Glycosylphosphatidylinositol (GPI) proteins of Saccharomyces cerevisiae contain ethanolamine phosphate groups on the alpha1,4-linked mannose of the GPI anchor.

In humans and Saccharomyces cerevisiae the free glycosylphosphatidylinositol (GPI) lipid precursor contains several ethanolamine phosphate side chains, but these side chains had been found on the protein-bound GPI anchors only in humans, not yeast. Here we confirm that the ethanolamine phosphate side chain added by Mcd4p to the first mannose is a prerequisite for the addition of the third mannose to the GPI precursor lipid and demonstrate that, contrary to an earlier report, an ethanolamine phosphate can equally be found on the majority of yeast GPI protein anchors. Curiously, the stability of this substituent during preparation of anchors is much greater in gpi7Delta sec18 double mutants than in either single mutant or wild type cells, indicating that the lack of a substituent on the second mannose (caused by the deletion of GPI7) influences the stability of the one on the first mannose. The phosphodiester-linked substituent on the second mannose, probably a further ethanolamine phosphate, is added to GPI lipids by endoplasmic reticulum-derived microsomes in vitro but cannot be detected on GPI proteins of wild type cells and undergoes spontaneous hydrolysis in saline. Genetic manipulations to increase phosphatidylethanolamine levels in gpi7Delta cells by overexpression of PSD1 restore cell growth at 37 degrees C without restoring the addition of a substituent to Man2. The three putative ethanolamine-phosphate transferases Gpi13p, Gpi7p, and Mcd4p cannot replace each other even when overexpressed. Various models trying to explain how Gpi7p, a plasma membrane protein, directs the addition of ethanolamine phosphate to mannose 2 of the GPI core have been formulated and put to the test.


From the Department of Medicine, University of Fribourg, CH-1700 Fribourg, Switzerland
In humans and Saccharomyces cerevisiae the free glycosylphosphatidylinositol (GPI) lipid precursor contains several ethanolamine phosphate side chains, but these side chains had been found on the protein-bound GPI anchors only in humans, not yeast. Here we confirm that the ethanolamine phosphate side chain added by Mcd4p to the first mannose is a prerequisite for the addition of the third mannose to the GPI precursor lipid and demonstrate that, contrary to an earlier report, an ethanolamine phosphate can equally be found on the majority of yeast GPI protein anchors. Curiously, the stability of this substituent during preparation of anchors is much greater in gpi7⌬ sec18 double mutants than in either single mutant or wild type cells, indicating that the lack of a substituent on the second mannose (caused by the deletion of GPI7) influences the stability of the one on the first mannose. The phosphodiesterlinked substituent on the second mannose, probably a further ethanolamine phosphate, is added to GPI lipids by endoplasmic reticulum-derived microsomes in vitro but cannot be detected on GPI proteins of wild type cells and undergoes spontaneous hydrolysis in saline. Genetic manipulations to increase phosphatidylethanolamine levels in gpi7⌬ cells by overexpression of PSD1 restore cell growth at 37°C without restoring the addition of a substituent to Man2. The three putative ethanolamine-phosphate transferases Gpi13p, Gpi7p, and Mcd4p cannot replace each other even when overexpressed. Various models trying to explain how Gpi7p, a plasma membrane protein, directs the addition of ethanolamine phosphate to mannose 2 of the GPI core have been formulated and put to the test.
Many glycoproteins of lower and higher eukaryotes are attached to the plasma membrane by means of a glycosylphosphatidylinositol (GPI) 1 (1,2). The carbohydrate structure link-ing the C-terminal end of GPI proteins to the lipid moiety is identical in GPI anchors from all organisms analyzed so far, namely protein-CO-NH-(CH 2 )-PO 4 -6Man␣1-2Man␣1-6Man␣1-4GlcNH 2 -inositol-PO 4 -lipid but the GPI anchors from various species differ widely with regard to the side chains attached to this core structure as well as the lipid moieties of the anchor (1). This report concerns the ethanolamine phosphate (EtN-P) side chains, which are often present on mannoses 1 and 2 of the core structure (Man1 and Man2 in Fig. 1). Indeed, an EtN-P is invariably found on Man1 of GPI proteins and on the GPI lipids in mammals and in Torpedo californica, but is not found in Trypanosoma brucei, Leishmania major, or Plasmodium falciparum (3)(4)(5)(6)(7)(8)(9)(10). EtN-P attached to Man1 and Man2, respectively, has also been identified on the complete GPI precursor lipid CP2 of Saccharomyces cerevisiae (11)(12)(13) although the identity of the substituent on Man2 has not been formally demonstrated. The presence of these side chains on CP2 came as a surprise in as much as a previous analysis of the pool of the protein-linked GPI anchors of S. cerevisiae had failed to reveal EtN-P or other substituents on Man1 or Man2 (14). Candidate EtN-P transferase genes required for the attachment of these substituents have been identified in humans and yeast: PIG-N and MCD4 are involved in the transfer of EtN-P from phosphatidylethanolamine (PE) onto Man1 (13,15,16), GPI7 in the transfer onto Man2 (12), and PIG-O and GPI13 in the transfer of EtN-P from PE onto Man3 (17)(18)(19)(20)(21). These genes are homologous to each other, are found throughout the eukaryotic kingdom, possess an N-terminal globular domain facing the lumen of the ER or the extracellular space and multiple transmembrane domains in their C terminus. They are good candidates for the EtN-P transferases themselves, as the N-terminal, globular domains have a distinct homology with verified phosphodiesterases and because the corresponding mutants accumulate GPI lipid intermediates that lack EtN-P residues. A further gene named PIG-F has been implied in the addition of EtN-P side chains, as PIG-F mutants fail to add EtN-P to Man3 in mammalian cells (8,9,22,23). GPI11, the yeast homologue of PIG-F, is not required for the addition of EtN-P to Man3 but may be required for the addition of a hydrofluoric acid (HF) labile substituent to Man2 (18). In yeast, deletion of MCD4, GPI11, or GPI13 is lethal, whereas deletion of GPI7 only compromises cell wall integrity (15,18,19,24). This is not unexpected for GPI13, as in all GPI proteins analyzed it invariably is the EtN-P on Man3 that links the GPI to the protein, and even in mutants that cannot add EtN-P to Man3 or cannot add Man3, the GPI proteins were never found to be attached through an EtN-P on Man1 or Man2.
Whereas at present we have a good overall picture of the major stages of the GPI biosynthesis pathway, many questions concerning the sequence of events and their importance for the biosynthetic process remain to be elucidated. One particular difficulty is to know which GPI lipid is attached to nascent proteins by the GPI transamidase complex in the ER. Even though there is a wealth of partial structures of lipids accumulating in mutants that are unable to synthesize or to attach GPI lipids to proteins, or of lipids that are synthesized by microsomes in vitro, it is doubtful that all these lipids are normal intermediates of the pathway. Thus, it cannot be excluded a priori that under physiological conditions certain EtN-Ps are added to GPI anchors only after the GPI lipid has been added to proteins or the protein has left the ER. Here we describe our attempts to address a few of the several unresolved questions about the biosynthetic route and the role of the EtN-P residues on Man1 and Man2 in yeast.
Testing Enzymes for Contaminating Phosphodiesterase Activity Removing EtN-P Side Chains-Radiolabeled CP lipids were generated by metabolic labeling of pmi40 with [ 3 H]mannose (30) or sec18 gpi8 double mutants with myo- [2-3 H]Ins. CP lipids were purified by preparative TLC and then incubated under exactly the same conditions as used for preparing GPI anchor peptides. In some experiments we added a mixture of the phosphodiesterase inhibitors pentoxifylline, dipyramidole, ethaverine, 3-isobutyl-1-methylxanthine and papaverine at 1, 0.25, 0.1, 0.1, and 0.25 mM final concentrations, respectively. To test for a phosphodiesterase activity that would remove EtN-P, CaCl 2 in GPI-PLD buffer was replaced by 10 mM EDTA so that the GPI-PLD itself was inactive (31). After incubation the potential degradation of CP lipids and the appearance of less polar lipids was assessed by TLC in solvent 1 followed by radioscanning/fluorography.
Analysis of Head Groups-Liberated head groups were subjected to acetolysis, JBAM and HF treatments, N-acetylation using methods listed in Ref. 11, and paper chromatography in solvent methyl ethyl ketone/pyridine/water (20:12:11) (9). myo-[ 14 C]Ins was added to each sample as an internal standard before paper chromatography allowing for exact positioning of Man-GlcNAc-Ins and GlcNAc-Ins peaks.
Plasmids and Plasmid Construction-A pUPRE expressing lacZ from an artificial promoter containing UPRE was obtained from Dr. Ralph Menzel, Berlin (32). For Fig. 10A we used multicopy TOp2141, TOp662, or YEp24 containing PSD1, ECM33, or no insert that all were contributed by Dr. Akio Toh-e (33). pBF111 containing GPI7 with a deletion of amino acids 192-321 was constructed by opening pBF43 (YCplac33 containing GPI7 behind its own promoter) with SmaI and BssHII. The gap was filled with the EcoRV-BssHII fragment excised from the same plasmid, pBF43. pBF112 containing GPI7 with a deletion of amino acids 54 -191 was generated by opening pBF43 with PflMI, blunting with Klenow polymerase, and then cutting with BssHII. The thus generated gap was filled with a SmaI-BssHII fragment from pBF43. pBF113 was constructed by amplifying the N-terminal part of Gpi7p by PCR using primers 5Ј-agacgttcaacaaattgatatcgt-3Ј and 5Ј-cgcgACGCGTACCGGTcaaaagaggataattataatttgt-3Ј having restriction sites MluI and AgeI (uppercase). The last 39 amino acids of Wbp1p (including its transmembrane domain and the KKXX motif), the stop codon plus the transcription terminator were amplified with primers 5Ј-cgcgACGCGTACCGGTtcttgggtttatattagcgccattt-3Ј and 5Ј-cgcg-gaattcgagctcGGTACCccttaatacaaactgcaaaagagttt-3Ј. The two PCR fragments were cut with MluI and EcoRV and MluI and KpnI, respectively, and ligated into pBF43 opened with EcoRV and KpnI. pBF114 was constructed by amplifying nucleotides 409 -1197 of GPI7 with primers 5Ј-tggCTGCAGcagttcatccaacata-3Ј and 5Ј-gGTTTAAACGG-TACCttacaattcatcgtgtgcagacttggttaacgtttcttga-3Ј. The PCR fragment was cut with PstI and KpnI and inserted into pBF43 opened with the same two enzymes. To obtain multicopy plasmids, the XhoI-KpnI fragments of pBF111, pBF112, pBF113, pBF114, and pBF43 were transferred into YEplac195, which was opened with SalI (same overhang as XhoI) and KpnI, thus yielding vectors pBF116, pBF117, pBF118, pBF119, and pBF120, respectively. In pBF121, Gpi7p contains an additional 6 amino acids (RSKKHQ) at its C terminus. pBF121 was obtained by amplifying the C terminus of Gpi7p with primers 5Ј-actt-GCGCGCgttttcttccaa-3Ј (uppercase is the RS of BssHII) and 5Ј-TCTA-TGTTAGTTTGTTTTTTTCGACCTatcaagagcgcaaaggaggg-3Ј and amplifying the transcription terminator of GPI7 with primers 5Ј-AGG-TCGAAAAAAACAAACTAAcatagaattgttcacgtggtctaaa-3Ј (sequence encoding RSKKHQ in uppercase) and 5Ј-agtgagtacatGGTACCaccttattat-3Ј. The two PCR fragments were used as templates for crossed PCR and the final product was cut with BssHII and KpnI and then ligated into pBF43 that was opened in the same way. The multicopy version pBF122 was obtained as described above.

Addition of Ethanolamine Phosphate to Man1
Is a Prerequisite for the Addition of Man3 by Gpi10p-During GPI biosynthesis, Man3 is transferred from dolicholphosphomannose to Man2 by PIG-B in mammals and its homologue GPI10 in yeast (11,34,35). Mammalian and yeast mutants in these genes accumulate Man-(EtN-P3)Man-GlcN-(acyl3)PI. When treated with the fungal inhibitor YW3548, yeast cells accumulate Man-Man-GlcN-(acyl3)PI and YW3548 has therefore been postulated to be an inhibitor of the EtN-P transferase PIG-N/ MCD4 (35,36) than of Gpi10p, as initially proposed. This notion was confirmed by the finding that overexpression of MCD4 made yeast cells resistant to YW3548, whereas overexpression of GPI10 had a much smaller, albeit significant effect, especially at low inhibitor concentrations (16,35). The implication is that Gpi10p cannot add onto Man-Man-GlcN-(acyl3)PI and requires a substrate that contains EtN-P on Man1. This hypothesis, however, was difficult to reconcile with the occurrence of major GPI lipids accumulating in mutants such as smp3 (lipid 3-1-2), gpi11 (lipid 11-2), and gpi13 (part of lipid a) as these lipids contain 3 or 4 mannoses but lack EtN-P on Man1. In this context it was of interest to know the struc-ture of the intermediates accumulating in mcd4 mutants. Several mcd4 mutants had previously been labeled with [ 3 H]inositol but accumulated only very small amounts of Ins-labeled GPI lipids that could not be structurally analyzed (15,19,37). We therefore decided to analyze the in vitro GPI biosynthesis in yeast strains carrying either a temperature-sensitive mutation in MCD4/SSU21 (38) or the wild-type MCD4 gene under control of the GAL1 promoter (19). Microsomes from a wild-type strain make the complete GPI precursor CP2 irrespective of the carbon source on which cells have been grown (Fig. 2, lanes 1 and 2). In contrast, Mcd4p-depleted microsomes accumulate a GPI intermediate termed lipid 4c (Fig. 2, lane 4) but are still able to make CP2, although considerably less than wild type. In microsomes from a temperature-sensitive mcd4, a stronger block of GPI biosynthesis is observed (Fig. 2, lane 5), and neither M4 nor CP2 is made. Lipid 4c is less polar than lipid 031b (Man-Man-Man-(EtN-P3)Man-GlcN-(acyl3)PI), suggesting that it is an earlier intermediate in GPI biosynthesis. Structural characterization of lipid 4c shows that lipid 4c contains a Man 2 -GlcN-Ins core structure and that both mannoses can be removed by treatment with JBAM ( Fig. 2B). A similar workup of the total pool of GPI lipids generated by mcd4 -174 microsomes showed that none of them had more than two mannoses (not shown). This finding is in agreement with the idea that in yeast, differently than in mammals, the addition of EtN-P is a prerequisite for the addition of Man3 as proposed before (16,35). It is still possible that lipids 3-1-2 and 11-2 of smp3 and gpi11 as well as the more polar lipids observed in mcd4-174 are generated by a parallel pathway that does not require EtN-P on Man1 (16,18,39), but it is clear that the latter pathway cannot provide substrates for the GPI transamidase that would allow cells to survive, and that this latter pathway cannot be detected by the standard microsomal in vitro system used here.
Technical Difficulties in Detecting Ethanolamine Phosphate Substituents on the GPI Anchor of Yeast Proteins-The confirmatory findings reported above suggest that the GPI lipids transferred onto proteins by the yeast GPI transamidase do contain EtN-P on Man1 and lead to the expectation, that, as many mammalian proteins, yeast proteins, at least initially, contain EtN-P on Man1. As our previous investigation had failed to detect any substituent on Man1 of protein anchors we decided to reinvestigate this issue. Indeed, in our previous report we prepared anchors in large quantities for analysis by GC-MS, NMR, exoglycosidase sequencing, and fast atom bombardment mass spectrometry on the basis of their biochemical and biophysical properties and without the preliminary purification of any particular GPI protein (14). The methods used previously for purification of GPI anchors might have led to the loss of phosphodiester-linked substituents because some endogenous phosphodiesterases might have survived the initial high pH treatment, pH 11, during glass bead disruption of cells and been active during subsequent membrane preparation steps. Also, the enzymes utilized for the further isolation and purification of the hydrophilic GPI anchor head groups such as Pronase, endoglycosidase H, and PI-PLC might have been contaminated by phosphodiesterases. We thus decided to take a different approach, i.e. to use [ 3 H]Ins-labeled cells as starting material and to follow the purification scheme depicted in Fig.  3. We reasoned that the initial denaturation in organic solvent and the subsequent boiling in SDS ought to rapidly and efficiently inactivate any untoward hydrolases. Moreover, we took care to screen all enzymes as well as ConA-Sepharose for the presence of contaminating phosphodiesterases by incubating purified radiolabeled CP lipids under the same conditions as used for preparing GPI anchor head groups. After incubation, the status of the radiolabeled CP was assessed by TLC and radioscanning/fluorography. As can be seen in Fig. 4, only minor quantities of more hydrophobic lipids running in the region of M4 were generated through 16-h incubations with purified bovine GPI-PLD, and the brands of PI-PLC of B. cereus or Pronase specified under "Experimental Procedures." The appearance of small amounts of material running in the region of M4 could not be prevented by phosphodiesterase inhibitors. When testing for phosphodiesterase activity in ConA-Sepharose we found that affinity chromatography degrades CP lipids and transforms them into a lipid comigrating with M4 as summarized in Table II. It appeared that degradation was not caused by ConA-Sepharose but largely by the buffer. Divalent cations were only partially responsible for the degradation and about 10% of degradation occurred even in their absence. Interestingly, degradation was strongly enhanced by NaCl. As the main degradation product comigrates with M4/1, the lipid that accumulates in gpi7 mutants and contains an EtN-P on Man1 but lacks a substituent on Man2 (12)  Lipids were extracted, desalted, and analyzed by TLC using solvent 1 and fluorography. B, microsomes from strain 521-17A-H42 were labeled with UDP-[ 3 H]GlcNAc and lipid 4c was purified by preparative TLC. The labeled head group of lipid 4c was isolated, treated or not with JBAM, an exomannosidase without any linkage specificity, followed by treatment with HF at 0°C to specifically cleave phosphodiesters and finally N-acetylated (NAc) using published methods (48). The generated fragments were analyzed by paper chromatography along with radiolabeled Man 0 -4 -GlcNAc-Ins standards. Their position is indicated by 0 to 4; I ϭ Ins.
Man2. We also verified that acidic conditions per se do not affect the stability of the EtN-P groups as the incubation of cells in 5% trichloroacetic acid at 0°C for 60 min before lipid extraction did not diminish the amount of CP lipid that could be extracted from [ 3 H]Ins-labeled gpi8 sec18 cells (not shown). Similarly, treatment of CP2 by acetolysis did not remove EtN-P from Man1 as described before (11) and this was verified during this experimental series by carrying purified CP2 through the same workup as the anchor peptides (not shown). Thus, after these preliminary tests we were confident that the enzymes used did not contain any activity transforming CP lipids into less polar compounds but we could not prevent that some substituents, nevertheless, might be lost through spontaneous hydrolysis during head group preparations.
Ethanolamine Phosphate Is Present on Man1 of the Majority of GPI Anchors in WT Cells-To detect the presence of EtN-P on Man1 we prepared anchor peptides as outlined in Fig. 3 and analyzed the resulting anchor GPI head groups by sequentially treating with acetolysis, a procedure that specifically cleaves the ␣1-6 glycosidic bond of the anchor (Fig. 1), then JBAM and finally HF. As shown in Fig. 5, A and B, and Table III, 15% of GPI anchors of WT cells contained an HF-sensitive substituent on Man1. Using the same protocol we tested several mutants as shown in Fig. 5, B-F, and Table III, and found that the percentage of substitution on Man1 was higher in gpi7⌬ and sec18-1, and that these effects were potentiated in sec18-1 gpi7⌬ double mutant cells, where almost all anchors were substituted on Man1. To make sure that the low percentage of Man1 substitution on GPI anchors of WT cells truly represented the status at the end of the metabolic labeling and were not the consequence of spontaneous or catalyzed hydrolysis during the following workup we performed the mixing experiments reported in Table IV. For this, labeled and non-labeled cells of various genotypes were mixed before starting the workup outlined in Fig. 3. These experiments revealed neither trans-acting destabilizing factors in WT nor stabilizing factors in sec18-1 gpi7⌬ cells. In addition, buffering the potential acidity of chloroform solutions with triethylamine or the inclusion of phosphodiesterase inhibitors did not influence the results. Previous studies had brought evidence for Man1-substituted complete precursor (CP) and M4 lipids but in these experi-ments the time of exposure of the lipids to organic solvents was significantly shorter than the time required for delipidation of proteins. We therefore continued to be concerned with the possibility of untoward hydrolysis of phosphodiester-linked side chains in organic solvents and we therefore modified the protocol of Fig. 3 by omitting the first step, i.e. the organic solvent extraction. Indeed, when labeled cells were directly boiled in SDS sample buffer, 52% of Man1 of WT anchors and 98% on sec18-1 gpi7⌬ anchors were found to be substituted ( Fig. 6 and Table V). Yet, by this procedure the yields of labeled anchors were 4 times lower than with the unmodified protocol. However, by either first breaking cells in Tris buffer using glass beads or preincubating cells in 0.1 M sodium hydroxide before boiling in sample buffer the yields were as high as with the original procedure of Fig. 3 and yet, 72% of WT anchors were found to be substituted (Table V). (The contribution of CP and M4 lipids to this improved yield could be excluded, as these lipids do not bind to ConA-Sepharose in the conditions used for anchor peptide preparation.) Thus, it appears that at the end of metabolic labeling, 72% of WT anchors carry a substituent on Man1 that is lost when proteins are delipidated using organic solvents. In contrast, 98% of the GPI anchors of sec18-1 gpi7⌬ carry a substituent on Man1, which seems to be largely resistant to organic solvent.
HF-sensitive Substituents on Man2-We tried to investigate if GPI anchors carried any substituents on Man2 by using limiting HF digestion, a procedure that had previously been used successfully to detect an HF-sensitive substituent on Man2 of [ 3 H]mannose-labeled CP2 of pmi40 cells (12). As can be appreciated from Fig. 7 and from the corresponding quantitation in Table VI, this method identifies a substituent on Man2 of [ 3 H]Ins-labeled CP from gpi8-1 sec18-1 as well, but it detects only minimal amounts of substitution on Man2 of WT of sec12 protein anchors. (Mutants in SEC12 block the exit of secretory proteins out of the ER.) Small amounts of Man 2 -GlcNAc-Ins were also generated from M4 in this experiment (see "Discussion"). However, it appears that limiting HF does not seem to work the same way on anchor peptides as on free GPI lipids in as much as also substituents on Man1 were not detected efficiently. Indeed, with the protocol used, 15% of the anchor peptides of WT and 94% of sec18-1 gpi7⌬ cells carried a After isolation, the head groups were further analyzed by chemical and enzymatic treatments as described in the text.

TABLE II Degradation of CP lipids by concanavalin A-Sepharose buffer
Purified CP lipids were incubated with ConA-Sepharose or the buffer that is used normally for ConA-Sepharose affinity chromatography but without Triton X-100 and lacking some ingredients. Products were extracted, analyzed by TLC, and quantitated. All radioactivity appeared as M4/1, M4/2, or CP and the sum of these three peaks was set as 100%. Experiment 4. Testing enzymes for phosphodiesterase activity that would remove EtN-P. Partially purified CP was incubated for 16 h at 37°C in the presence (ϩ) or absence (Ϫ) of the phospholipases (PL), Pronase (PRO), and a phosphodiesterase inhibitor mixture (PDI) as indicated. After incubation the lipids were extracted into butanol, dried, and separated on TLC with chloroform/methanol/water (10:10:3). Lanes A7 and B8 -B10 contain the partially purified CP used as starting material and lanes A8 and B1 an aliquot of the [ 3 H]Ins-labeled lipid extract of sec18 gpi8, from which CP (*) was purified. A, incubation with GPI-PLD was done in GPI-PLD buffer in the presence of 10 mM EGTA to block GPI-PLD activity. PI-PLC in lanes 4 -6 is from ICN and Roche Applied Science, respectively. Note that PI-PLC cannot act on CP because of the presence of an acyl on the Ins. To have identical migration conditions the compounds not added for incubation were added only before butanol water partitioning. B, Pronase from Streptomyces griseus was either from Sigma (lanes 2-4) or Roche Applied Science ( lanes 5-7); a, addition of inhibitors after incubation. substituent on Man1 (Table III) but only a small fraction could be detected with limiting HF. Curiously, however, in sec18-1 gpi7⌬ the method detected as much substituents on Man2 as on Man1, although we know that M4, the free anchor lipid accumulating in gpi7⌬ cells does not carry any substituent on Man2. As these experiments are rather extensive and the results are difficult to interpret, we did not repeat or do more experiments of this kind, but we note that with this method we detect very little substituents on Man2 in GPI anchors of WT cells. This is in agreement with the above observed lability of a substituent on Man2 of CP (Table II).
Whereas the presence of EtN-P on Man1 and Man2 of mammalian anchors is firmly established, in yeast EtN-P has formally been shown to be present only on Man1 and Man3 (11,13,14). The HF-sensitive substituent on Man2 was surmised to be EtN-P based on the accumulation of M4, a GPI lipid lacking a substituent on Man2, in gpi7⌬ and on the homology of GPI7 with MCD4 and GPI13, which are required to add EtN-P onto Man1 and Man3, respectively (12,15,18,19). However, a recent report has clearly established the presence of a GlcNAc-P residue attached to Man2 in some human GPI anchors (40). To address the question if yeast anchors carry EtN-P on Man2, sec18 gpi8-1 cells were double labeled with [ 3 H]serine and [ 14 C]Ins at 37°C. The lipid extract was separated in two-dimensional TLC, the plate was radioscanned, and the spots corresponding to CP and M4 were scraped and counted in a scintillation ␤-counter with the windows set for separate detection of tritium and carbon 14. In M4 the 3 H: 14 C ratio was 3.0, in CP it was 4.0. As M4 has 2 and CP either 2 or 3 EtN-Ps, we would have expected that the two ratios are either the same or differ by a factor of 1.5. Yet, the labeling lasting 2 h, the specific activity of [ 3

H]serine/[ 3 H]PE and [ 14 C]Ins/[ 14 C]
PI not being constant during that time, we cannot reasonably expect the ratios to differ by a factor of exactly 1.5. The higher 3 H: 14 C ratio in CP suggests that the residue on its Man2 is indeed an EtN-P.
Is CP Made in the ER and Is Gpi7p Essential for Adding the Substituent on Man2?-The foregoing and published data are compatible with the view of Gpi7p being an EtN-P transferase for Man2. Previous studies have indicated that Gpi7p is a glycoprotein located at the plasma membrane and that its N-glycans undergo extensive elongation in the Golgi. A minor core N-glycosylated form disappeared upon incubation of cells in cycloheximide (Chx) and was presumed to represent the immature ER form of Gpi7p in transit to the cell surface (12). To test the working hypothesis that ER-localized Gpi7p in transit is responsible for the transfer of the HF-sensitive substituent on Man2 of CP lipids, Gpi8p-depleted cells were labeled with [ 3 H]Ins after a 40 -70-min preincubation with Chx that ought to allow newly made Gpi7p to be transported out of the ER. This protocol leads to the complete disappearance of CP but also of the less hydrophilic GPI lipids M4/1 and M4/2 (Fig.  8, lanes 7 and 8), indicating that Chx was blocking GPI biosynthesis at an earlier stage. M4, the abnormal GPI lipid accumulating in gpi7⌬ was also drastically reduced when Chx was used (Fig. 8A, lanes 9 and 10). The same phenomenon was also observed in gpi8-1 sec18-1 cells labeled at 37°C, at which temperature all vesicular traffic is blocked. These cells made significantly more CP than gpi8-1 mutants (not shown) but Chx preincubation blocked the biosynthesis of mature GPI lipids (Fig. 8A, lanes 1 and 5). Incidentally, this result demonstrates that CP biosynthesis does not require any vesicular transport of GPI lipids out of the ER. Cycloheximide blocked GPI biosynthesis also when just added before [ 3 H]Ins (Fig. 8, A,  lanes 3 and 4, and B, lane 8). Cycloheximide, however, did not induce their breakdown; as shown in Fig. 8B, M4/1, M4/2, and CP lipids, which had been accumulating over 60 min in the absence of Chx, remained entirely stable during a subsequent period of 20 min in the presence of Chx. Thus, although, as judged by Western blotting, the core-glycosylated form of Gpi7p could efficiently be accumulated by preincubation of gpi8-1 sec18-1 at 37°C (not shown), cells were, nevertheless, unable to make mature GPI lipids in the presence of Chx. As little as 1 g/ml Chx was sufficient to produce this effect. As shown in Fig. 9A, Chx had no effect on GPI biosynthesis when added to microsomes, raising the possibility that it was blocking the biosynthesis of glucosamine or UDP-GlcNAc. However, the addition of glucosamine (up to 100 mM) to sec18 gpi8 while they were labeled with [ 3 H]Ins in vivo did not allow cells to make M4 or CP lipids in the presence of Chx (not shown). The arrest of GPI biosynthesis caused by Chx was surprising as we previously had found that Chx stimulated the incorporation of [ 3 H]mannose into CP and M4 lipids in pmi40, a temperaturesensitive mutant unable to make mannose at 37°C (30,41). We excluded a potential effect of Chx on mannose biosynthesis as the addition of mannose to the medium during [ 3 H]Ins labeling did not allow GPI biosynthesis in the presence of Chx (not shown). One possible explanation for the discrepancy would be that the amounts of CP present in gpi8 sec18 are much higher than in pmi40. Indeed, the comparison of the ratio of were labeled at 37°C and head groups were isolated, subjected to acetolysis, and then treated with JBAM. The resulting products were dephosphorylated with HF, N-acetylated, and desalted before being analyzed by paper chromatography. In panel B, JBAM treatment was omitted. Note that acetolysis conditions are such that only part of the anchors are cleaved to preserve glycosidic bonds that are not ␣1,6. Quantitation of the data and of control experiments are reported in Table III in vivo and we therefore could not assess the role of the ER form of Gpi7p for GPI biosynthesis in living cells.
As vesicular traffic is not required for making CP (Fig. 8A,  lane 5), it seemed impossible that surface-located Gpi7p contributed to the biosynthesis of CP2 unless the GPI could reach the surface by a non-vesicular pathway. Indeed, earlier studies had shown that mature GPI lipids were accessible at the cytosolic leaflet of microsomes (42). In view of more recent data showing that the mannosyl and EtN-P transferases involved in GPI biosynthesis had their catalytic site on the lumenal side of the ER (34,43) these earlier data may be taken as evidence that lumenally synthesized GPI lipids can flip and are in equilibrium over the two leaflets of the ER membrane. It thus appeared possible that M4 made in the ER would reach the plasma membrane through a cytosolic route and be transformed into CP by Gpi7p residing at the cell surface. To put this hypothesis to the test we asked whether the removal of Gpi7p from spheroplasts used for preparation of microsomes would FIG. 6. Analysis of protein head groups isolated avoiding organic solvents. X2180-1A and FBY49 cells were labeled at 37°C for 1 h with [ 3 H]Ins, cells were directly boiled in sample buffer and then further processed as outlined in Fig.  3. Head groups were subjected to acetolysis, subjected or not to JBAM, products were dephosphorylated with HF, N-acetylated, and desalted prior to analysis by paper chromatography.

TABLE III
Quantitation of data in Fig 5 Numbers indicate the amount of Man-GlcNAc-Ins as percentage of Man-GlcNAc-Ins plus GlcNAc-Ins after various treatments as calculated from paper chromatography profiles (1/1 ϩ 0 in Fig. 5). As sometimes small amounts of GlcNAc-Ins were present already after acetolysis/HF/NAc, the percentage of EtN-P-substituted Man1 was obtained by dividing the % of Man-GlcNAc-Ins after acetolysis/JBAM/HF/NAc by the % of Man-GlcNAc-Ins after acetolysis/HF/NAc. Profiles shown in Fig. 5 are in italics. The control in the third column demonstrates that all labeled head groups were completely sensitive to JBAM after HF treatment.  FBY49 (sec1-8 ⌬gpi7) cells were labeled for 60 min at 37°C (labeled cells are indicated by an asterisk). After the labeling 5 A 600 of labeled cells were mixed with 5 A 600 of unlabeled cells and processed as outlined in Fig. 3. Cells were broken in chloroform/methanol, 1:1, extensively delipidated with chloroform/methanol/water, 10:10:3, then Lester solvent or alternatively with chloroform, methanol, 1.5 mM triethylamine, pH 7.5 (10:10:3)(ϩTEA). After solubilization of proteins in sample buffer further incubations were in some cases done in the presence of a phosphodiesterase inhibitor mixture (ϩ PDI). Head groups were subjected to acetolysis and then either treated with JBAM (acetolysis/JBAM/ HF/NAc) or left untreated (acetolysis/HF/NAc). All products were dephosphorylated with HF, N-acetylated and desalted before being analyzed by paper chromatography. Numbers indicate the Man-GlcNAc-Ins as percentage of Man-GlcNAc-Ins plus GlcNAc-Ins after various treatments as calculated from paper chromatography profiles. affect the biosynthesis of CP lipids in microsomes. We exploited the fact that Zymolyase consistently removes the high molecular weight form of Gpi7p from cells, whereas Quantazyme, a recombinant ␤1,3-glucanase leaves Gpi7p intact (12). Indeed, microsomes prepared from Zymolyase-treated spheroplasts made less CP than the ones made from Quantazyme-treated spheroplasts (Fig. 9C, lanes 2 and 4). Although the inhibition of GPI biosynthesis by Chx is not clearly understood, we tried to ask if preincubating living cells with Chx would affect a sub-sequent microsomal GPI biosynthesis assay. Experiments to address this question are shown in Fig. 9C, and are quantitated in Table VII. It clearly appears that microsomes prepared from cells preincubated with Chx incorporate less [ 3 H]GlcNAc into CP and the same applies to M4 lipids. However, even the combination of Chx preincubation and the use of Zymolyase did not result in a complete suppression of GPI biosynthesis (lane 5). Thus, whereas Zymolyase and Chx are efficient in eliminating the high molecular weight and core-glycosylated form of Gpi7p (data not shown), their combined action did not completely abrogate the biosynthesis of CP. We tried to get independent evidence for transfer of GPI lipids between microsomes during the in vitro GPI biosynthesis assay by mixing microsomes from gpi mutants blocked at various stages of the biosynthetic pathway. As shown in Fig. 9B, there was no biochemical complementation of gpi7⌬ microsomes by microsomes from gpi1 or gpi10 cells in that these mixtures did not produce any CP and this was true even though concentrated cytosol (a source of possible lipid transfer proteins) had been added to the standard assays. In summary these studies indicate that the microsomal GPI biosynthesis assay generates CP2 in the absence of any vesicular fusion or transfer of GPI intermediates from one vesicle to the other suggesting the possibility that the transferase that adds a substituent to Man2 in microsomes is not Gpi7p.
Effects of Overexpression of PSD1, PSD2, and ECM33 in gpi7⌬-Previous studies showed that certain mcd4 alleles render cells auxotrophic for ethanolamine if they are placed into a psd1⌬ background, and other studies showed that overexpression of PSD1 can rescue the growth phenotype of a temperature-sensitive mcd4 allele (fsr2-1) (33,37). PSD1 is the mitochondrial phosphatidylserine decarboxylase that generates the bulk of phosphatidylethanolamine, the lipid that is the immediate donor of EtN-P for Mcd4p and Gpi13p (13,20). The above findings thus suggested that PE may be limiting for the lumenal EtN-P transferase Mcd4p in certain conditions and that overexpression of PSD1 can increase lumenal PE levels. Curiously, overexpression of PSD1 (or ECM33) also restored normal growth at 37°C to gpi7⌬ cells, again raising the possibility that some protein other than Gpi7p may add EtN-P to GPI lipids if the PE levels are high enough (33). The experiment shown in Fig. 10A indicates that the overexpression of PSD1 or ECM33 does not decrease, but rather increases the accumulation of M4 in gpi7⌬ cells and that it does not lead to the accumulation of CP lipids in a thermosensitive gpi7 gpi8 double mutant, a mutant, which accumulates CP when complemented with GPI7 (12). Similarly, the overexpression of PSD2, a phosphatidylserine decarboxylase of the secretory apparatus, does not abolish the accumulation of M4 when overexpressed in gpi7⌬ that simultaneously overexpress MCD4 or ECM33 (not shown). Thus, overexpression of PSD1 or ECM33 seem to improve the growth of gpi7⌬ cells not by allowing for CP biosynthesis but rather by increasing M4 biosynthesis sug- X2180-1A (wt) and SF226-1C (sec12-4) cells were labeled with [ 3 H]Ins at 37°C and anchor peptides were prepared as outlined in Fig. 3. The gpi7⌬ and gpi8-1 sec18-1 cells were labeled with [ 3 H]Ins at 37°C to obtain radiolabeled GPI lipids M4 and CP, respectively. These lipids were obtained by two consecutive runs of preparative TLC. Head groups were treated for 12 h with HF (limiting HF treatment), desalted, treated with JBAM, retreated with HF for 48 h, N-acetylated, desalted, and finally analyzed by paper chromatography.   Fig. 7  FBY49 (sec18 -1 gpi7⌬) cells were labeled and processed along with other cells shown in Fig. 7. Boxheads indicate the various species of Man x -GlcNAc-Ins containing from 0 to 4 mannoses. Numbers indicate the fraction of Man x -GlcNAc-Ins as a percentage of all peaks. Values obtained from the profile of Fig. 7  We incidentally also tested if overexpression of Gpi7p would rescue either mcd4⌬ or gpi13⌬ mutants. In mcd4⌬ or gpi13⌬, the endogenous Gpi7p may fail to add EtN-P onto Man1 or Man3 not because of its inappropriate specificity but because of its inappropriate location outside the ER. We had noted before that the overexpression of GPI7 results in the accumulation of Gpi7p in a core-glycosylated form (51). Yet, as shown in Fig. 10B, overexpression of GPI7 did not improve the growth of Mcd4p-or Gpi13p-depleted cells.
Identification of Functional Domains in Gpi7p-Our previous study (12) on Gpi7p had revealed that the protein, apart from EtN-P transfer, was required for a second enzymatic activity, the remodeling (exchange) of the lipid moiety of GPI anchors in the late compartments of the secretory pathway. The structure of Gpi7p contains a 400-amino acid globular ectocytoplasmic domain followed by a 430-amino acid long hydrophobic domain containing 8 -11 membrane spanning regions (Fig. 11A). The Gpi7p constructs shown in Fig. 11A were introduced into gpi7⌬ cells to test their capacity to abolish the abnormal accumulation of the GPI lipid M4 and to restore GPI anchor remodeling. Western blotting with anti-Gpi7p antibody demonstrated that all constructs were expressed and that the products of pBF118 and pBF119 were only core glycosylated, suggesting that they are efficiently retained in the ER, whereas those of pBF121 and pBF122 were of high molecular weight indicating that the KKXX motif added to the C terminus of Gpi7p was lumenal or otherwise non-functional (Fig. 11B). M4 accumulation was suppressed by pBF43, pBF120, pBF121, and pBF122, whereas all other constructs, even when overexpressed, were inefficient in this respect (not shown). Similarly, only WT Gpi7p forms could restore remodeling of GPI proteins and calcofluor resistance (not shown). In summary, we could not dissect the protein into functional domains simply by making the large deletions indicated in Fig. 11A. DISCUSSION The finding that the addition of EtN-P to Man1 of yeast GPI lipids is an obligatory step of GPI biosynthesis in yeast led us to reinvestigate the structure of the yeast GPI anchor and led us to the conclusion that indeed, most GPI anchors contain an HF-sensitive substituent on Man1. The study also led us to realize some intricacies that render the study of EtN-P substituents tricky and may explain the substoichiometric amounts of EtN-P substituents found in GPI anchors of other organisms. Indeed, the substituent on Man1 is relatively labile during lengthy delipidation of GPI proteins in organic solvents, but it somehow gets resistant to organic solvents in gpi7⌬ sec18 mutants. The effect of the gpi7⌬ sec18 double mutation is discernible even if organic solvents were avoided, in which case 98% of anchors had a substitution on Man1 in gpi7⌬ sec18 but only 72% in WT (Table V). There is no obvious explanation for this finding but several hypothesis can be formulated: 1) if there is no substituent on Man2 the stability of the EtN-P on Man1 may increase because of altered electrostatic interactions. Alternatively, a spontaneous transesterification reaction that translocates the EtN-P from the C atom it was attached to by Mcd4p to another C atom may be enhanced or prevented by the presence of a substituent on Man2. 2) Deletion of GPI7 induces the cell wall integrity pathway (46), and the UPR. These responses may induce enzymes that change the configuration of the substituent on Man1.
3) The small amounts of Gpi7p present in the ER may not, or not only be important for the addition of EtN-P on Man2 but may also hydrolyze or transesterify the EtN-P residue on Man1. Absence of Gpi7p thus could stabilize the EtN-P on Man1. It must be said that it is unclear if the substituent is primarily stable and loses stability in the context of WT cells, or it is primarily unstable and gains stability in the context of gpi7⌬ sec18 cells. It also is surprising that the substituent is only partially stabilized in gpi7⌬ single mutants indicating that the environment inside and outside the ER influences the substituent on Man1 differently. Clearly, to resolve this puzzle, it will be necessary to compare the site of attachment of the Man1 substituent on anchors of WT and gpi7⌬ sec18 mutants. (The short exposure to organic solvents required for lipid extraction leaves the Man1 substituents of CP intact but it could well be that part of CP initially present after labeling looses its substituent and is transformed into M4/2 (Fig. 8A) during extraction.) The substituent on Man2 most likely is an EtN-P, but it is at present not clear if this substituent is stably associated with FIG. 9. Influence of Zymolyase and cycloheximide preincubation on biosynthesis of CP in vitro. Cells were transformed into spheroplasts using either Zymolyase (Z) or Quantazyme (Q), microsomes were prepared and incubated with UDP-[ 3 H]GlcNAc, and lipid extracts were analyzed by TLC/fluorography as described under "Experimental Procedures." The lipid extract of [ 3 H]Ins-labeled WT cells is shown in lanes 1 and 6. A, the assay was carried out either in the presence or absence of Chx (100 g/ml). Lane 4 contains [ 3 H]Ins. B, WT, gpi1, gpi10, and gpi7 cells were treated with Quantazyme, microsomes were incubated in the presence of cytosol either individually (lanes [2][3][4][5] or by mixing microsomes from two strains as indicated (lanes 8 -10). C, cells were preincubated in the presence or absence of Chx (100 g/ml) (Chx pre.) for 20 min at 30°C. For Chx-treated cells, Chx was also present during spheroplasting and subsequent washes although the spheroplasting buffer in itself is inhibitory for protein synthesis (azide, absence of glucose).  FBY115 (gpi7⌬, lanes 1-3), YAT2626 (gpi7⌬, lanes 4 -6), and FBY122 (gpi7 gpi8, lanes [7][8][9][10][11][12] were transformed with multicopy vectors containing PSD1 (P), ECM33 (E), or nothing (⌽) and were labeled at the indicated temperature with [ 3 H]Ins. Lipids were extracted and analyzed by TLC/fluorography. The relative percentage of M4 lipid as compared with the total of radioactivity per lane is given at the bottom of each lane. B, 10-fold dilutions of FBY1104 (mcd4::HIS3 GAL1 UAS -MCD4), FBY1102 (gpi13::HIS3-GAL1 UAS -GPI13), and FBY15 (gpi7-1) harboring MCD4, GPI13, and GPI7 behind their own promoter on a URA3-based 2 multicopy vector or harboring empty vector (⌽) were spotted on plates containing galactose or glucose and calcofluor white as indicated on the top of each column and were incubated at 30°C for 3 days.

TABLE VII
Influence of Zymolyase and cycloheximide (Chx) preincubation on biosynthesis of GPI lipids in vitro Radioscanning of TLC plates allowed to quantitate the relative amount of CP and M4 lipids as % of total radioactivity in a given lane for the experiment shown in Fig. 9, A and C (here experiments 1 and 3, respectively) and a further identical experiment. Values obtained by scanning the fluorograms of Fig. 9 are in italics. the anchors in the ER because it is not very stable in aqueous saline on CP lipids (Table VI). Yet, it cannot be excluded that under physiological conditions the substituent is stable but that the organic solvent used for delipidation of proteins changes the configuration of the substituent on Man2. It also has to be taken into account that if a GPI anchor spontaneously looses a substituent on Man2 in the ER, this also may secondarily affect the stability of the EtN-P on Man1 (see above). As Gpi7p is a surface protein, it is difficult to conceive that it is the EtN-P transferase that adds the EtN-P onto Man2, at least of the lipids that are made by WT microsomes in vitro. Indeed, several experimental results seem to raise the possibility that Gpi7p is not directly involved in the addition of EtN-P to Man2, but has an indirect role. 1) Occasionally we found small amounts of substituents on Man2 of M4 using limiting HF (Fig. 7), although this may also stem from inadvertent hydrolysis of glycosidic bonds during HF dephosphorylation. 2) Protein anchors of gpi7⌬ sec18 cells do not contain less but more substituents on Man2 than anchors from sec12 cells or WT cells (Table VI). 3) CP lipid was made in vitro by microsomes, derived from cells that had lost Gpi7p through preincubation with Chx and lyticase treatment (Fig. 9C, lane 5, and Table VII). 4) The head group of lipid 031b, the major lipid accumulating in Gpi13p-depleted cells was shown to contain a Man 4 -GlcN-Ins core with a single EtN-P on either Man1 (80%) or Man2 (20%) (19). Analysis of Gpi13p-depleted gpi7⌬ cells shows that they similarly accumulate 031b, and its preliminary characterization revealed 20% of it to be substituted on Man2 (not shown). 5) Close GPI7 homologues are only found in fungi, although certain mammalian GPI anchors reportedly also contain an EtN-P substituent on Man2 (47). 6) A further hint in this direction comes from the fact that overexpression of PSD1 suppresses the thermosensitivity of gpi7⌬ as well as of certain mcd4 and gpi13 mutants (33), although, as shown here, this effect is not sufficient to abolish the accumulation of M4 in gpi7⌬.

Preincubation of cells with Chx
If Gpi7p were not directly involved in the addition of the HF-sensitive substituent of Man2, another gene would have to contain this activity. Gpi11p is a candidate as it is required for the addition of this substituent, but it may not be the active component of the transferase as it does not have a phosphodiesterase motif and, based on studies in mammalian cells, is believed to be required for stabilization of the EtN-P transferases (18). We suspected that Mcd4p and/or Gpi13p might serve this function in the absence of Gpi7p. However, overexpression of these genes improved the growth of gpi7 mutants on calcofluor white only very little. Thus, whatever the identity of the EtN-P transferase for Man2, its activity remains dependent on GPI7. The role of Gpi13p, Gpi7p, and Mcd4p in the transfer of EtN-P to GPI anchors will have to be further studied by in vitro activity tests with purified enzymes and substrates to unambiguously establish their exact function in this process.

FIG. 11. Dissection of Gpi7p into partially functional domains.
A, the indicated constructs were constructed and expressed in gpi7⌬ mutants (FBY182). A plus sign (ϩ) indicates the position of the conserved phosphodiesterase/nucleotidepyrophosphatase motif, whereas a asterisk (*) indicates the two motifs of the globular domain that are generally conserved in the MCD4/GPI7/GPI13 family (12,47). The conserved Pfam-B_2878 domain extends from position 77 to 276. Deletions are in black, the hydrophobic C terminus is hatched. B, cells expressing the various constructs were extracted with sample buffer containing 20 mM EDTA and probed by Western blotting using affinity purified anti-Gpi7p (12).