract (500000 cpm) were incubated overnight at 37°C with the enzymes or corresponding buffers as controls as follows: (A) untreated; (B) PLA, from bee venom; (C) PtdIns-PLC; (D) alkaline phosphatase (AP): (E) glycosyl-Ptdlns-PLD (GPI-PLD). The distribution of the radioactivity was measured by radioscanning of the plate. Solid lines correspond to enzyme-treated samples and dashed lines to the buffer controls. The major peak in A comigrated with a commercial ['HIPtdlns standard. and other peaks were identified as outlined in the text.

which cleaves terminal mannose groups (data not shown). This suggested that there were no mannose groups attached to lipid a and, together with the evidence discussed above, led us to propose that lipid a is GlcN(acy1)Ptdlns.

Lipid a contains the sequence GlcN-Ins. To obtain additional evidence for a GlcN-inositol linkage in lipid a, the radiolabeled TLC band corresponding to lipid a' in Fig. 1 E was hydrolyzed i n base and the lipid a band i n Fig. 1 A was hydrolyzed in acid, and the labeled hydrolysis products were analyzed by cation- exchange chromatography as shown in Fig. 2. The GlcN-Ins

I A 6oo 4 GlcN-Ins

InsIlnsP

GlcN-Ins :::/A , A , , , 0 I=?===

0 20 40 60 120

Time (min)

Fig. 2. Cation-exchange chromatography of fragments of lipid a. (A) Silica gel bands corresponding to the position of lipid a' in Fig. 1 E were scraped from the plate. The sample was eluted from the beads with methanol, dried, resuspended in mild base (7 M NH,OH in methanol) for 4 h at 37"C, dried, and partitioned between n-butanol and water. The aqueous phase ( 1 600 cpm) was dried and taken for analysis. (B) Silica gel bands corresponding to the position of lipid a in Fig. 1 A were hy- drolyxd directly ' in 6 M HCI, and 4000 cpm of the hydrolyzate was analyzed. Acid hydrolysis and cation-exchange chromatography conditions are given in Experimental Procedures: radioactivity in 1 -ml fractions was measured by scintillation counting. The protonated ammonium form of the GlcN-Ins conjugate was assigned based on the close coinci- dence of its retention time with that of the protonated form of a GlcN(Me)?-Ins standard (2 min later than the standard; data not shown). The peak labeled lns/lnsP was not retained on the column and presumably reflects ['Hlinositol and ['Hlinositol phosphate generated during acid hydrolysis. Recovery of radiolabel from the chromatography column (corrected for quenching) was >90%.

conjugate has a characteristic retention time on a cation exchange column, and detection of this conjugate was recently proposed as another diagnostic tool for glycosyl-PtdIns species (Deeg et al., 1992a). When lipid a' obtained by glycosyl-PtdIns- PLD cleavage of lipid a was deacylated in base, the only fragment observed corresponded to GlcN-Ins (Fig. 2A). This observation tended to rule out attachment of additional mannose residues, since such residues result in earlier column retention times (Deeg et al., 1992a). The same GlcN-Ins fragment was obtained after HCI hydrolysis of lipid a (Fig. 2 B), but some free { 'Hlino- sitol and ['H]inositol phosphate also were produced because the GlcN linkage to inositol is partially susceptible to strong acid hydrolysis (Deeg et al., 1992a). These data further support the assignment of lipid a as GlcN(acy1)PtdIns.

Purification of lipid a by chromatography on octyl-Sepha- rose. Further structural characterization required purification of lipid a to remove labeled and unlabeled lipid contaminants. Oc- tyl-Sepharose chromatography was a useful adjunct to TLC in the large-scale purification of a family of glycoinositol phospholipids from Lei.vhmaniu major (McConville and Bacic, 1989). While this technique gave little resolution on the basis of glycolipid head group, we reasoned that it might well separate lipids on the basis of the number of acyl or alkyl chains that they contained. Since lipid a appeared to be acylated on inositol and thus to contain three hydrocarbon chains, a procedure that could separate it from bulk phospholipids with two chains was attractive. Figs 3 and 4 demonstrate that chromatography on octyl- Sepharose did indeed separate lipid a from bulk phospholipids. Fig. 3 A shows the profile from a ['H]inositol-labeled lipid extract from HeLa cells prior to dilution with unlabeled lipids. Radioactivity eluted in three peaks denoted 1-111 with near baseline resolution. TLC analysis indicated that the radiolabel in

388 Sevlever et al. (Eur: J. Bioclzenz. 2-73)

pool I corresponded to 2-lysoPtdlns, that i n pool 11 to Ptdlns, and that in pool 111 to lipid a (see Fig. 9). When a large excess of unlabeled lipid extract was mixed with the radiolabeled lipid extract for preparative scale chromatography as shown in Fig. 3B, some resolution was lost. Peak I now appeared as a shoulder on peak 11, and peak 111 appeared only partially resolved from peak I1 (Fig. 3 B). TLC analysis of pools taken from near the centers of these peaks shows that pool I now contained some PtdIns in addition to 2-IysoPtdIns (Fig. 3 C), but Ptdlns in pool 11 (Fig. 3 D) and lipid a in pool I11 (Fig. 3E) were radio- chemically quite well resolved. The bulk phospholipids in the extract were monitored by phosphate determination and were recovered near the Ptdlns peak in pool I 1 (Fig. 3 B).

To improve the resolution of lipid a in preparative octyl- Sepharose chromatography, we took advantage of the observation in Fig. 1 B that this lipid is resistant to bee venom PLA,. Two cycles of incubation of the lipid extract with bee venom PLA, removed virtually all of the radiolabeled Ptdlns (Fig. 4A). Labeled lipid products resulting from the PLA, treatment eluted at the position of IysoPtdIns in pool I, but the sample load was also decreased because, as noted above, about a half of the labeled products were removed in the aqueous phase of the 17-

butanol partition prior to column application. Phosphate determination indicated that bulk phospholipids were also cleaved by bee venom PLA, and recovered as lyso phospholipids in pool I. Based on inositol determination, the amount of putative GlcN- (acy1)PtdIns recovered in pool 111 was about 40 nmol/lOX cells. To assess whether lipid contamination of this GlcN- (acy1)Ptdtns in pool III was decreased by the PLA, treatment, total fatty acids were determined as fatty acid methyl esters. Pool 111 from Fig. 3 B had a stoichiometric ratio of about 10 mol fatty ac idshol inositol, while pool 111 after PLAL treatment in Fig. 4 A had a ratio of about 5 , within a factor of 2-3 of that expected for pure GlcN(acy1)PtdIns.

The TLC profiles of purified lipid a in pool I11 from octyl- Sepharose chromatography revealed a small amount of a more polar component (Figs 3E and 4D) that is denoted b in Fig. SA. Additional information about the source and structure of peak b is presented below.

Glucosamine is present in the putative GlcN(acy1)PtdIns. As a first step in characterizing the chemical composition of purified lipid a and confirming its structure as GlcN(acyl)Ptdlns, we focused on the component containing the amine group. After the base or acid treatment and cation-exchange chromatography in Fig. 2, this component was consistent with GlcN linked to inositol. To confirm the presence of the free amine group in intact lipid a, three amine modification reactions were conducted on purified lipid a. Treatment with nitrous acid increased the TLC mobility of the main peak, consistent with deamination of the free amine group on GlcN(acy1)Ptdlns and cleavage of the GlcN glycosidic bond to give Ptdlns with a acyl group on inositol (Fig. S B ; see Urakaze et al., 1992). N-acetylation of lipid a also yielded a faster-migrating product consistent with GlcNAc- (acy1)PtdIns (Fig. 5 C ; see Hirose et al., 1991). Peak b also was susceptible to both treatments and appeared to give parallel shifts in TLC mobility. These data indicate that both lipids a and b contain a hexosamine with a free amine group. This hexosamine was identified as GlcN by reductive radiomethylation, a technique used previously to detect and quantitate GlcN and Etn in glycosyl-PtdIns-anchored proteins (Haas et al., 1986; Medof et al., 1986; Fatemi et al., 1987; Lee et a]., 1992) and i n free glycosyl-Ptdlns species (Deeg et al., 1992b). The main peaks detected by cation-exchange chromatography of acid hydrolyzates from [ "Clmethylated lipid a had retention times corresponding to GlcN(Me)2 and Gl~N(Me)~1ns (Fig. 6). Reductive

4 0

2

1

0

0 50 100 150

Fraction number

II u

0 50 100 150

Fraction number

2-lyso GI c N ( a c y I) Ptdlns-P Ptdlns Ptdlns Ptdlns

I l l I

=- 40

0 o p - - . ~ - I D ................. 1 L ...... ~

c ._ > 0 m

U m [I

.- t.

.-

' " L A - 0 0 8 10 12 16

Migration (cml

Fig. 3. Octyl-Sepharose chromatography of ['Hlinositol-labeled lipids. The HeLa S3 subline was incubated with ['Hlinositol and a radiolabeled lipid extract was prepared as in Fig. 1. (A) An aliquot of the extract (150000 cpm in 600 pl) was subjected to analytical scale chromatography on octyl-Sepharose as outlined in Experimental Procedures. Fractions (600 pl) were taken to measure radioactivity by liquid scintillation counting; the recovery of radioactivity was 80%. Selected fractions were combined in three pools (1-111) as indicated by the solid horizontal lines for TLC analysis. (€3) A radiolabeled extract (2Xl0"cpm) was combined with a similar extract from 5X107 unlabeled cells for preparative scale chromatography, as outlined in Experimental Procedures, and 600-pl fractions were collected. A 5-pl aliquot was taken to measure radioactivity (-) by liquid scintillation counting and a 30+1 aliquot for phosphate (......) determination (Ames, 1966); data are shown/fraction. The recovery of the radioactivity was 87%. Aliquots (2000 cpm) of the indicated pools 1-111 were analyzed by TLC as shown in C-E: (C) pool 1, fractions 47-52; (D) pool 11, fractions 74-79; (E) pool Ill, fractions 102-108. An unfractionated sample was run in a parallel TLC lane, and the migration of each of the four radiolabeled lipids identified i n Fig. 1 is indicated. Radioactivity (%) indicates the percentage of detected total radioactivity in each 0.78-mm segment of the TLC lane. Lipid a is provisionally indicated as GlcN(acy1)PtdIns.

methylation not only identified the free amine as GlcN but also established that lipid a was free of other detectable amine contaminants. Furthermore, inositol determination on the HCI hydrolyzates established a molar ratio of close to 1 for GlcNhnosi- to1 in lipid a (Table 1 ) .

Sevlever ct al. (ELII:

I

1 2 0 1 ii jJ,/JJ1 1

0 0

0 50 100 150

Fraction number

J. Biochem. 233) 389

a

GlcN(acy1)Ptdlns 1 A

40 1

2-lyso GlcN(acy1) Ptdlns-P Ptdlns Ptdlns Ptdlns

1 I 1 I

2. 0 J __ ___. . ..... 1111 c .-

0 8 10 12 16

Migration (crnl

Fig. 4. Octyl-Sepharose chromatography of ["Hlinositoi-labeled lipids after digestion with bee venom PLA,. (A) Radiolabeled and unlabeled lipid extracts were combined as in Fig. 3 B and incubated with PLAz prior to preparative-scale chromatography as outlined in Experi- mental Procedures. The 'H (-) and phosphate/fraction (. . . . .) were measured and the indicated pools 1-111 were taken for TLC analysis as in Fig. 3B. The column onput contained 2x10" cpm, and the recovery of radioactivity in the column fractions was 85%. (B-D) TLC analysis of pools 1-111: (B) pool I, fractions 47-52; (C) pool 11, fractions 78- 82; (D) pool 111, fractions 102- 108. Migration of the indicated standards was measured as in Fig. 3. Radioactivity (%) is defined and lipids are designated as in Fig. 3.

Identification of GlcN-inositol in lipid a by mass spectrometry. Treatment of purified lipids a and b in pool I11 of Fig. 4A with glycosyl-PtdIns-PLD resulted in somewhat greater cleavage of lipid a than was observed with the crude lipid extract in Fig. 1 E. Close to 70% of lipid a was cleaved by glycosyl- PtdIns-PLD in Fig. 7B, and a new peak appeared at a TLC position identical to that of lipid a' in Fig. 1 E. In contrast, lipid b did not appear to be altered by glycosyl-PtdIns-PLD (Fig. 7 B). The labeled glycosyl-PtdIns-PLD product expected from GlcN(acy1)PtdIns is GlcN(acyl)inositol, and the same product should be generated by treatment of GlcN(acyl)Ptdlns with HF, a reagent that cleaves phosphodiester bonds. HF treatment did in fact release a major fragment from lipids a and b in pool 111 of Fig. 3 A that comigrated with the glycosyl-PtdIns-PLD product (Fig. 7 C). The putative GlcN(acy1)inositol bands generated by both reagents were broad and fluorography revealed a closely spaced doublet. However, the source of this heterogeneity was not clear. The doublet produced by HF did not appear to arise from fatty acid heterogeneity (see below). Treatment of the HF fragment with jack bean a-mannosidase had no effect on the doublet pattern (data not shown), suggesting that the heterogeneity also did not arise from partial mannosylation.

deamination

B r c - 2 0 1 c 0 N-acetylation 2 1 c " ._

(r 6ooL 0 0 8 12 16

Migration (crnl

Fig. 5. Lipid a contains a free amine group. Dried samples (5000 cpm) of pool III from Fig. 3A containing ['Hlinositol-labeled lipid a, provisionally identified as GlcN(acyl)PtdIns, were given the following treatments as indicated in Experimental Procedures: (A) buffer control for deamination with nitrous acid; (B) deainination with nitrous acid; (C) N-acetylation with acetic anhydride. Treated samples were partitioned into n-butanol. developed in TLC with chloroform/methanol/l M NH,OH (10: 10: 3) as a solvent, and analyLed as in Fig. I . The control for N-acetylation (not shown) gave the same TLC profile as untreated samples.

)1 c .- > 0 m 0 -0 Q IT

._ c

.-

0 40 80 120

Time Irnin) Fig. 6. Identification of GlcN in lipid a. A dried sample (2000 cpm) of pool 111 from Fig. 4A containing ~'H]inositolLlabeled lipid a was radiomethylated with ['YJHCHO and NaCNBH, and run on TLC; the major "C-labeled silica band was hydrolyzed in 6 M HCI as outlined in Experi- mental Procedures. An aliquot (3000 I4C cpm) from the dried hyroly- zates was mixed with ['H]GlcN(Me), and ['H]Etn(Me)? standards (30000 .'H cpm each) and analyzed as in Fig. 2 except that a dual 'W '*C channel program was employed in the scintillation counter. Only the values for "C are shown. The '"C recovery (corrected for quenching) was > 90%. The elution positions of [.'H]GlcN(Me), and [ 'H]Etn(Me), were fractions SO and 86, respectively.

Silica containing the band corresponding to lipid a', provisionally GlcN(acyl)inositol, was scraped from a TLC lane corresponding to that in Fig. 7C, N-acetylated, methylated, and analyzed by ESI-MS. Fatty acid esters are cleaved during the methylation procedure. A predominant MH' ion with m l ~ . of 532.2 was observed (Fig. 8), consistent with the sodium adduct of HexN(Me)Ac(OMe),-Ins(OMe),. This evidence together with the GlcN/inositol ratio in Table 1, demonstrates conclusively that lipid a is GlcN(acy1)PtdIns.

Lipid analysis of GlcN(acy1)PtdIns. Although bee venom PLAz failed to cleave GlcN(acy1)PtdIns either in the crude lipid extract (Figs 1B and 4) or after purification on octyl-Sepharose (data not shown), Crotalus PLA2 did cleave this glycosyl-PtdIns

390 Sevlever et al. (Eur: J. Hiochem. 233)

GlcN(acyl1Ptdlns I

GPI-PLD

' 4 1 c I

HF

0 I

1 4 1 E l , PLA2 control

8 12 16 0

0

Migration fcm)

Fig. 7. Cleavage of lipid a by glycosyl-Ptdlns-PLD, HF, and Crotalus PLA,. Dried samples (5000 cpm) of pool 111 from Fig. 3A containing ['Hlinositol-labeled lipid a, provisionally identified as GlcN(acyl)PtdIns, were treated with the following agents as indicated in Experimental Pro- cedures : (A) untreated control: (B) glycosyl-PtdIns-PLD (GPI-PLD) ; (C) HF; (D) Crofuliis PLA,; (E) buffer control for Crotalus PLA,. Treated samples were analyzed by TLC as in Fig. 1. The migrations of lipid b, GlcN(acy1)PtdIns and GlcN(acy1)Ins are indicated. Radioactivity (%) is defined as in Fig. 3.

loo 1 532.2

Ac

W I I I I ' I I I ' I

300 400 500 600

m/z Fig. 8. Electrospray ionization mass spectrum of N-acetylated and permethylated GlcN-Ins. GlcN(acy1)inositol was derivatized and subjected to mass spectrometry as outlined in Experimental Procedures. A molecular ion of rnlz 532.2 corresponding to the indicated structure is identified.

Table 1. Compositional analysis of GlcN(acy1)PtdIns and GlcN(acy1)- inositol. Determinations were conducted on three GlcN(acy1)PtdIns preparations (1 -3) corresponding to pool I11 in Fig. 4A that contained about 10 nmol based on inositol. One determination also was made on 7 nmol GlcN(acy1)inositol produced by HF treatment of pool 111 in Fig. 4A. One aliquot of sample 1 was radiomethylated, repurified on TLC, and analyzed for radiomethylated GlcN as in Fig. 6 . The quantity of GlcN was calculated from the previously calibrated specific radioactivity of stock [ 14C]HCH0 (69 cpdpmol) (Haas and Rosenberry, 1985; Lee et al., 1992). Samples were run on TLC, and the radioactive band corresponding to GlcN(acy1)PtdIns or GlcN(acy1)inositol were scraped, subjected to acid methanolysis, and analyzed for fatty acid methyl esters. Quantities are expressed relative to the amount of ioositol determined by GC-MS. Methods are outlined in Experimental Pro- cedures. ~

Component Amount in

GlcN(acy1)PtdIns GlcN- (acy1)-

1 2 3 inositol

mollmol inositol

GlcN 1.04 5 0.02 Fatty acids 1610 0.50 1.07 0.93 0.88 18:O 0.92 1.06 1.10 <0.03 18: 1 0.38 0.37 0.34 <0.03

both before (data not shown) and after purification (Fig. 7D). Cleavage was quantitative (>98% in Fig. 7D), indicating that GlcN(acy1)PtdIns was composed entirely of a glycerol-based inositol phospholipid and was devoid of any subpopulation based on an inositol-containing ceramide. Assignment of the products resulting from Crotdus PLA, treatment was compli- cated by a substantial redistribution between GlcN(acy1)PtdIns and lipid b in the buffer control (Fig. 7E). Nearly equal amounts of these two lipids together with smaller amounts of more polar labeled species were apparent after overnight incubation in this pH-8 buffer, indicating that lipid b is produced from GlcN- (acy1)PtdIns (see Discussion). We presume that the largest peak generated by Crotulus PLA, treatment in Fig. 7 D is the sn- l - acyl-glycerol derivative of GlcN(acy1)PtdIns because the difference in its migration from GlcN(acy1)PtdIns was the same as the difference in migration between PtdIns and 2-IysoPtdIns and because it retained susceptibility to glycosyl-PtdIns-PLD (data not shown). It was unclear whether the other peaks in Fig. 7D derived from the largest peak or from lipid b.

To determine the fatty acid composition of GlcN(acy1)- Ptdlns, a sample free of lipid contaminants was required. The purity of pool III from Fig. 4A was assessed by running a sample on TLC and exposing the plate to iodine. With a sample containing 10 nmol inositol, the only band detected by iodine was close to the position of lipid b. The band corresponding to radiolabeled GlcN(acy1)PtdIns was revealed by fluorography, scraped from the plate, and analyzed for fatty acids. A total of about 2.3 mol fatty acids/mol inositol was obtained (Table l), consistent with a predominant composition for GlcN(acy1)PtdIns that includes a diacylglycerol and a third fatty acyl group on inositol. About two-thirds of the fatty acids were identified as 18:O and 18:1, and the remaining third was 16:O. In order to establish which fatty acid was attached to inositol, a 7-nmol sample of GlcN(acy1)inositol was prepared by HF treatment of GlcN(acy1)PtdIns and isolated by TLC as in Fig. 7C. Acid rneth- anolysis of this sample yielded only 16:O with a stoichiometry relative to inositol of 0.88 (Table 1). Thus virtually all of the

391 Sevlever et al. (EUK J . Bioclzern. 233)

GlcN(acy1)F'tdlns -

F'tdlns -

2-IysoPtdlns -

origin -

Fig. Y. Base-resistant ['H]inositol-labeled glycerolipids. Dried samples (50000 cpm) of pools 1-111 from the octyl-Sepharose Chromatography in Fig. 3 A were incubated in methanol with (+) or without (-) 0.2 M KOH for 30 min at room temperature. The samples were neutralized with acetic acid, dried, partitioned into n-butanol, and run on TLC as outlined in Experimental Procedures; the plate was then submitted to fluorography. The base-resistant sn-1 -alkyl-glyceroiipid products expected from PtdIns and GlcN(acy1)PtdIns were not resolved in this TLC solvent system. Lipids are designated as in Fig. 3 .

16:0 fatty acid in GlcN(acy1)Ptdlns is accounted for in the acyl group attached to inositol, and the glycerolipid is composed primarily of 18: 0 and 18: 1 fatty acids. Acid methanolysis of the iodine-stained TLC band containing lipid b revealed a somewhat smaller total amount of fatty acids than in the GlcN(acy1)PtdIns band. However, the ratio of fatty acids to inositol in the lipid b band was much larger than 3, indicating that this band was heterogeneous.

To compare the alkylglycerol contents of GlcN(acylfPtd1ns and lipid b in pool I11 with those of PtdIns and 2-lysoPtdIns, samples from each of the three pools in Fig. 3 A were subjected to base methanolysis, partitioned into n-butanol, and analyzed by TLC as shown in Fig. 9. ['H]Inositol-labeled phospholipids containing diacylglycerol should be converted entirely to water- soluble labeled products by this treatment, whereas those containing alkylacylglycerol should be converted to labeled species containing alkylglycerol that remain in the n-butanol phase. No detectable radioactivity from the 2-IysoPtdIns pool in Fig. 3 A remained in the n-butanol phase after base methanolysis, but both the PtdIns and GlcN(acy1)Ptdlns pools contained labeled base-resistant lipids with TLC mobilities close to that of 2-Iy- soPtdIns (Fig. 9). Scintillation counting of the n-butanol phases after base treatment and scanning of the TLC plates indicated that about 1 % of the PtdIns pool and about 15 % of the GlcN- (acy1)PtdIns pool were resistant to base. Base methanolysis of the silica beads containing GlcN(acy1)PtdIns and lipid b revealed that both bands contained about the same percentage of ['Hlinositol-labeled base-resistant species, ruling out the possibility that lipid b was enriched in these species.

DISCUSSION

In this paper we report the accumulation of GlcN(acy1)- PtdIns, an intermediate in the glycosyl-Ptdlns biosynthetic path-

way, in a HeLa S3 cell subline. This intermediate is undetectable in most cultured cell lines, but it has been identified in the murine lymphoma mutant E (Urakaze et al., 1992) and the yeast mutant dpml (Costello and Orlean, 1992), both of which fail to express glycosyl-PtdIns-anchored proteins. These mutants are defective in Dol-P-Man synthase and make no Dol-P-Man, a cofactor necessary both for mannosylation steps immediately downstream from GlcN(acy1)PtdIns in the glycosyl-PtdIns biosynthetic pathway and for the synthesis of N-glycans. Leaky mutants of CHO cells (Singh and Tartakoff, 1991) and T cell hybridomas (Thomas et al., 1992) also have been reported that make very small amounts of Dol-P-Man. Expression of glycosyl-PtdIns-anchored proteins in these mutants is increased from low levels to near normal by culturing with tunicamycin, an agent that blocks N-glycan synthesis but does not inhibit the glycosyl-PtdIns pathway. Another CHO mutant (Lec35), which makes Dol-P-Man but cannot utilize i t efficiently for N-glycan or glycosyl-Ptdlns synthesis in vivo, has been suggested from indirect evidence to accumulate GlcN(acy1)PtdIns (Camp et al., 1993). The HeLa S3 subline used here differs from all these mutants in making essentially normal amounts of Dol-P-Man, mannosylated free glycosyl-PtdIns species, and glycosyl-PtdIns- anchored proteins whose levels are at most slightly affected by culturing with tunicamycin (Sevlever, Schiemann and Rosen- berry, unpublished results). Further studies are necessary to determine whether GlcN(acy1)PtdIns accumulates in these cells because of a partial defect in its subsequent mannosylation, because it rapidly escapes the cellular compartment in which mannosylation occurs, or because of its accelerated biosynthesis.

The amount of GlcN(acy1)PtdIns in these cells, based on inositol determination by GUMS, is about lo7 molecules/cell. This is an order of magnitude more than the amount of H8, the most abundant mannosylated glycosyl-PtdIns in these cells (Ueda et al., 1993; data not shown), and is similar to the amounts calculated for leishmania1 glycosyl-PtdIns species and the glycosyl- Ptdlns-anchored variant surface glycoprotein from Trypanosoma brucei (McConville and Ferguson, 1993), the most abundant glycosyl-PtdIns-containing molecules reported to date. The high levels of GlcN(acy1)PtdIns permitted its purification by octyl- Sepharose chromatography. This technique resolved ['HIInosi- tol-labeled phospholipids according to their number of hydrocarbon chains: one in 2-lysoPtdIns, two in PtdIns, and three in GlcN(acy1)PtdIns (Fig. 3). Purification was also aided by prior treatment of the lipid extract with bee venom PLA,. Trypanoso- ma1 glycosyl-PtdIns species were previously reported to be resistant to bee venom PLA, (Field et al., 1991) and a similar resistance was observed here for GlcN(acy1)PtdIns (Fig. 1 B). Prior treatment of the extract converted most of the labeled and unlabeled phospholipids to lyso phospholipids, further improving both the resolution of GlcN(acy1)PtdIns from lipid contaminants on octyl-Sepharose (Fig. 4) and the purity of GlcN- (acy1)PtdIns in pool 111. As a result of this PLA, treatment, the ratio of fatty acids to inositol in pool III decreased by about a factor of two, and phosphatidylethanolamine contaminants revealed by radiomethylation were removed. The fatty acidhnosi- to1 molar ratio of about 5 in pool I11 after the treatment indicated that a final TLC step was still required for complete purification. However, octyl-Sepharose was important in achieving GlcN- (acy1)PtdIns purification because the fatty acidhnositol molar ratio was about 4 in the radiolabeled GlcN(acy1)PtdIns band from TLC of preparative-scale lipid extracts prior to chromatography on octyl-Sepharose. Since most free glycosyl-Ptdlns species identified in mammalian cells to date appear to be acylated on inositol and thus to contain three hydrocarbon groups, octyl- Sepharose chromatography may be useful in the isolation of additional glycosyl-PtdIns species.

392 Seviever et al. (Euv. J . Bicichenz. 2-73)

The structure of GlcN(acy1)PtdIns was documented through several approaches. Its susceptibility to glycosyl-PtdIns-PLD (Figs 1 E and 7B) showed that it was a glycosyl-PtdIns, and its GlcN-Ins head group was established by several techniques including ESI-MS analysis. Another procedure that revealed this head group is introduced in this report: cation-exchange chromatography of acid hydrolyzates of the ['Hlinositol-labeled GlcN(acy1)PtdIns band scraped from a TLC plate. This procedure takes advantage of the relative stability of the linkage between GlcN and inositol to identify the GlcN-Ins conjugate by its characteristic retention time (Fig. 2). The procedure is based on previous documentation that GlcN(Me),-Ins can be identified by cation-exchange chromatography following fragmentation of radiomethylated glycosyl-Ptdlns species (Deeg et al., 1992a; Fig. 6). Radiomethylation was also used here to quantitate the equimolar ratio of GlcNhosi tol in purified GlcN(acy1)PtdIns (Table 1). Fatty acid analysis indicated an overall stoichiometry of about 2.3 mol fatty acids/mol inositol in purified GlcN(acy1)Ptdlns. Analysis of GlcN(acy1)inositol produced by H F fragmentation showed that palmitoyl (1 6:0) was the acyl group attached to inositol, leaving stearoyl (18:O) and oleoyl (18: l ) as the fatty acyl groups in the glycerolipid (Table 1). Previous lipid analysis of the glycosyl-PtdIns anchors of human erythrocyte acetylcholinesterase (Roberts et al., 1988a,b) and decay accelerating factor (Walter et al., 1990) showed palmitoyl attached to inositol, and thus this feature is conserved in the GlcN(acy1)PtdIns isolated here. Although its biological importance is not known, fatty acid acylation of inosi- to1 is known to convey Ptdlns-PLC resistance to glycosyl-PtdIns species (Roberts et al., 198813).

The lipid component of glycosyl-PtdIns species is also of interest because of another unresolved issue : mammalian free glycosyl-PtdIns species differ from glycosyl-PtdIns anchors in their glycerolipid structure. Most mammalian glycosyl-PtdIns anchors analyzed to date are derived from erythrocytes and are based exclusively on alkylacylglycerol (Roberts et al., 1988b; Walter et al., 1990; see McConville and Ferguson, 1993). Inter- estingly, a recent report on another mammalian glycosyl-PtdIns- anchored protein, CD-52 from human spleen, found only diacylglycerol-based anchor species (Treumann et al., 1995). The glycerolipid structures of free glycosyl-PtdIns species are not as well characterized; based on resistance to base treatment, < 5 % (Singh et al., 1994) to > 7 0 % (Puoti and Conzelmann, 1993) of the free glycosyl-PtdIns species in various cultured cell lines contain alkylglycerol. Base methanolysis indicated that about 15 5% of the GlcN(acy1)PtdIns purified here contained alkylglycerol, a value similar to that reported for GlcN-Ptdlns synthesized in vitro from murine lymphoma cell lysates (Stevens and Raetz, 1991) but less than that reported for a putative GlcN(acyl)PtdIns in lymphoma mutant E (Puoti and Conzelmann, 1993). Since less than 1 % of the PtdIns analyzed here was resistant to base, a value in good agreement with that reported for PtdIns from human erythrocytes (Butikofer et al., 1992), the difference may reflect a higher activity of GlcNAc transferase toward PtdIns species containing alkylacylglycerol than toward PtdIns species containing diacylglycerol. A potential further complication arose from the observation that the glycosyl-PtdIns species of some glycosyl-PtdIns-anchored proteins in yeast are based on ceramide (Fankhauser et al., 1993). The observation that GlcN- (acy1)PtdIns is completely cleaved by Crotalus PLA, (Fig. 7D), in contrast to its complete resistance to bee venom PLA,, ruled out any possibility that a subpopulation of the GlcN(acy1)PtdIns isolated here was based on ceramide.

The substantial conversion of GlcN(acy1)PtdIns to lipid b in even mildly alkaline buffers was quite unexpected (Fig. 7E). We confirmed that the conversion was promoted by alkaline condi-

tions by observing that about 30% of the 3H in purified GlcN- (acy1)PtdIns (pool 111 of Fig. 3A) shifted to lipid b after overnight incubation at 37°C in SO% n-propanol, 20mnM Tris/CI pH 9.0. In contrast, no increase in lipid b was observed after parallel incubation in 50% n-propanol, 200 mM sodium acetate pH 5.0 (data not shown). The substituents of GlcN(acy1)PtdIns that we expect to be most labile to alkaline buffers, fatty acid esters of hydroxyl groups, typically are not hydrolyzed at significant rates below pH 10-11 ; we confirmed by TLC that ['HIPtdIns was unaltered after overnight incubation at 37 "C in this pH-9.0 buffer. This suggested that a special feature of GlcN(acy1)PtdIns was responsible for its breakdown to lipid b. The breakdown did not appear to destroy the GlcN-Ins head group, as conjugates corresponding to GlcN-Ins were observed in acid hydrolyzates of TLC bands containing ['H]inositol-labeled lipid b (as in Fig. 2B) and radiomethylated lipid b (as in Fig. 6; data not shown). Deacylation was a possibility. However, the mobility of lipid b did not correspond to that observed for the lyso lipid produced by Crotalus PLA, cleavage of the sn-2 acyl group on glycerol (Fig. 7D), and lipid b was not susceptible to glycosyl-PtdIns-PLD (Fig. 7 B ) or to PtdIns-PLC (data not shown) although it should be if it resulted from loss of the acyl group on inositol. Furthermore, the lipid b product produced at pH 9.0 chromatographed at the position of GlcN(acy1)PtdIns on octyl-Sepharose (i.e. as a phospholipid still with three hydrocarbon chains; data not shown). Migration of the acyl group to another position on the inositol ring was also a possibility, but this structural change also does not explain the glycosyl-PtdIns- PLD resistance or the significant change in TLC mobility. Furthermore, acyl migration should be facilitated by acidic conditions (Ferguson, 1992) and not by base treatment. Although the current data do not allow proposal of a structure for lipid b, the breakdown of GlcN(acy1)Ptdlns to lipid b may be important because it appears to occur in vivo. When cells were labeled with ['H]inositol as in Fig. 1 and the label was then chased by incubation i n fresh medium containing unlabeled inositol for four days, the ratio of lipid b/GlcN(acyl)PtdIns in the lipid extract increased twofold over that in Fig. 1 A (data not shown).

We have noted above that this HeLa S3 subline appears rather unique in expressing GlcN(acy1)PtdIns as well as glycosyl-PtdIns-anchored proteins. This lipid was undetectable in the CCL-2 HeLa cell line and present at much lower levels in the S3 parental line (Sevlever et al., 1994). Identification of a cell line that produces large quantites of GlcN(acy1)PtdIns is important, not only because it permitted the structural characterization described here, but also because it allowed us to demonstrate that exogenous inositol phospholipids can enter the glycosyl- PtdIns biosynthetic pathway (Wongkajornsilp and Rosenberry, 1995). It is also important to determine whether the GlcN(a- cy1)PtdIns accumulated in these cells actually serves as an intermediate in the glycosyt-PtdIns biosynthetic pathyway, and this question may be addressed by studies of the uptake and metabolism of radiolabeled GlcNPtdIns which should soon be available.

This work was supported by Grant DK38181 from the National Insti- tutes of Health. We thank Dr Bruce Reinhold of the Harvard School of Public Health for obtaining the ESI-MS spectra.

REFERENCES Ames, B. N. (1966) Assay of inorganic phosphate, total phosphate and

phosphatases, Methods Enzyrnol. 8, 115-117. Bordier, C. (1981) Phase separation of integral membrane proteins in

Triton X-114 solution, J . Biol. Chem. 256, 1604-1607. Butikofer, P., Zollinger, M. & Brodbeck, U. (1992) Alkylacyl glycero-

phosphoinositol in human and bovine erythrocytes. Molecular spe- c k s composition and comparison of glycosyl-inositolphospholipid

Sevlever et al. (Eur: 1. Biochem. 233) 393

anchors of erythrocyte acetylcholinesterases, Eur: J Biochern. 208,

Camp, L. A., Chauhan, P., Farrar, J . D. & Lehrman. M. A. (1993) De- fective mannosylation of glycosylphosphatidylinositol in Lec35 Chi- nese hamster ovary cells, J. Biol. Chem. 268, 6721 -6728.

Ciucanu, 1. & Kerek, F. (1984) A simple and rapid method for the permethylation of carbohydrates, Carbohyd?: Res. 131, 209-21 7.

Costello, L. C. & Orlean, P. (1992) Inositol acylation of a potential gly- cosy1 phosphhoinositol anchor precursor from yeast requires acyl coenzyme A, J. Biol. Chern. 267, 8599-8603.

Deeg, M. A., Humphrey, D. R., Yang, S. H., Ferguson, T. R., Reinhold, V. N. & Rosenbeiry, T. L. (1992a) Glycan components in the glycoinositol phospholipid anchor of human erythrocyte acetylcholinesterase. Novel fragments produced by tritluoroacetic acid, J . B i d . Chem. 267, 18 573 - 18 580.

Deeg, M. A., Murray, N. R. & Rosenberry, T. L. (1992b) Identification of glycoinositol phospholipids in rat liver by reductive radiomethylation of amines but not in H4IIE hepatoma cells or isolated hepato- cytes by biosynthetic labeling with glucosamine, J . Biof. Cheni. 267,

Doering, T. L., Masterson, W. J., Englund, P. T. & Hart, G. W. (1989) Biosynthesis of the glycosyl phosphatidylinositol membrane anchor of the trypanosome variant surface glycoprotein. Origin of the non- acetylated glucosamine, J . Biol. Cliem. 264, 11 168-11 173.

Englund, P. T. (1993) The structure and biosynthesis of glycosyl phosphatidylinositol protein anchors, Annu. Rev. Biochem. 62, 121 -138.

Fankhauser, C., Homans, S. W., Thomas-Oates, J. E., McConville, M. J., Desponds, C., Conzelmann, A. & Ferguson, M. A. J . (1993) Struc- tures of glycosylphosphatidylinositol membrane anchors from Saccharomyces cerevisiae, J. B id . Chem. 268, 26 365 -26 374.

Fatemi, S. H., Haas, R., Jentoft, N., Rosenberry, T. L. & Tartakoff, A. L. (1987) The glycophospholipid anchor of Thy-1. Biosynthetic labeling experiments with wild-type and class E Thy-1 negative lym- phomas, J . Biol. Chem. 262, 4728-4732.

Ferguson, M. A. J. (1992) Chemical and enzymic analysis of glycosylphosphatidylinositol anchors, in Lipid nzodifcations of proteins; a prucrical approach (Turner, A. J. & Hooper, N., eds) pp. 191 -230, IRL Press, Oxford.

Field, M. C., Menon, A. K. & Cross, G. A. M. (1991) Developmental variation of glycosylphosphatidylinositol membrane anchors in Try- panosoma hrucei. Identification of a candidate biosynthetic precursor of the glycosylphosphatidylinositol anchor of the major procyclic stage surface glycoprotein, J . B i d . Chem. 266, 8392-8400.

Guther, M. L. S., Masterson, W. J. & Ferguson, M. A. J. (1994) The effects of phenylmethylsulfonyl fluoride on inositol-acylation and fatty acid remodeling in African trypanosonies, J. Bin/. Client. 269, 18 694- 18 701.

Haas, R. & Rosenberry, T. L. (1985) Quantitative identification of N- terminal amino acids in proteins by radiolabeleed reductive methylation and amino acid analysis : Application to human erythrocyte acetylcholinesterase, Anal. Biochem. 148, 154- 162.

Haas, R., Brandt, P. T., Knight, J. & Rosenberry, T. L. (1986) Identifica- tion of amine components in a glycolipid membrane-binding domain at the C-terminus of human erythrocyte acetylcholinesterase, Bio- chemistry 25, 3098 - 31 05.

Henner, D. J., Yang, M., Chen, W., Hellmiss, R., Rodriguez, H. & Low, M. G. (1988) Sequence of the Bacillus fhuringiensis phosphatidylinositol specific phospholipase C, Nucleic Acids Res. I6, 10383.

Hirose, S . , Ravi, L., Hazra, S. & Medof, M. E. (1991) Assembly and deacetylation of N-acetylglucosaminyl-plasmanylinositol in normal and affected paroxysmal nocturnal hemoglobinuria cells, Proc. Nut1 Acad. Sci. USA 88, 3762-3766.

Hirose, S., Prince, G. M., Sevlever, D., Ravi, L., Rosenberry, T. L., Ueda, E. & Medof, M. E. (1992) Characterization of putative glycoinositol phospholipid anchor precursors in mammalian cells. Localization of phosphoethanolamine, J. B i d . Chem. 267, I6 968- 16 974.

Homans, S. W., Ferguson, M. A. J., Kjwek, R. A,, Rademacher, T. W., Anand, R. & Williams, A. F. (1988) Complete structure of the glycosyl phosphatidylinositol membrane anchor of rat brain Thy-l glycoprotein, Nature 333, 269-272.

Kamitani, T., Menon, A. K., Hallaq, Y., Warren, C. D. & Yeh, E. T. H. (1992) Complexity of ethanolamine phosphate addition in the biosynthesis of glycosylphosphatidylinositol anchors in mammalian cells, J . Biol. Chem. 267, 24611-24619.

677 -683.

18 581 -1 8588.

Lee, H.-C., Shoda, R., Krall, J. A,, Foster, J. D., Selhub, J. & Rosenberry, T. L. (1992) Folate binding protein from kidney brush border membranes contains components characteristic of a glycoinositol phospholipid anchor, Biochemi.qj 31, 3236-3243.

Masterson, W. J . , Doering, T. L., Hart, G. W. & Englund, P. T. (1989) A novel pathway for glycan assembly: Biosynthesis of the glycosylphosphatidylinositol anchor of the trypanosome variant surface glycoprotein, Cell 56, 793-800.

McConville, M. J . & Bacic, A. (1989) A family of glycoinositol phospholipids from Leislimnnia major. Isolation, characterization, and an- tigenicity, J. B id . Chem. 264, 757-766.

McConville, M. J . & Ferguson, M. A. J. (1 993) The structure, biosynthesis and function of glycosylated phosphatidylinositols in the parasitic protozoa and higher eukaryotes, Biochem. J. 294, 305-324.

Medof, M. E., Walter, E. I., Roberts, W. L., Haas, R. & Rosenberry, T. L. (1986) Decay accelerating factor of complement is anchored to cells by a C-terminal glycolipid, Biochemistry 25, 6740-6747.

Menon. A. K., Schwarz, R. T., Mayor, S. & Cross, G. A. M. (1990) Cell-free synthesis of glycosyl-phosphatidyliiiositol precursors for the glycolipid membrane anchor of Trypano.roma brucei variant surface glycoproteins. Structural characterization of putative biosynthetic intermediates, J. B id . Chenz. 265, 9033 -9042.

Mohney, R. P., Knez, J. J., Ravi, L., Sevlever, D., Rosenberry, T. L., Hirose, S. & Medof, M. E. (1994) Glycoinositol phospholipid anchor-defective K562 mutants with biochemical lesions distinct from those in the Thy-I murine lymphoma mutants, 1. Biol. Chern. 269,

Puoti, A. & Conzelmann, A. (1992) Structural characterization of free glycolipids which are potential precursors for glycophosphatidylino- sitol anchors in mouse thymoma cell lines, J. B i d Chem. 267,

Puoti, A. & Conzelmann, A. (1993) Characterization of abnormal free glycophosphatidylinositols accumulating in mutant lymphoma cells of classes B, E, E and H, J . Biol. Chem. 268, 7215-7224.

Roberts, W. L., Santikarn, S., Reinhold, V. N. & Rosenberry, T. L. (1988a) Structural characterization of the glycoinositol phospholipid membrane anchor of human erythrocyte acetylcholinesterase by fast atom bombardment mass spectrometry, J. B id . Chem. 263, 18776- 18784.

Roberts, W. L., Myher, J. J. , Kuskis, A., Low, M. G. & Rosenbeny, T. L. (1988 b) Lipid analysis of the glycoinositol phospholipid membrane anchor of human erythrocyte acetylcholinesterase. Palmitoylation of inositol results in resistance to phosphatidylinositol-specific phospholipase C, J . Biol. Chenz. 243, 18 766- 18 775.

Rosenberry, T. L., Sevlever, D. & Medof, M. E. (1994) Metabolism of GPIs in mammalian cells, Bmz. J. Med Biol. Res. 27, 151-159.

Sevlever, D. & Rosenberry, T. L. (1993) Mannosamine inhibits the synthesis of putative glycoinositol phospholipid anchor precursors in mammalian cells without incorporating into an accumulated intermediate, J. B id . Chem. 268, 10938-10945.

Sevlever, D., Humphrey, D. R., Schiemann, D. & Rosenberry, T. L. (1994) Differential expression of GlcN(acy1)PI among Hela cell lines, Mol. B id . Cell 5, 1869.

Singh, N . & Tartakoff, A. M. (1991) Two different mutants blocked in synthesis of dolichol-phosphoryl-mannose do not add glycophospholipid anchors to membrane proteins : Quantitative correction of the phenotype of a CHO cell mutant with tunicamycin, Mol. Cell. Biol.

Singh, N., Zoeller, R. A,, Tykocinski, M. L., Lazarow, P. B. & Tartakoff, A. M. (1994) Addition of lipid substituents of mammalian protein glycosylphosphoinositol anchors, Mol. Cell. B i d . 14, 21 -31.

Stahl, N., Baldwin, A,, Hecker, R., Pan, K.-M., Burlingame, A. L. & Prusiner, S. B. (1992) Glycosylinositol phospholipid anchors of the scrapie and cellular prion proteins contain sialic acid, Biochemistry 31, 5043-5053.

Stevens, V. L. & Raetz, C. R. H. (1991) Defective glycosyl phosphatidylinositol biosynthesis in extracts of three Thy-1 negative lymphoma cell mutants, J. Biol. Chem. 266, 10039-10042.

Stevens, V. L. (1 993) Regulation of glycosylphosphatidylinositol biosynthesis by GTP. Stimulation of N-acetylglucosamine-phosphatidylinositol deacetylafon, J. B id . Cliern. 268, 971 8-9724.

Stevens, V. L. & Zhang, H. (1994) Coenzyme A dependence of glycosylphosphatidylinositol biosynthesis in a mammalian cell-free system, J. Biol. Chem. 269, 31 397-31 403.

6536 - 6542.

22673-22680.

11, 391 -400.

394 Sevlever et al. ( E U K J. Biochem. 233)

Sugiyama, E., DeGasperi, R., Urakaze, M., Chang, H.-M., Thomas, L. J., Hyman, R., Warren, C. D. & Yeh, E. T. H. (1991) Identification of defects in glycosylphosphatidylinositol anchor biosynthesis in the Thy-I expression mutants, J. Biol. Chern. 266, 12 1 19- 12 122.

Taguchi, R., Hamakawa, N., Harada-Nishida, M., Fukui, T., Nojirna, K. & Ikezawa, H. (1994) Microheterogeneity in glycosylphosphatidylinositol anchor structures of bovine liver S’hucleotidase, Bio- chemistry 33, 1017-1022.

Thomas, L. J., Urakaze, M., DeGasperi, R., Karnitani, T., Sugiyama, E., Chang, H.-M., Warren, C. D. & Yeh, E. T. H. (1992) Differential expression of glycosylphosphatidylinositol-anchored proteins in a murine T cell hybridorna mutant producing limiting amounts of the glycolipid core. Implications for paroxysmal nocturnal hemoglobinuria, J . Clin. Invest. 89, 1172-1 177.

Treumann, A., Lifely, M. R., Schneider, P. & Ferguson, M. A. J. (1995) Primary structure of CD52, J. Biol. Chem. 270, 6088-6099.

Ueda, E., Sevlever, D., Prince, G. M., Rosenberry, T. L., Hirose, S. & Medof, M. E. (1993) A candidate mammalian glycoinositol phospholipid precursor containing three phosphoethanolamines, 1. Biol. Clzern. 258, 9998-10002.

Urakaze, M., Kamitani, T., DeGasperi. R., Sugiyama, E., Chang, H.-M., Warren, C. D. & Yeh, E. T. H. (1992) Identification of a missing link in glycosylphosphatidylinositol anchor biosynthesis in mammalian cells, J . Biol. Clzem. 267, 6459-6462.

Walter, E. I., Roberts, W. L., Rosenberry, T. L., Ratnoff, W. D. & Medof, M. E. (1990) Structural basis for variations in the sensitivity of human decay accelerating factor to phosphatidylinositol-specific phospholipase C cleavage, J. Im~nunol. 144, 1030- 1036.

Wongka,jornsilp, A. & Rosenberry, T. L. (1995) Uptake of exogenous sn-I-acyl-2-lyso-phosphatidylinositol into HeLa S3 cells. Reacyla- tion on the cell surface and metabolism to glucosaminyl(acy1)phosphatidylinositol, J . B i d . Chern. 270, 9147-9153.

Documents

Compositional analysis of Glucosaminyl(acyl)phosphatidylinositol Accumulated in HeLa S3 Cells