Requirement of PIG-F and PIG-O for Transferring Phosphoethanolamine to the Third Mannose in Glycosylphosphatidylinositol*

Many eukaryotic proteins are anchored by glycosylphosphatidylinositol (GPI) to the cell surface membrane. The GPI anchor is linked to proteins by an amide bond formed between the carboxyl terminus and phosphoethanolamine attached to the third mannose. Here, we report the roles of two mammalian genes involved in transfer of phosphoethanolamine to the third mannose in GPI. We cloned a mouse gene termed Pig-o that encodes a 1101-amino acid PIG-O protein bearing regions conserved in various phosphodiesterases.Pig-o knockout F9 embryonal carcinoma cells expressed very little GPI-anchored proteins and accumulated the same major GPI intermediate as the mouse class F mutant cell, which is defective in transferring phosphoethanolamine to the third mannose due to mutantPig-f gene. PIG-O and PIG-F proteins associate with each other, and the stability of PIG-O was dependent upon PIG-F. However, the class F cell is completely deficient in the surface expression of GPI-anchored proteins. A minor GPI intermediate seen inPig-o knockout but not class F cells had more than three mannoses with phosphoethanolamines on the first and third mannoses, suggesting that this GPI may account for the low expression of GPI-anchored proteins. Therefore, mammalian cells have redundant activities in transferring phosphoethanolamine to the third mannose, both of which require PIG-F.

In yeast, two homologous gene products involved in side chain modification of mannose residues were characterized. Gpi7p is involved in the addition of a side chain, probably EtNP, to Man2, which is not essential for growth (12). Mcd4p, which is essential for growth, is probably involved in EtNP transfer to Man1 (15,16). Mouse F9 cells defective in Pig-n, a mouse homologue of MCD4, accumulate GPI precursors without EtNP on Man1 (16). YW3548 (17) inhibits the addition of EtNP to Man1 in both yeast and mammalian cells by inhibiting Mcd4p and Pig-n (8,16), indicating that yeast Mcd4p is involved in transferring EtNP to Man1. Mcd4p and Gpi7p are large proteins of about 120 kDa having an amino-terminal lumenal domain and multiple transmembrane domains in the carboxyl-terminal portion. They have regions conserved in phosphodiesterases and nucleotide pyrophosphatases within the amino-terminal lumenal domain (12,15). It is likely that they are involved in transfer of EtNP to mannoses, although their enzyme activities remain to be demonstrated.
In the yeast genome, there is a third gene, YLL031c, homologous to MCD4 and GPI7 (12,15). Although YLL031c has not been shown to be involved in GPI biosynthesis, its disruption in yeast caused a phenotype similar to those of other GPI anchoring mutants; namely, germinated spores had lethal growth defect after one to two generations with heterogeneous bud sizes (18). Therefore, it is likely that YLL031cp is involved in EtNP transfer to Man3. Man1 is modified at position 2, whereas Man2 and Man3 are modified at position 6. Consistent with these different positions of modification, Gpi7p and YLL031cp share an amino acid identity (38%) that is higher than that of Gpi7p and Mcd4p (24%) and that of YLL031cp and Mcd4p (21%) (12,16).
Mammalian GPI biosynthesis mutant cells of complementation class F are defective in EtNP transfer to Man3 (10,19,20), for which mutation in Pig-f is responsible (21,22). However, PIG-F protein does not have structural similarity to YLL031cp, and yeast has a PIG-F homologue, YDR302wp. Here, we report molecular cloning of a mouse YLL031c homologue termed Pig-o (phosphatidylinositol glycan, class O) and show that PIG-O and PIG-F are involved in the addition of EtNP to Man3. We also show that mammalian cells have another mechanism of adding EtNP to Man3 that requires PIG-F but not PIG-O.
Preparation of cDNA and Genomic Clones of Pig-o-We searched the GenBank TM data base using the BLAST software (26) for sequences homologous to yeast YLL031c and found a mouse genomic sequence. We designed two primers, 5Ј-ATAGACGCTCTGCGGTTTGACTTTGCC (forward) and 5Ј-CATTATCTTCCACTATAGCATGGCTGG (reverse), on the basis of the genomic sequence and amplified its cDNA from a mouse testis cDNA library (a gift from Dr. H. Nojima, Osaka University). The sequence of the PCR product coincided with the corresponding genomic sequences. Using a forward primer designed in a vector and the reverse primer above, fragments containing the 5Ј region were amplified and cloned from the cDNA library. The longest clone was 1.2 kb long. Using the forward primer above and a reverse vector primer, a 3.3-kb fragment containing the 3Ј region was cloned. The longest 5Ј and the 3Ј fragments were ligated to generate a full-length cDNA. Exon positions in the genomic sequence were determined.
Disruption of Pig-o Gene in F9 Cells-A 2.0-kb NotI-XbaI genomic fragment containing the 5Ј region flanking exon 1 was amplified by PCR with forward primer 5Ј-TGCAGCGCGGCCGCTCACTGTGGCTGAG-FIG. 1. A, alignment of the amino acid sequences of PIG-O and YLL031cp. Identical and similar amino acids are shaded black and gray, respectively. Alignment was done using the CLUSTAL W program (40). Regions conserved in many phosphodiesterases and nucleotide pyrophosphatases are underlined. B, hydropathy profile of PIG-O drawn by the method of Kyte and Doolittle (41).
CAGTATGTCAGCG and reverse primer 5Ј-TCCCAATTGACAGAGTT-TGCTTAGCATTCGC and cloned into pPNT (27). (The resulting plasmid was named pPNT-B.) A 6.9-kb SalI-EcoRI genomic fragment containing exons 4 -10 was amplified by PCR with forward primer 5Ј-TTTGTCGACAGGGTTTCATTATGTAGCTTTGGCTGACCAGG and reverse primer 5Ј-GTGCTCTTACTGAAGATTCATCTCTCCGGC. This fragment and a 1.9-kb XbaI-SalI fragment of PGKneo or a 1.8-kb XbaI-SalI fragment of PGKpuro were ligated together in the XbaI-EcoRI site in pPNT-B. The resulting plasmids were used as the first and second targeting vectors to delete exons 1-3. The gene targeting strategy was followed as described (28). Briefly, F9 cells (2 ϫ 10 7 ) were electroporated with 50 g of NotI-cut targeting vector at 230 V and 500 F using a GenePulser (Bio-Rad) and seeded on plates coated with 0.1% gelatin. Positive selection was started with 380 g/ml G418 or 2 g/ml puromycin on the next day. Three days later, negative selection with 2 M gancyclovir was performed. Eight to 11 days after transfection, we screened colonies by PCR for recombinants, in which PCR primers, one in the PGKneo or PGKpuro region and another outside of the genomic region used in vector construction, were used. Positive colonies were confirmed by Southern blot analysis.
FACS Analysis-To examine the surface expression of GPI-anchored proteins, cells were stained with biotinylated-anti-Thy-1 G7 antibody (29), followed by phycoerythrin-conjugated streptavidin (Biomedica, Foster City, CA). They were analyzed in a FACScan (Becton-Dickinson, Franklin Lakes, NJ). Biotinylated monoclonal rat IgG2c antibody with nonrelevant specificity was used as an isotype-matched control.
Subcellular Localization of Pig-o-CHO cells (2 ϫ 10 7 ) were transfected with 4 g of pMEori-FLAG-PIG-F (22) and 21 g of pMEEB-GST-Pig-o and cultured for 2 days. They were collected and suspended in 3 ml of buffer containing 0.25 M sucrose, 10 mM HEPES/NaOH (pH 7.4), 1 mM dithiothreitol, 0.1 mM TLCK, and 1 g/ml leupeptin; disrupted by homogenization; and treated with 1 g/ml DNase I for 20 min on ice. After centrifugation at 10,000 ϫ g for 15 min, the postnuclear supernatants were fractionated by discontinuous sucrose density gradient centrifugation as described (16,30). Membranes were collected from fractions by centrifugation at 100,000 ϫ g for 2 h and dissolved in 1% Nonidet P-40 -150 mM NaCl. GST-tagged Pig-o and FLAG-tagged PIG-F were affinity-precipitated with glutathione beads (Amersham Pharmacia Biotech) and anti-FLAG antibody M2 beads (Sigma), respectively, and analyzed by SDS-polyacrylamide gel electrophoresis and Western blotting. Western blotting was carried out with biotinylated anti-FLAG monoclonal antibody M2 plus horseradish peroxidase-conjugated streptavidin (Amersham Pharmacia Biotech) or with anti-GST antibody (Amersham Pharmacia Biotech) plus HRP-conjugated antigoat IgG antibody (Organon Teknika, N.V.) and visualized with chemiluminescence (Renaissance, DuPont) (31). The band intensities were quantitated by measuring chemiluminescence with a Fuji Image analyzer LAS1000 (Fuji Film Co., Tokyo, Japan) (28). Fractions were also characterized by assaying membrane marker enzymes, alkaline phosphodiesterase I for the plasma membrane, ␣-mannosidase II for the Golgi, and dolichol phosphate mannose synthase for the ER, as described (32).

FIG. 2. Disruption of Pig-o.
A, targeting vectors for homologous recombinations. The first and second targeting vectors were designed to disrupt exons 1-3 and to contain neomycin (neo) and puromycin (puro) resistance genes, respectively. Herpes simplex virus thymidine kinase gene (tk) was included at the 3Ј end to select against random integration. All selection marker genes were driven by PGK promoter and had PGK poly(A) signals. Exons are numbered. B, the structures of intact (middle) and disrupted (top and bottom) Pig-o genes. To screen homologous recombinations, PCR primers were used as indicated (arrows). The forward primer was outside the targeting vectors. The genomic structure of Pig-o was obtained from mouse genomic sequence in the GenBank TM data base. The start codon is located in exon 1. in 50 mM HEPES/NaOH (pH 7.4) containing 0.1 mM TLCK and 1 g/ml leupeptin for 20 min on ice and further disrupted by homogenization. Membranes were obtained by centrifugation at 100,000 ϫ g for 1 h. The collected membranes were solubilized in 1% digitonin in 50 mM HEPES/ NaOH (pH 7.4), 25 mM KCl, 5 mM MgCl 2 , 0.1 mM TLCK, and 1 g/ml leupeptin. Insoluble materials were removed by centrifugation at 100,000 ϫ g for 1 h, and supernatants were used for affinity precipitation. Precipitated proteins were separated by SDS-polyacrylamide gel electrophoresis and analyzed by Western blotting as described above. In Vivo Mannose Labeling and Characterization of Mannolipids-Cells (2 ϫ 10 6 ) were briefly washed twice with glucose-free medium and incubated for 1 h in 1 ml of a medium containing 20 mM HEPES/NaOH (pH 7.4), 100 g/ml glucose, 10% dialyzed fetal calf serum, and 10 g/ml tunicamycin. Then, 40 Ci/ml [ 3 H]mannose (Amersham Pharmacia Biotech) was added, and the cells were further incubated in the same medium for 45 min. After five washes with phosphate-buffered saline, lipids were extracted from pelleted cells with chloroform/methanol/ water (10:10:3) and partitioned into n-butanol (10).

Stabilities of PIG-O in Class F Cells and of PIG-F in Pig-o Knockout
Mannolipids were separated by TLC on Kiesel gel 60 (Merck, Germany) with a solvent system of chloroform/methanol/water (10:10:3) and detected by a Fuji Image analyzer BAS1500 (Fuji Film Co.). In some cases, the glycolipids were treated with GPI-specific phospholipase D (GPI-phospholipase D) (34) and jack bean ␣-mannosidase (Sigma) before TLC analysis. To study the effect of YW3548/BE49385A on GPI biosynthesis in mammalian cells, cells were incubated for 3 days in a medium containing 10 M YW3548/BE49385A (16) and subjected to mannose labeling and lipid extraction, as described above.
Purification of Mannose-labeled Lipids-Cells (1.6 ϫ 10 7 ) were labeled with [ 3 H]mannose, extracted, and separated by TLC as described above. Silica gel particles in active spots were scratched from the TLC plate and collected into tubes containing 400 l of a solvent consisting of isopropanol/hexane/water (50: 25: 20). Tubes were sealed and sonicated in an ultrasonic bath for 10 s and centrifuged at 12,000 ϫ g for 10 min. Supernatants were decanted to new tubes. Extraction was re-peated against the remaining pellets once more. A total of 800 l of solution was dried up in vacuum and used as purified lipid.

RESULTS
Cloning of Pig-o-We found in the GenBank TM data base human and mouse DNA sequences (accession numbers AC004472 and AC005259, respectively) that have homology to yeast YLL031c. We then amplified and cloned a corresponding mouse cDNA from a testis library. We obtained a 4.2-kb cDNA that encodes 1101 amino acids (Fig. 1A) and named the gene Pig-o (for phosphatidylinositol glycan-class O) (DDBJ/Gen-Bank TM /EMBL accession number AB038560). Both Pig-o and its human orthologue PIG-O consisted of 10 exons, the latter being localized on chromosome 9p13 (35). PIG-O had 27% amino acid identity to YLL031cp (Fig. 1A) and several regions highly conserved in two other homologues, Gpi7p and Mcd4p (12,15). PIG-O had two potential N-linked glycosylation sites (Asn-45 and Asn-276) and regions conserved in various phosphodiesterases and nucleotide pyrophosphatases (Fig. 1A) (12,15). There was one predicted transmembrane domain near the amino terminus (amino acids 12-34) that was followed by a hydrophilic portion of about 400 amino acids (Fig. 1B). Within the carboxyl-terminal portion, there were 15 transmembrane domains, as predicted by analyses with the SOSUI (36) and TMHMM (37) programs. Therefore, it is predicted that PIG-O, like Mcd4p and Gpi7p, is a membrane protein with a large hydrophilic domain near the amino terminus and multiple transmembrane domains in the carboxyl-terminal portion.
Disruption of Pig-o in F9 Cells-To see whether Pig-o is involved in GPI anchor biosynthesis, we disrupted Pig-o with homologous recombination in F9 embryonal carcinoma cells. We used targeting vectors in which about a 1.3-kb 5Ј flanking region and exons 1-3 were replaced by neomycin or puromycin resistance gene (Fig. 2A). A region encoded by exons 1-3 that corresponds to the amino-terminal 25% of the protein contains regions conserved in YLL031cp and various phosphodiesterases and nucleotide pyrophosphatases (Fig. 1A). After the first and second homologous recombinations, genomic DNAs were Southern blotted against a cDNA probe containing exons 1-6 (Fig. 2C). The probe should recognize a 9.3-kb fragment before disruption and 7.4-and 7.3-kb fragments after the first and second disruptions, respectively (Fig. 2B). As expected, the 9.3-kb fragment disappeared after the second disruption with the appearance of 7.4/7.3-kb fragments, indicating that Pig-o was completely knocked out (Fig. 2C). To confirm that there was no contamination by nontargeted cells in double disruptant cells, we amplified exon 1 and exon 6 by PCR from genomic DNA (Fig. 2D). Exon 1 was amplified from DNA of wild-type and single disruptant cells (Fig. 2D, lanes 1 and 2) but not double disruptant cells (lane 3), whereas exon 6 was amplified from all of these cells (lanes 5-7). Taken together, the results show that both alleles of Pig-o were successfully disrupted.
Pig-o Is Involved in but Not Essential for GPI Anchor Biosynthesis-We investigated the GPI-anchored protein expression on the surface of Pig-o knockout cells. The expression of Thy-1 was greatly decreased on double disruptant cells (Fig.  3A, middle panel) compared with wild-type cells (left panel). The remaining expression of Thy-1 was significant because disruption of Gaa1, a gene essential for attachment of GPI to protein, completely eliminated the surface Thy-1 expression (Fig. 3A, right panel) (23). The partial decrease was also seen in the expression of ScaI, another mouse GPI-anchored protein (data not shown). Therefore, Pig-o is involved in but not essential for GPI anchor biosynthesis.
Transfection of Pig-o cDNA into Pig-o knockout cells normalized the surface Thy-1 expression as expected (Fig. 3B, left  panel). Transfections of cDNA of Pig-n (16), a homologous gene involved in transfer of EtNP to Man1, and Pig-f cDNA (22), as well as a control ALDH cDNA, did not restore the surface Thy-1 expression (Fig. 3B, right three panels).
PIG-O Is an ER Membrane Protein and Forms a Complex with PIG-F-To begin to clarify the functional relationships between PIG-O and PIG-F, we determined the subcellular localization of these two proteins. We cotransfected cDNAs of GST-tagged PIG-O and FLAG-tagged PIG-F into CHO cells and separated the postnuclear membranes with sucrose density gradient centrifugation. Most of the PIG-O and PIG-F was found in fraction 5, which contained the ER membranes, but not in the fractions containing the plasma membranes and the Golgi membranes (Fig. 4A). We next tested whether the two proteins associate with each other. FLAG-tagged PIG-F was cotransfected with GST-tagged Pig-o, ALDH, and Pig-n separately into CHO cells. ALDH (38) and Pig-n (16) are ER membrane proteins. After the membranes had been dissolved with 1% digitonin, the GST-tagged proteins were precipitated with glutathione beads, and coprecipitation of FLAG-tagged proteins was analyzed by Western blotting (Fig. 4B). FLAG-tagged PIG-F was coprecipitated (Fig.  4B, middle panel) with GST-tagged PIG-O (Fig. 4B, lane 1) but not with GST-tagged ALDH (lane 2) or GST-tagged Pig-n (lane 3). Therefore, PIG-O specifically associated with PIG-F. It was noted that much of PIG-F existed without association with PIG-O (Fig. 4B, lane 1, middle versus bottom panel).
PIG-F Stabilizes PIG-O-Because PIG-O and PIG-F were bound with each other, one might affect the expression of the other. To investigate this, we made a stable transfectant of class F cells (19,20) in which GST-tagged PIG-O and dolichol phosphate mannose 1, an ER membrane protein (33), were expressed. Then the transfectant cells were transiently transfected with PIG-F cDNA or an empty vector and expression of GST-tagged proteins was assessed (Fig. 5A). Expression of GST-tagged PIG-O was much higher in the presence (Fig. 5A,  (Fig. 5B). The expression of PIG-F was not affected by PIG-O because a ratio of PIG-F to ALDH did not change with (Fig. 5B, lane 1) or without (lane 2) PIG-O. These results showed that PIG-O was associated with PIG-F in the ER, that stable expression of PIG-O was dependent upon PIG-F and that PIG-F was stable in the absence of PIG-O.
Transfection of GST-tagged PIG-O without PIG-F cDNA caused a weak but significant expression of GST-tagged PIG-O in class F cells (Fig. 5A, lane 2). However, the surface expression of Thy-1 was not restored at all under these conditions (data not shown). Therefore, PIG-O requires the presence of PIG-F to restore GPI anchor biosynthesis, suggesting a functional role of PIG-F in addition to its ability to stabilize PIG-O.
The  (Fig. 6). In wild-type F9 cells, H8, a mature GPI anchor competent for transfer to proteins, was seen (Fig. 6, lane 1). Gaa1 knockout cells (23), in which a gene for GPI transamidase component is disrupted, accumulated various GPI intermediates (lane 2). Class B (lane 3) and F (lane 4) cells accumulated major spots, B and H6, respectively, which were previously characterized as GPIs containing two and three mannoses with EtNP on Man1, respectively (19,20,39). See Fig. 7 for a schematic representation of the structures of various GPI intermediates and mature GPIs. Pig-o knockout cells accumulated H6 as a major spot (Fig. 6, lane 5

), similarly to class F cells (lane 4).
Pig-o knockout cells accumulated two minor spots, KO-1 and KO-2 (Fig. 6, lane 5) whereas class F cells accumulated one, F-1 (lane 4). KO-1 and F-1 had similar mobilities. GPI biosynthesis in Pig-o knockout cells was normalized by transfection of cDNAs of PIG-O and GST-tagged PIG-O (Fig. 6, lanes 6 and 7), indicating that the GST-tagged PIG-O used in the experiments shown in Figs. 4 and 5 was functional. Transfection of Pig-n cDNA (Fig. 6, lane 8), PIG-F cDNA (lane 9), and an empty vector (lane 10) had no effect. These results suggest that Pig-o knockout and class F cells accumulate the same major GPI but different minor GPIs.
To clarify this difference, we characterized KO-1, KO-2, and F-1 as well as H6. KO-1 and KO-2 were completely cleaved and H6 was partially cleaved by GPI-phospholipase D, showing that they are GPIs (Fig. 8A). We purified H6, KO-1, KO-2, and F-1 from TLC plates. Purified H6 from Pig-o knockout and class F cells (Fig. 8B, lanes 3 and 5), as well as unpurified H6 from Gaa1 and Pig-o knockout cells (lanes 1 and 6), had the same mobility. After digestion with jack bean ␣-mannosidase, they migrated to the same position that aligned with H5 (Fig.  8B, lanes 2 and 4), consistent with their identity as H6. PIG-O, therefore, like PIG-F, is involved in transfer of EtNP to Man3.
Purified KO-1 and F-1 showed similar migration (Fig. 8C,  lanes 3 and 5). After digestion with ␣-mannosidase, their migration shifted similarly (Fig. 8C, lanes 2 and 4), suggesting that KO-1 and F-1 are the same glycolipids. The number of mannoses removed by ␣-mannosidase was not clear but may be two, as judging from the extent of mobility shift. Purified KO-2 (Fig. 8C, lane 7) migrated more slowly than a GPI termed H7, seen in Gaa1 knockout cells (lane 8). KO-2 was sensitive to ␣-mannosidase and migrated to a position similar to that of H7 (Fig. 8C, lane 6), indicating that KO-2 had one or more unmodified mannoses at its terminus. H7 has three mannoses, with EtNPs on Man1 and Man3, and is thought to be competent for transfer to proteins. It is suggested, therefore, that KO-2 would have at least one more mannose than H7 and be able to anchor proteins. This idea that KO-2 may account for the residual surface expression of GPI-anchored proteins on Pig-o knockout cells is consistent with the fact that class F cells, which are completely deficient in the expression of GPI-anchored proteins, have no KO-2.
To further confirm that KO-1 and KO-2 have EtNP on Man1, we treated wild-type and Pig-o knockout cells with 10 M YW3548/BE49385A, a drug that inhibits addition of EtNP to Man1 (16,17). In the presence of the drug, wild-type F9 cells accumulated H7Ј and H7Љ as reported previously (Fig. 9, lanes  3 and 4) (16), indicating that EtNP addition to Man1 was inhibited under these conditions. The drug inhibited generation of KO-1 and KO-2 ( Fig. 9, lanes 1 and 2), indicating that they had EtNP on Man1. DISCUSSION In the present study, we cloned and characterized Pig-o and analyzed the roles of PIG-O and PIG-F in GPI biosynthesis. The first conclusion of this study is that PIG-O and PIG-F act  lanes 4 and 5), and KO-2 (lanes 6 and 7) were treated with jack bean ␣-mannosidase (lanes 2, 4, and 6) or buffer alone (lanes 3, 5, and 7). Lanes 1 and 8 were unpurified mannolipids from Pig-o knockout and Gaa1 knockout F9 cells, respectively. together in transferring EtNP to Man3. This is based on two results: 1) Pig-o knockout cells and class F cells defective in PIG-F accumulated the same major GPI intermediate, H6, which bears three mannoses lacking EtNP on Man3; and 2) PIG-O and PIG-F formed a protein complex in the ER, and the stability of the former was dependent upon the latter. The second conclusion is that although this mechanism of EtNP transfer to Man3 is predominant, there must be a second mechanism that also requires PIG-F but not PIG-O. This conclusion is supported by two observations: 1) Pig-o knockout cells express a low level of GPI-anchored proteins, whereas class F cells are completely deficient in the surface expression of GPIanchored protein; and 2) Pig-o knockout but not class F cells generated a minor GPI with more than three mannoses that has EtNPs on Man1 and Man3.
Association of PIG-O and PIG-F-Both PIG-O and PIG-F were found exclusively in the ER, where the GPI anchor is synthesized, and they specifically associated with each other. PIG-F, which is a very hydrophobic protein, may associate with the carboxyl-terminal hydrophobic region of PIG-O.
The expression of PIG-O was dependent upon PIG-F, i.e. the level of PIG-O expression was at least three times higher in the presence than in the absence of PIG-F. However, the expression of PIG-F was not affected by a lack of PIG-O. Presumably, most PIG-O was associated with PIG-F, whereas some PIG-F may exist free from PIG-O. These results suggest that PIG-O acts together with PIG-F in the ER and that PIG-F may have at least one more partner that is involved in the second mechanism of transferring EtNP to Man3 (see below for further discussion).
Common and Different Phenotypes of Pig-o Knockout and Class F Mutant Cells-Pig-o knockout and class F cells share common defective phenotypes but they have some differences too. Both cells are deficient in the surface expression of GPIanchored proteins. However, low levels of Thy-1 and ScaI remained on the Pig-o knockout F9 cells. In contrast, class F cells are completely deficient in the surface Thy-1 expression. Therefore, PIG-O is involved in but not essential for GPI-anchoring of proteins, whereas PIG-F is essential for it. Both Pig-o knockout and class F cells did not generate a mature GPI, H8, but accumulated a GPI intermediate, H6. H6 has three mannoses and EtNP modification on Man1 but lacks EtNP on Man3 (10). Therefore, both cells are defective in the transfer of EtNP to Man3 (see Fig. 7 for the pathway). Class F cells generated one minor GPI, termed F-1, whereas Pig-o knockout cells generated two minor GPIs, termed KO-1 and KO-2. We concluded that F-1 and KO-1 are the same GPI (see below for discussion), but KO-2 was seen only in Pig-o knockout cells.
Possible Structures of GPIs, KO-1/F-1 and KO-2-KO-1 and F-1 had similar mobilities on TLC, slightly slower than that of H7. H7 bears three mannoses with EtNPs on Man1 and Man3. After the treatment with ␣-mannosidase, KO-1 and F-1 behaved similarly and became a GPI migrating slightly more slowly than H6. H6 bears three mannoses with EtNP on Man1. KO-1 must have EtNP on Man1 because its generation was inhibited by YW3548/BE49385A. Although we did not analyze the number of mannose residues in KO-1 and F-1, we speculate based on these results that they have three or four mannoses with EtNP on Man1 and Man2 (Fig. 7).
KO-2 migrated more slowly than H7, and after ␣-mannosidase treatment, its product behaved similarly to H7 on TLC. KO-2 has EtNP on Man1 because its generation was inhibited by YW3548/BE49385A. KO-2, therefore, has more than three, most likely four, mannoses with EtNP on Man1 and Man3.
Two Mechanisms of Transfer of EtNP to Man3 in Mammalian Cells-The complex of PIG-O and PIG-F should be respon-sible for the addition of EtNP to Man3 in H6 to generate H7 and subsequently H8. It is very likely that PIG-O bears a catalytic site because three family members, Mcd4p/Pig-n, Gpi7p, and PIG-O, correspond to EtNP additions to Man1, Man2, and Man3, respectively, and because they have regions with homology to various phosphodiesterases and nucleotide pyrophosphatases. Consistent with the idea that these regions are involved in the catalytic site, the temperature-sensitive mutant allele of MCD4 had a mutation within one of these regions (15). These regions are within the lumenal domains, suggesting that transfer of EtNP to Man3 occurs on the lumenal side of the ER.
PIG-F is required for the stable expression of PIG-O. It is unlikely that this is the only role of PIG-F because PIG-O expressed in the absence of PIG-F did not cause the surface Thy-1 expression in class F cells. PIG-F may play some role in the addition of EtNP to Man3.
In the absence of PIG-O, GPI-anchored proteins were expressed at low levels. Because KO-2 may have Man3 with EtNP on it, it seems possible that KO-2 accounts for the residual GPI-anchoring in Pig-o knockout cells. Generation of KO-2 should require PIG-F because it was not seen in class F cells. Consistent with the notion that KO-2 is competent for attachment to proteins, class F cells are completely deficient in the surface expression of GPI-anchored proteins. Because PIG-F may not be a catalytic component, it would act together with a catalytic component other than PIG-O to generate KO-2. Because Gpi7p is responsible for addition of EtNP to Man2 at position 6 in yeast and EtNP on Man3 is also at position 6, a mammalian homologue of Gpi7p seems to be a candidate of the second partner for PIG-F. Human and mouse sequences homologous to Gpi7p are found in the GenBank TM data base. They should be cloned and characterized to determine whether they represent a functional homologue of Gpi7p and act with PIG-F in transferring EtNP to Man3.