Phosphatidylethanolamine is the donor of the ethanolamine residue linking a glycosylphosphatidylinositol anchor to protein.

Numerous cell surface glycoproteins from eukaryotic organisms including African trypanosomes and budding yeast (Saccharomyces cerevisiae), are anchored to the lipid bilayer by a glycophospholipid, glycosylphosphatidylinositol, covalently linked to the carboxyl terminus of the protein via a phosphoethanolamine bridge. In this paper we describe metabolic labeling experiments aimed at identifying the biosynthetic origin of the ethanolamine residue in the phosphoethanolamine bridge. Using yeast mutants generated by disruption of the ethanolaminephosphotransferase (EPT1) and cholinephosphotransferase (CPT1) genes, we report data consistent with the proposal that the ethanolamine residue is derived from phosphatidylethanolamine.

Numerous cell surface glycoproteins from eukaryotic organisms including African trypanosomes and budding yeast (Saccharomyces cerevisiae), are anchored to the lipid bilayer by a glycophospholipid, glycosylphosphatidylinositol, covalently linked to the carboxyl terminus of the protein via a phosphoethanolamine bridge. In this paper we describe metabolic labeling experiments aimed at identifying the biosynthetic origin of the ethanolamine residue in the phosphoethanolamine bridge. Using yeast mutants generated by disruption of the ethanolaminephosphotransferase ( E P T I ) and cholinephosphotransferase (CPTI) genes, we report data consistent with the proposal that the ethanolamine residue is derived from phosphatidylethanolamine.
A wide variety of cell surface glycoproteins are associated with the plasma membrane through a glycosylphosphatidylinositol (GPI)' membrane anchor. This complex glycophospholipid is covalently linked to the carboxyl-terminal amino acid of the mature protein via a phosphoethanolamine bridge (1-4). GPI-anchored proteins have been found in numerous eukaryotic organisms, including trypanosomes, budding yeast (Saccharomyces cerevisiae) (5-9), and mammals. Current data suggest that GPI anchors are synthesized as precursor glycolipids which are then covalently coupled to target proteins immediately (within 1-5 min) after completion of protein synthesis. Candidate GPI anchor precursors have been identified in African trypanosomes (10-15) and in mammalian cells (16)(17)(18)(19), and the transfer of structurally characterized precursors to protein has been demonstrated in vitro (20).
Recent biochemical studies using trypanosome membrane preparations indicate that the GPI anchor precursor (ethanolamine-P-Man3GlcN-PI) is assembled by sequential addition of the various components (glucosamine, mannose, phosphoethanolamine) to phosphatidylinositol (21,22). N-Acetylglucosamine is incorporated directly from the sugar nucleotide (23), but mannose first combines with endogenous dolichol phosphate to form dolichol-P-mannose and is then transferred to the GPI anchor core (24) (Fig. 1). Phosphoethanolamine addition in vitro neither requires nor is affected by exogenously added CDP-ethanolamine (21), suggesting that this moiety is derived from an endogenous precursor. A major ethanolamine-containing compound in cells and a potential candidate for this donor is phosphatidylethanolamine (PE). Because the cell-free systems used to study GPI biosynthesis require the presence of cellular membranes, the dependence on this phospholipid has not been tested. Attempts to demonstrate that [3H]ethanolamine is transferred from PE to GPI with metabolic radiolabeling studies have been complicated by the fact that exogenously supplied ethanolamine is first converted to CDP-ethanolamine before being incorporated into PE. Although both GPI and PE are radiolabeled under these conditions, the presence of CDP-ethanolamine makes it difficult to rule out this compound as the ethanolamine donor. To circumvent this problem, we have performed metabolic labeling experiments using yeast mutants unable to synthesize PE from ethanolamine via CDP-ethanolamine because of disruptions in the ethanolaminephosphotranferase (EPT1) and cholinephosphotransferase (CPT1) genes (25-28) (Fig. 1). The mutant cells grow normally and have a normal phospholipid composition* because of the redundancy in the pathways for the biosynthesis of phosphatidylethanolamine and phosphatidylcholine (PE can be synthesized by decarboxylation of phosphatidylserine, and phosphatidylcholine can be synthesized by methylation of phosphatidylethanolamine) (29,30).
Metabolic Labeling-Single yeast colonies were transferred to synthetic medium and incubated overnight (with shaking) in a water bath at 30 "C. Growth was monitored by measurement of absorbance at 600 nm using disposable polystyrene cuvettes (1-cm path) and a CARY 219 spectrophotometer. Aliquots of dense suspensions were * S. Morash, personal communication.  anchor precursor (ethanolamine-P-MannGlcN-PI) is assembled by sequential addition of the various components (glucosamine, mannose, phosphoethanolamine) to phosphatidylinositol. The glucosamine residue is derived from UDP-GlcNAc, and the three mannose residues are derived from dolichol-phosphomannose. The data presented in this paper are consistent with the proposal that the ethanolamine residue is derived from phosphatidylethanolamine (PE). The reaction catalyzed by the EPTl gene product is indicated (the same reaction is also catalyzed inefficiently by the CPTl gene product); CDP-ethanolamine is generated from ethanolamine via ethanolamine-phosphate. PE is synthesized via the CDP-ethanolamine pathway, or via decarboxylation of phosphatidylserine (PS). GPI species containing a fatty acid esterified to the inositol residue (see Refs. 12,14,22,50,and 51) have been omitted from the diagram for simplicity. PI, phosphatidylinositol; OAc, acetate; GlcN, glucosamine; GlcNAc, N-acetylglucosamine; Man, mannose; dol-P-Man, dolicholphosphomannose; EtN, ethanolamine; P, phosphate.
diluted -10-fold into fresh medium, and the incubation was continued to obtain a suitable number of logarithmically growing cells (Asw < 1.2) for metabolic labeling. For ['HH]ethanolamine labeling the cells (-15 ml, Am -0.7) were pelleted by centrifugation, resuspended in -1 ml of complete medium, and incubated with 50-300 pCi of [l-'HH] ethanolamine (Amersham Corp., 29.5 Ci/mmol) for 2 h in a water bath (with shaking) at 30 "C. In some experiments, the incubation was continued for an additional 1.5-3 h after adding -1 ml of fresh medium to the culture. For [3H]inositol labeling, logarithmically growing cells were washed and resuspended in vitamin-free medium (cell densities and volumes as above), then incubated with 50-300 pCi of [2-'HJinositol (Amersham Corp. or American Radiolabeled Chemicals Inc., -20 Ci/mmol) as above. Labeled samples were taken for lipid or protein analysis as described below. Lipid Extraction-At the end of the incubation, [3H]ethanolaminelabeled cells were washed, resuspended in 160 pl of fresh medium, and processed by the method of Bligh and Dyer (31). Briefly, 600 pl of chloroform/methanol(l:2, v/v), and a small quantity of glass beads (Sigma, 425-600 pm, acid-washed) were added to each sample, and the cells were broken by agitation on a vortex mixer for 5-15 min at 4 "C. Then 200 pl of chloroform and 200 p1 of water were added to each sample, and after vigorous mixing on a vortex mixer, the extracts were separated into two phases by centrifugation. Aliquots of each phase were taken for liquid scintillation counting and thin layer chromatographic analysis using solvent system A (see below).
Protein Extraction and Analysis-Radiolabeled proteins were extracted into Triton X-114 and subjected to PI-PLC treatment essentially as described by Orlean (32). Proteins in the detergent and aqueous phases after PI-PLC treatment were precipitated with trichloroacetic acid, washed with acetone, dissolved in SDS-PAGE sample buffer, and analyzed by electrophoresis in 10% polyacrylamide gels followed by autoradiography. Depending on the number of cells used and the amount of radiolabeled precursor, exposures of 1-5 weeks were required for visualization of labeled proteins.
Acid Hydrolysis-In order to determine the identity of the radioactivity in labeled lipids and water-soluble metabolites extracted from [3H]ethanolamine-labeled cells, aliquots of each phase of a Bligh-Dyer extract (see above) were subjected to acid hydrolysis. Hydrolysis was performed in 1-ml Reacti-Vials (Pierce Chemical Co.). Samples were dried in the vials using a Speed-Vac evaporator (Savant Instruments). Then 400 pl of 6 N HCl (Pierce) was added to each vial, and the vials were tightly sealed and incubated at 100 "C for 6 h. At the end of the incubation, the hydrolysates were dried in a Speed-Vac evaporator, resuspended in water, and analyzed by thin layer chromatography using solvent system B (see below).

RESULTS
The major water-soluble metabolites produced by yeast biosynthetically labeled with [3H]ethanolamine were ethanolamine phosphate and CDP-ethanolamine ( Fig. 2A, wt). These compounds were also efficiently synthesized by yeast mutants with single ( e p t l ) or double (eptl, cptl) gene disruptions ( Fig.  2A). The major phospholipids radiolabeled in wild-type yeast co-migrated with standards for phosphatidylethanolamine (PE) and phosphatidylcholine (PC) when analyzed by thin layer chromatography (Fig. 2B, wt). The eptl null mutant produced reduced levels of these two phospholipids (Fig. 2B,  e p t l ) . Despite the disruption of the EPTl gene, phosphoethanolamine is transferred from CDP-ethanolamine to form PE urd'-52 trpl-289) (referred to as wild-type, wt) and its derivatives HJ051 (eptl-AI::URAS) (referred to as eptl) and HJOOO (eptl-Al::URAJ cptkLEU2) (referred to as eptl, cptl) were metabolically labeled for 2 h at 30 "C, and lipids were extracted as described under "Materials and Methods." Identical aliquots of each upper phase were applied to silica 60 thin layer plates and chromatographed using solvent system B (panel A ) . Different aliquots of each lower phase (wt, 7.5 pl; eptl, 30 pk eptl, cptl, 60 pl, out of a total of 300 pl in each case) were chromatographed on silica 60 using solvent system A (panel B). Radiolabeled standards were analyzed alongside; radioactivity was detected by spraying the plates with EN'HANCE (Du Pont-New England Nuclear) and exposing them to Kodak XAR-5 film for 5 days at -70 "C. The material chromatographing ahead of CDP-EtN in panel A was not identified. The minor labeled spots in panel B are likely to be lyso PE and lyso PC. Only ["Hlethanolamine was recovered after acid hydrolysis of the upper phase; ['HJethanolamine and ['Hlcholine were recovered after acid hydrolysis of the lower phase. EtN, ethanolamine, EtN-P, ethanolamine-phosphate; CDP-EtN, CDP-ethanolamine; PE, phosphatidylethanolamine; PC, phosphatidylcholine.

Phosphatidylethanolamine Is the GPI Ethanolamine Donor
in these cells because the reaction is catalyzed, albeit inefficiently, by the CPTl gene product in uiuo (26). A double null mutant (eptl, cptl ) showed enormously reduced activity (-1%) compared with wild-type (Fig. 223, middle lane), and did not synthesize radiolabeled PE from ['H]ethanolamine (the faint residual labeling probably represents the contribution of the base-exchange pathway to PE biosynthesis (33), although such an activity has not been previously identified in yeast). Metabolic labeling with ['H]inositol showed that GPI-anchor biosynthesis was unimpaired in the double mutant (Fig.  3). Both wild-type and mutant showed identical patterns of labeled proteins (analyzed after detergent phase separation using Triton X-114; Fig. 3, lanes 2 and 5 ) , consisting of a characteristic smear of high molecular weight material as well as a ladder of lower molecular weight bands. Treatment with phosphatidylinositol-specific phospholipase C (PI-PLC) reduced the intensities of all the labeled bands in the detergent phases (Fig. 3, compare lanes 3 and 7 with lanes 2 and 5 ) , and some of these were recovered in the aqueous phases (Fig. 3,   lanes 4 and 8). The most prominent of these was a band at -125 kDa that probably corresponds to the GPI-anchored GASl gene product (5, 7, 34); a fainter band at -110 kDa could also be visualized (Fig. 3, lanes 4 and 8; see also Fig. 4,  lanes 2 and 4 ) . The inefficient recovery of the other cleaved proteins has been noted previously (5,7,32) but is yet to be explained.
If the ethanolamine bridge in the GPI anchor of proteins is derived from PE, then yeast cells deficient in the E P T l and CPTl gene products will not be able to incorporate ["HI ethanolamine into GPI. Wild-type and mutant yeast strains were grown in ['Hlethanolamine and labeled proteins were extracted, Triton X-114-selected, treated with or without PI-PLC, and analyzed by SDS-PAGE and fluorography.  strains (wt, and eptl, c p t l ) were metabolically labeled with ['Hlethanolamine or ['H]inositol and extracted with Triton X-114 as described under "Materials and Methods." Detergent phases were dissolved in PI-PLC buffer on ice, then incubated with or without PI-PLC at at 30 "C. Phases were separated, and radiolabeled proteins in the aqueous phase were analyzed by SDS-PAGE. The figure shows the portion of the autoradiogram containing the PI-PLC-sensitive -125-and -110-kDa proteins described in Fig. 3. Both proteins could also be seen by Coomassie staining in lanes 2 , 4 , 6 , and 8. The positions of molecular weight markers are indicated on the left of the figure.

DISCUSSION
A set of yeast proteins, including a major 125-kDa membrane glycoprotein, that can be metabolically labeled with ['Hlethanolamine, ['Hlpalmitic acid, and ['H]inositol have been shown to be GPI-anchored by a number of criteria including susceptibility to hydrolysis by phosphatidylinositolspecific phospholipase C and the presence of the cross-reacting determinant found in other GPI anchors (this paper, and Refs. 5-9). That these proteins are not labeled by ["Hlethanolamine in mutant yeast unable to transfer this compound from CDP-ethanolamine to PE argues strongly that the phospholipid is the immediate donor of the ethanolamine residue linking a GPI anchor to protein, rather than CDP-ethanolamine.
Additional evidence that PE is the donor of the ethanolamine bridge in GPI can be obtained by taking advantage of the redundancy in the biosynthetic pathways for this phospholipid. Because PE can be produced by decarboxylation of phosphatidylserine, growth of cells in ["]serine should result in the metabolic labeling of PS, PE, and, if derived directly from PE, GPI. Unfortunately, such an experiment is not technically possible in yeast. To date, ethanolamine-containing GPI precursors generated either in uiuo or in uitro from yeast have not been identified and characterized. Therefore, protein-associated GPI would have to be analyzed for radiolabeled ethanolamine derived from ["]serine. To do this, a GPI-modified peptide free of serine must first be isolated. Then, the GPI anchor must be acid-hydrolyzed and the components analyzed to verify that the radiolabel is in ethanolamine. Although we have attempted this experiment with yeast, we have been unable to isolate sufficient amounts of radioactive material to complete the analyses.? This is probably because the efficiency of uptake and utilization of exogenously supplied compounds by yeast is extremely low. Such ' A. K. Menon, unpublished data. experiments have been performed with African trypanosomes. Radiolabeled serine was found in PE and various GPI species as ethanolamine when trypanosomes were metabolically labeled with [3H]serine4; no radiolabeled water-soluble ethanolamine metabolites were generated. The complementing evidence that PE is the donor of the GPI ethanolamine from the yeast and trypanosome systems suggests that this finding is not specific for either organism but is true for all eukaryotes.
The experimental strategy employed in this paper cannot indicate wLether or not the phosphoethanolamine moiety is transferred as a group. Again, the poor labeling efficiency of yeast prohibit the experiment in which cells are grown in ["'Plphosphate and [3H]ethanolamine, and the GPI anchor analyzed for the appearance of both radiolabels to address this question. However, in vitro labeling experiments using trypanosome membranes indicate that the phosphoethanolamine is indeed transferred as a single m~i e t y .~ Related reactions involving the transfer of phosphocholine (35,36), phosphoethanolamine (37, 38), and phosphoinositol (39) to ceramide have been previously described in mammalian cells and in yeast, and glycerophospholipids have been shown to be the donors of the phosphoethanolamine and phosphoglycerol substituents on the membrane-derived oligosaccharides of Escherichia coli (40,41).
The result that a lipid donor is involved in ethanolamine addition has implications for the membrane topology of GPI biosynthesis. GPI anchor precursors are expected to be lumenally oriented in the membrane of the ER in order to be transferred to newly translocated polypeptides. GPI assembly may be initiated in the lumenal leaflet of the ER since both UDP-GlcNAc and PI have access to the ER lumen (42,43). Based on the observation that dolichol-phosphomannose is the mannosyl donor for GPIs (16, 24, 32,45-47), it has been suggested that mannosylation steps in GPI assembly occur in the lumenal leaflet of the ER membrane (24, 32, 47; see also Ref. 18). Since PE is found in both leaflets of the ER membrane (44), the final ethanolamine addition step could also take place in the lumenal leaflet to generate GPI species appropriately located for transfer to proteins.
In addition to the phosphoethanolamine bridge linking the GPI anchor to the carboxyl terminus of the protein, the GPI anchors of some proteins (48,49) as well as free GPIs have additional phosphoethanolamine substituents ( 19)5 on the mannose residue proximal to the glucosamine, and possibly on the middle mannose of the tri-mannose glycan core? The immediate precursor of these extra phosphoethanolamine groups is not known, but if they are acquired as a late event during transit of the GPI-anchored protein or free GPI through the secretory pathway to the cell surface, the precursor must be available in the lumenal aspect of the appropriate intracellular organelle. Since CDP-ethanolamine is not known to be transported across organellar membranes, these extra phosphoethanolamine residues may also be derived from PE.