YLL031c Belongs to a Novel Family of Membrane Proteins Involved in the Transfer of Ethanolaminephosphate onto the Core Structure of Glycosylphosphatidylinositol Anchors in Yeast*

MCD4 and GPI7 are important for the addition of glycosylphosphatidylinositol (GPI) anchors to proteins in the yeast Saccharomyces cerevisiae. Mutations in these genes lead to a reduction of GPI anchoring and cell wall fragility. Gpi7 mutants accumulate a GPI lipid intermediate of the structure Manα1–2[NH2-(CH2)2-PO4→]Manα1–2Manα1–6[NH2-(CH2)2-PO4→]Manα1–4GlcNα1–6[acyl→]inositol-PO4-lipid, which, in comparison with the complete GPI precursor lipid CP2, lacks an HF-sensitive side chain on the α1–6-linked mannose. In contrast, mcd4–174 accumulates only minor amounts of abnormal GPI intermediates. Here we investigate whether YLL031c, an open reading frame predicting a further homologue of GPI7and MCD4, plays any role in GPI anchoring. YLL031c is an essential gene. Its depletion results in a reduction of GPI anchor addition to GPI proteins as well as to cell wall fragility. YLL031c-depleted cells accumulate GPI intermediates with the structures Manα1–2Manα1–2Manα1–6[NH2-(CH2)2-PO4→]Manα1–4GlcNα1–6[acyl→]inositol-PO4-lipid and Manα1–2Manα1–2Manα1–6Manα1–4GlcNα1–6[acyl→]inositol-PO4-lipid. Subcellular localization studies of a tagged version of YLL031c suggest that this protein is mainly in the ER, in contrast to Gpi7p, which is found at the cell surface. The data are compatible with the idea that YLL031c transfers the ethanolaminephosphate to the inner α1–2-linked mannose, i.e. the group that links the GPI lipid anchor to proteins, whereas Mcd4p and Gpi7p transfer ethanolaminephosphate onto the α1–4- and α1–6-linked mannoses of the GPI anchor, respectively.

The structural analysis of GPI 1 anchors revealed that different organisms add different side chains to the universally conserved Man␣1-2[NH 2 -(CH 2 ) 2 -PO 4 3]Man␣1-2Man␣1-6Man␣1-4GlcN␣1-6inositol-PO 4 -lipid core structure ( Fig. 1) (1). In particular, human and all other mammalian cells invariably add an EtN-P group onto Man1 (Fig. 1), and in a few cases they also add an EtN-P group to Man2 (for a review, see Ref. 2). Although no EtN-P side chains had been found in a study of a pool of protein-derived GPI anchors of Saccharomyces cerevisiae (3), it recently was shown that the GPI intermediate M2* accumulating in a gpi10 mutant has the structure Man␣1-6[NH 2 -(CH 2 ) 2 -PO 4 3]Man␣1-4GlcN␣1-6[acyl3]inositol-PO 4lipid (4). Moreover, the complete GPI anchor precursor lipid CP2 that can be observed in mutants such as pmi40 or in transamidase mutants (5-7) contains phosphodiester-linked side chains on both Man1 and Man2 (4,8). This suggests that the side chains on Man1 and Man2 may have been invented before the separation of the lineages leading to present day yeast and mammals. Recent studies have implicated MCD4 and GPI7 in the transfer of EtN-P groups onto GPI structures in yeast. MCD4 is essential, and mcd4 -174 cells show an almost complete defect in GPI anchor addition to proteins at the restrictive temperature (9,10). Deletion of PIG-N, a mammalian homologue of MCD4, leads to a reduced surface expression of GPI proteins and loss of an enzyme activity that adds EtN-P to Man1 (11). GPI7 is not essential, but its deletion slows the GPI anchor addition to newly synthesized proteins, produces cell wall fragility, and blocks biosynthesis of GPI lipids at a premature stage. Of the several abnormal GPI lipids accumulating in ⌬gpi7 cells, even the most complete one lacks the phosphodiester-linked substituent on Man2, a finding that relates GPI7 to the addition of a side chain to Man2 (8). The idea that MCD4, GPI7, and their homologue YLL031c may all be involved in the transfer of EtN-P groups has been fostered by their pronounced homology with bacterial, viral, plant, and mammalian phosphodiesterases, phosphatases, and pyrophosphatases (8,10).
Here we investigate whether YLL031c plays any role in GPI biosynthesis and, specifically, in the addition of EtN-P-groups to GPI structures.
Materials were obtained from the sources described recently (4 Conditional Expression of YLL031c and MCD4 -Conditional expression of YLL031c was achieved by the insertion of the GAL1,10 promoter in front of the chromosomal YLL031c gene as described (13). Briefly, the HIS3 marker flanked by the GAL1,10 promoter was PCR-amplified using pTL26 as template and the following two adapter primers: 031GalFor (5Ј-AAGATCAAAAAAGGAATAGAAGCATATGTTTTAAG-GGCAACGCCGctcttggcctcctctag-3Ј) with 17 nt of homology to the pTL26 vector (lowercase) and 45 nt of homology to a 5Ј-flanking sequence of YLL031c (uppercase) and oligonucleotide 031GalRev (5Ј-AAG-AATCGACTTTTTAATTGTCTTTTCATCCATATTACGGGAGCTcgaattccttgaattttcaaa-3Ј) with 45 nt of homology to the 5Ј-end of YLL031c ending 12 nt upstream of the start codon (boldface type) and 21 nt of homology to the pTL26 vector (lowercase type). This PCR-generated DNA fragment was used to transform the strains FCEN010a and FBY413, yielding FBY1103 and FBY1102, respectively. Correct targeting of the inserted promoter was verified by whole yeast cell PCR (14), using primers A3Gal (5Ј-gagcagttaagcgtattactg-3Ј) and 031A4 (5Ј-catcaatgaaagtcggtaagg-3Ј) yielding a 0.7-kilobase pair DNA fragment.
The genomic MCD4 was placed under the control of the GAL1,10 promoter by the same strategy as described above for YLL031c, using the following two adapter primers for PCR amplification from the pTL26 as the template: 165GalFor (5Ј-GAACCGTTCTTTACTATATAT-TCAACAACCCATCTTCGACCAAAGctcttggcctcctctag-3Ј) with 17 nt of homology to the pTL26 vector (lowercase type) and 45 nt of homology to the MCD4 flanking sequence (uppercase type) and oligonucleotide 165GalRev (5Ј-GACACCAACAGCCAGAAGCGTCGTTCTGGTTTTGT-TCCACATTTTcgaattccttgaattttcaaa-3Ј) with 45 nt of homology to MCD4 ending 3 nt upstream of the start codon (boldface type) and 21 nt of homology to the pTL26 vector (lowercase type). This PCR-generated DNA fragment was used to transform the strain FCEN010a. Correct targeting of the HIS3-marked GAL1,10 promoter was verified by whole yeast cell PCR using primers A3Gal (5Ј-gagcagttaagcgtattactg-3Ј) and 165A4 (5Ј-agcaatcatagcaacatgacc-3Ј) yielding a 0.5-kilobase pair DNA fragment.
Disruption of YLL031c-For deletion of YLL031c, a short flanking homology (SFH) replacement cassette was synthesized by PCR amplification of the KanMX4 module of pFA6a-KanMX4 (15) with the following primers: primer 031-S1 (5Ј-GATGAAAAAATAATATACAAATCGC-GAATAAAGAAATTTCAAcgtacgctgcaggtcgac-3Ј) with 42-nt homology to amino acids 16 -29 of YLL031c and 18-nt homology to the KanMX4 module (lowercase type) and primer 031-S2 (5Ј-TATAGTATATTTGT-AAGTAAAGAGTGGAAATGAAGTTCGTCATTatcgatgaattcgagctcg-3Ј) with 19 nt of homology to the KanMX4 module (lowercase type) and 44 nt of homology to a sequence near the 3Ј end of YLL031c (uppercase type) comprising the stop codon (boldface type). Since gene replacement with this SFH cassette proved to be difficult, the SFH cassette was used as a template to construct a long flanking homology cassette. This long flanking homology replacement cassette was synthesized as described (16). In a first step, the primer pair K1 (5Ј-CATGGTACAATTGCAAA-GT-3Ј) and L2 (5Ј-TTGAAATTTCTTTATTCGCGATTTGTATATTAT-3Ј) as well as the primer pair L3 (5Ј-AATGACGAACTTCATTTCCACTC-TTTACTTACAAATATAC-3Ј) and A4 (5Ј-GCTGCAGAAAAGAGATGC-3Ј) were used to amplify two fragments of genomic DNA corresponding to the promoter region including the first 29 codons of the open reading frame and the terminator region, respectively. The 5Ј-parts of primers L2 and L3 were designed to anneal, respectively, to the 5Ј-and 3Ј-ends of the SFH replacement cassette, which are homologous to YLL031c. In the second step, a PCR was performed using the SFH cassette as template and the two PCR-generated fragments from the first step as primers, yielding a replacement cassette with long flanking regions having homology to the YLL031c gene. The long flanking homology cassette was used to transform the diploid WT strain FY1679, homozygous for YLL031c. The correct targeting of the KanMX4 replacement cassette to the YLL031c locus was verified by whole yeast cell PCR using primer K1, primer 031-A5 (5Ј-catcaatgaaagtcggtaagg-3Ј) annealing inside the YLL031c coding sequence, and the K2 primer of the KanMX4 module (5Ј-gtcgcacctgattgcccg-3Ј), yielding a 900-bp DNA fragment for the disrupted gene and a 748-bp DNA fragment for the WT YLL031c gene.
Tagging of YLL031c-A Myc tag was inserted at the C terminus of the genomic copy of YLL031c by homologous recombination with an insertion cassette containing the Myc tag and the selectable KanMX6 marker as described (17). The insertion cassette was obtained by using plasmid pFA6a-13Myc-kanMX6 (17) as template and the oligonucleotides F2 and R1 as target gene-specific primers (F2, 5Ј-AGTGGGAGA-TTGATTAAGCACATAAATGACATTTTTTGGAAAcggatccccgggttaattaa-3Ј, with 20-nt homology to the pFA6a-13Myc-kanMX6 sequence (lowercase type) and 42-nt homology to the sequence of YLL031c immediately upstream of the stop codon (uppercase type), including the codon of the last amino acid (boldface type); R1, 5Ј-ATATATAGTATA-TTTGTAAGTAAAGAGTGGAAATGAAGTTCGgaattcgagctcgtttaaac-3Ј with 42-nt homology to the genomic sequence of YLL031c immediately downstream of the stop codon (uppercase) and 20-nt homology to the pFA6a-13Myc-kanMX6 plasmid sequence). This PCR fragment was used to transform the WT strain FBY413, yielding FBY1107. The correct targeting of the PCR-made module was verified by whole yeast cell PCR using the primer K2 of the kanMX module and oligonucleotide C1-Myc (5Ј-CTGACACTGTGGTCACAGCC-3Ј), producing a 1577-nt fragment.
Cellular Localization and Protease Sensitivity of YLL031c-Myc Protein-The subcellular localization of YLL031c-Myc was determined essentially as described (18). Briefly, 100 optical density units of exponentially growing FBY1107 cells were washed with 10 mM NaN 3 and converted to spheroplasts by incubation for 1 h at 30°C with Quantazyme in 1.4 M sorbitol, 10 mM NaN 3 , 50 mM K 2 HPO 4 , pH 7.5, 40 mM 2-mercaptoethanol. Spheroplasts were resuspended in lysis buffer (20 mM HEPES-KOH, pH 6.8, 150 mM potassium acetate, 250 mM sorbitol, 1 mM magnesium acetate, 20 g/ml phenylmethylsulfonyl fluoride, 5 g/ml antipain, 0.5 g/ml leupeptin, 0.7 g/ml pepstatin) and homogenized by 20 strokes in a Dounce homogenizer. The crude lysate was centrifuged twice at 1000 ϫ g for 5 min to remove unlysed spheroplasts. The cleared supernatant was then centrifuged at 13,000 ϫ g for 15 min to generate pellet P13 and supernatant S13. S13 was centrifuged at 100,000 ϫ g for 60 min in a Sorvall AH-650 swing out rotor to generate pellet P100 and supernatant S100. Pellet fractions were dissolved and denatured in high urea buffer (8 M urea, 5% SDS, 200 mM Tris-HCl, pH 6.8, 20 mM EDTA, bromphenol blue, 15 mg/ml dithiothreitol) by sonication and incubation at 95°C for 5 min. The S100 fraction was trichloroacetic acid-precipitated before denaturation in high urea buffer.
For probing the surface exposure of YLL031c-Myc, exponentially growing FBY1107 cells were converted to spheroplasts as described above but using zymolyase-20T. Spheroplasts were either mock-treated or treated with proteinase K in the presence or absence of 0.5% Triton X-100 for 30 min on ice. Phenylmethylsulfonyl fluoride was then added to a final concentration of 4 mM, and samples were kept on ice for 15 min. Finally, spheroplasts were placed on a cushion of 1.5 M sorbitol, 50 mM K 2 HPO 4 , pH 7.5, 20 mM EDTA, 10 mM NaN 3 , 10 mM NaF and were centrifuged. YLL031c protease sensitivity in microsomes was examined by protease K digestion essentially as described (8), except that the membrane pellet was resuspended in 600 l of lysis buffer before protease digestions. All samples were denatured during 5 min at 95°C in reducing sample buffer containing 20 mM EDTA and analyzed by SDS-polyacrylamide gel electrophoresis in 6 -10% gels (19) followed by Western blotting.
Miscellaneous Methods-Western blots were revealed using the chemiluminescence ECL kit from Amersham Pharmacia Biotech.
Previously described procedures were used to label cells with [2-3 H]Ins and for lipid extraction (20).

RESULTS
Characterization of YLL031c-YLL031c was disrupted in a diploid strain by the replacement of most of the open reading frame by the KanMX4 kanamycin resistance gene. This deletion strain was constructed because in the YLL031c deletion strain produced in the EUROFAN project, part of YLL030c has been deleted along with YLL031c (24). Sporulation and dissection of tetrads yielded in all cases two growing colonies per tetrad. None of these colonies were kanamycin-resistant, indicating that YLL031 is essential. A partial depletion of YLL031 was achieved by the insertion of the glucose-repressible GAL1,10 promoter immediately upstream of the genomic YLL031c. When such cells, growing in liquid culture, were shifted from galactose to glucose, thus allowing for the partial depletion of the YLL031c protein, their doubling time increased, after 24 h, to 3.5 h, and cells thereafter continued to grow at this rate, whereas WT cells grew on glucose with a doubling time of 1.3 h. When the same experiment was done in a ⌬gpi7 background, the doubling time of YLL031c-depleted cells, after 24 h on glucose, had risen to 7.7 h, while ⌬gpi7 mutants grew as fast as WT cells. Thus, the deletion of GPI7 synthetically enhances the growth retardation caused by depletion of YLL031c. The reduced growth rate of YLL031c-depleted cells could also be observed on YPD plates, and this effect was significantly enhanced by the presence of CFW at a concentration that did not affect growth of WT cells but totally blocked the growth of ⌬gpi7 cells (Fig. 2). This result indicates that partial depletion of YLL031c leads to cell wall fragility, a phenotype that is commonly observed in mutants affected in the GPI biosynthesis pathway.
Depletion of YLL031c Leads to the Accumulation of Immature GPI Proteins-If Gas1p does not receive a GPI anchor in the ER, it fails to be transported to the Golgi (25), because, as shown by in vitro experiments using a vesicle budding assay, nonanchored Gas1p is not packaged into COPII-coated transport vesicles budding off the ER (26). Therefore, the characteristic mass increase due to elongation of N-and O-glycans in the Golgi does not take place, and the maturation of the immature 105-kDa ER form into the mature 125-kDa form is delayed. GPI anchoring deficiencies also can lead to the relative depletion of the mature forms of Gas1p and other GPI proteins (27). Using Western blotting and antibodies to detect well characterized GPI proteins, we could observe a significant depletion of mature GPI proteins upon depletion of YLL031c. As can be seen in Fig. 3, while in WT cells the carbon source affects the amount of Gas1p only moderately, cells partially depleted of YLL031c by growth on glucose contain much less of the mature 125-kDa form of Gas1p than cells grown on galactose. This phenomenon is even more drastic when YLL031c depletion occurs in a ⌬gpi7 background. Other GPI proteins such as Yap3p and Cwp1p are also affected. Although the details of the  3. Depletion of YLL031c leads to a depletion of mature and accumulation or depletion of immature forms of GPI proteins. Cells indicated at the top were grown for at least four generations at 30°C on either Gal or Glc except for sec18 cells, which were grown at 24°C but were shifted to 37°C 2 h before extraction. Exponentially growing cells were lysed by boiling for 5 min in reducing sample buffer and processed for SDS-polyacrylamide gel electrophoresis on 7.5 or 10% gels (19) followed by Western blotting with antibodies against Gas1p, Yap3p, Cwp1p, or carboxypeptidase Y (CPY). The expected amounts of Gpi7p and YLL031c protein in these cells are indicated as in Fig. 2. biosynthesis of these proteins have not been reported and part of their mature forms gets covalently attached to the cell wall, the mass of their ER forms can be inferred form the peptides accumulating in sec18, a mutant that blocks the protein traffic from ER to Golgi at 37°C (28,29). YAP3 predicts a 60-kDa translation product, but upon shift to 37°C, sec18 cells rapidly accumulate lower molecular weight forms (Fig. 3). Similar forms also accumulate upon depletion of YLL031c, suggesting a maturation defect of these proteins due to a delay in GPI anchor addition. Cwp1p is a cell wall protein that can be released as soluble 55-60-kDa protein by ␤-glucanase (30,31). There are also intracellular, detergent-soluble forms of Cwp1p of 48 and 58 kDa, and these forms are drastically increased when the secretory pathway is blocked for 2 h at 37°C in sec18 (Fig. 3). In YLL031c-depleted cells, these detergent-soluble forms are severely diminished. This also can be taken as evidence for some disturbance of GPI protein maturation, although in this case the protein does not seem to accumulate but may be rapidly degraded. In contrast, YLL031c-depleted cells do not accumulate the typical ER proform p1 of carboxypeptidase Y, a vacuolar hydrolase, nor do they show a thinning of mature carboxypeptidase Y (Fig. 3). This suggests that the maturation defect of YLL031c-depleted cells affects only GPI proteins.
031a and 031b are less polar than M4, suggesting that they have smaller head groups and are earlier intermediates of GPI biosynthesis than M4. It therefore is not unexpected that the accumulation of 031b is epistatic to the accumulation of M4 both in vitro and in vivo. Indeed, the combination of ⌬gpi7 with YLL031c depletion strongly reduces the accumulation of M4 and increases the accumulation of 031b (Fig. 4, A, lane 6, and  C, lanes 5 and 6). Since the depletion of YLL031c is only partial, it is understandable that in vitro there still is residual biosynthesis of CP2 or M4 in single or double mutants, respectively (Fig. 4C, lanes 4, 6, and 11). The severely affected mcd4 -174 mutant has been reported to accumulate only trace amounts of abnormal [ 3 H]Ins-labeled lipids (10). We find the same if Mcd4p is depleted using the repressible GAL1,10 promoter (Fig. 4A, lane 5).
Structural Characterization of Lipids 031a and 031b-YLL031c-depleted cells were labeled with [ 3 H]Ins, and lipids 031a and 031b were purified by preparative TLC. Their structural analysis yielded the following information. Both lipids are effectively cleaved by GPI-specific phospholipase D, indicating that they are GPIs. They are totally resistant to bacterial, phosphatidylinositol-specific phospholipase C, suggesting that they contain a protecting acyl group on the Ins. Their label becomes entirely hydrophilic upon deacylation by mild alkaline hydrolysis.
Analysis of the hydrophilic head group of lipid 031a indicates that it contains a Man 4 -GlcN-Ins core structure (Fig. 6A) and that all of its four mannoses can be removed by JBAM not only after but also before HF dephosphorylation (Fig. 6, C and B). These data strongly suggest that lipid 031a has the structure Man␣1-2Man␣1-2Man␣1-6Man␣1-4GlcN␣1-6[acyl3]inositol-PO 4 -lipid. Thus, the GPI structures accumulating upon YLL031c depletion contain four mannoses and may contain HF-sensitive side chains on Man1 or Man2 but lack EtN-P on Man3 (Fig. 1).
Subcellular Localization of YLL031c-YLL031c predicts a translation product of 116 kDa. To localize the YLL031c protein, the endogenous, chromosomally encoded gene was modified by the insertion of a Myc tag at its C terminus. This YLL031c-Myc translation product (including the N-terminal signal sequence) has a predicted molecular mass of 136 kDa. The corresponding protein was detected in Western blots as a major band at about 136 kDa plus a heterogeneously glycosylated smear with another distinct band at about 300 kDa. The cells containing the tagged YLL031c grew at the same rate as WT cells and were completely resistant to CFW, indicating that the tagged version of the protein is functional. As shown in Fig.  7B, when intact spheroplasts were treated with variable amounts of zymolyase and proteinase K, YLL031c-Myc as well as ER proteins such as Wbp1p and Gpi8p remained intact, whereas Gas1p was strongly diminished, as expected for a surface protein and as described before (8). Gpi7p could not be detected on any blot in this experiment (not shown) because of its previously reported susceptibility to zymolyase (8). When spheroplasts were prepared with Quantazyme, both YLL031c and Gpi7p were preserved, but Gpi7p was readily destroyed by proteinase K, whereas YLL031c-Myc was completely resistant (not shown). Thus, unlike Gpi7p, YLL031c-Myc does not reside at the cell surface. Differential centrifugation of microsomes indicated that YLL031c-Myc cofractions with ER markers Wbp1p and Gpi8p, since it could be found mainly in P13, which is enriched in ER and plasma membrane (Fig. 7A). Minor amounts were present in P100, a fraction that is enriched in FIG. 5. Analysis of the head group of lipid 031b using HF, JBAM, and A. satoi ␣-mannosidase (ASAM). FBY1102 was labeled with [ 3 H]Ins at 30°C, and lipid 031b was purified by two rounds of preparative TLC and used to prepare head groups. Head groups were subjected to the following sequential treatments. A, HF followed by N-acetylation; B, HF followed by N-acetylation, desalting, and then A. satoi ␣-mannosidase; C, JBAM followed by HF and then N-acetylation; D, HF followed by desalting, then JBAM, and then N-acetylation. The thus generated fragments were separated by paper chromatography, and radioactivity contained in 1-cm-wide strips was determined through scintillation counting. Standards 0 -4 run on the same paper are Man x -GlcNAc-Ins with x ϭ 0, 1, 2, 4.
FIG. 6. Analysis of the head group of lipid 031a using HF and JBAM. FBY1102 cells were labeled with [ 3 H]Ins at 30°C, lipid 031a was purified by two rounds of preparative TLC, and head groups were prepared. Head groups were subjected to the following treatments. A, HF followed by N-acetylation; B, JBAM followed by HF and then N-acetylation; C, HF followed by desalting, JBAM, and then N-acetylation. The thus generated fragments were separated as in Fig. 5.   FIG. 7. Subcellular localization of YLL031c. A, FBY1107 cell lysates were subjected to differential centrifugations at 13,000 and 100,000 ϫ g. These centrifugations generated pellet P13, which contains ER, plasma, and vacuolar membranes, and pellet P100, which contains Golgi membranes. The 100,000 ϫ g supernatant (S100) was precipitated with trichloroacetic acid. B, FBY1107 cells were converted to spheroplasts using various concentrations of zymolyase-20T and then digested with proteinase K at the indicated concentrations. Lane 6 shows the reactivity of the anti-Myc antibody and secondary antibody with an extract from FBY413 WT cells not harboring any Myc-tagged proteins. This control extract was prepared as described in the legend to Fig. 3. C, FBY1107 microsomes were sedimented at 13,000 ϫ g for 15 min, and the membrane pellet was thoroughly resuspended and digested with the indicated amounts of proteinase K at 0°C on ice for 20 min in the presence or absence of 0.5% Triton X-100. Samples were processed for SDS-polyacrylamide gel electrophoresis and Western blotting.
the Golgi marker Och1p and is largely free of Gpi8p and Wbp1p (Fig. 7). The existence of the 300-kDa form is indicative of massive glycan elongation in the Golgi on part of YLL031c-Myc. Since only little YLL031c-Myc is found in this organelle at steady state, it may be that the protein is recirculating between Golgi and ER as has been reported for other proteins such as Emp47p, Sec12p, and Sed5p. When microsomes were incubated with proteinase K, even low concentrations of protease (10 g/ml) readily destroyed the immunoreactivity of YLL031c-Myc (Fig. 7C). In contrast, Gpi8p, which has a large N-terminal luminal and a small C-terminal cytosolic domain of about 14 amino acids was slightly reduced in size but not destroyed by proteinase K. In view of the massive glycosylation of part of YLL031c-Myc and the presence of an N-terminal hydrophobic domain that qualifies as potential signal sequence, we interpret these findings in the sense that the N-terminal part of YLL031c containing six N-glycosylation sites is oriented luminally. The fact that proteinase K treatment of microsomes does not generate any immunoreactive low molecular weight product suggests that the C-terminal Myc tag is exposed cytosolically or that YLL031c-Myc contains a proteinase K-sensitive site close to its C terminus. The membrane topology of YLL031c thus appears to be the same as in the homologous Gpi7p (8). DISCUSSION In this study, we investigate the potential function of YLL031c in GPI anchoring. As shown in Fig. 8, YLL031c predicts a membrane glycoprotein of 1017 amino acids with an N-terminal hydrophilic domain and a C-terminal hydrophobic sequence containing numerous potential transmembrane domains, a feature that initially led to its classification as potential facilitator of membrane permeation (32).
YLL031c is homologous to GPI7 and MCD4 of S. cerevisiae, but closer homologues can be found in other species such as Homo sapiens, Drosophila melanogaster, Caenorhabditis elegans or Schizosaccharomyces pombe. The GPI7/MCD4/ YLL031c gene family can be subdivided into three subfamilies of more closely related genes. MCD4, GPI7, and YLL031c each belong to a different subfamily. All genes in this family predict proteins that have the same general structural attributes; i.e. they have an N-terminal signal sequence followed by a large hydrophilic domain and a C-terminal hydrophobic sequence containing numerous potential transmembrane domains. The hydrophilic domains of YLL031c, GPI7, and MCD4 have a distinct homology with mammalian enzymes classified as phosphodiesterases, phosphatases, or nucleotide pyrophosphatases. This homology extends over 240 amino acids. YLL031c and GPI7 also contain the distinct, universally conserved motif PTXTX 8 TGX 2 P (double bar above the sequence in Fig. 8A; stippled in Fig. 8B), which is found in mammalian but also some plant and bacterial enzymes. The homology among the members of the GPI7/MCD4/YLL031c gene family encompasses 400 amino acids at the N terminus of these proteins. The family contains several motifs that are not present or are only partially conserved in phosphatases (e.g. HXLGXDXXGH and DH-GMXXXGXHG (single bar above the sequence in Fig. 8A; hatched in Fig. 8B). These motifs fall into regions in which also FIG. 8. Homology of YLL031c with GPI7 and MCD4. A, the first 395 amino acids of YLL031c are shown. Potential Nglycosylation sites and the N-terminal hydrophobic sequence for translocation of the protein into the ER are underlined. Homologies with the other S. cerevisiae (OSC) genes GPI7 and MCD4 were identified using the Clustal W (version 1.8) program at EBI. This introduced a few gaps into the YLL031c sequence (dashes). In parallel, each of these three genes was aligned with its nearest neighbors in other species, namely humans, mice, C. elegans, and S. pombe. GenBank TM accession numbers and Geninfo identifiers of sequences aligned with YLL031c were T02245 GI:7513075, AAB93646 GI:2734088, and T40030 GI:7491546; aligned with GPI 7 were Q09782 GI:1175452 and T21487 GI:7500059, and aligned with MCD4 were NP_036459 GI: 6912500, NP_038812 GI:7305383, and T40715 GI:7491747. These alignments reveal homologies for the YLL031c (C31) subfamily, the GPI7 (equivalent to YJL062w) (C62) and the MCD4 (equivalent to YKL165c) (C65). Asterisks, colons, and periods indicate identity, strong similarity, and weak similarity, respectively. B, in the Kyte-Doolittle plot of YLL031 are integrated the positions of the nine potential N-glycosylation sites (vertical lines). The phosphatase motif (double bar), and two motifs conserved between the YLL031c, GPI7, and MCD4 subfamilies (single bars) are also indicated. the highest degree of homology is observed within each subfamily. As shown in Fig. 8A, the degree of consensus within each subfamily is significantly higher than between YLL031c, GPI7, and MCD4 (compare lines OSC, C31, C62, and C65). Assuming that all of these genes evolved from a common ancestor, this suggests that the divergence and functional diversification of these three subfamilies occurred prior to the separation of the lineages leading to S. cerevisiae, S. pombe, C. elegans, and humans.
The phenotype of YLL031c-depleted cells strongly suggests that YLL031c indeed is involved in GPI anchoring. The cell wall fragility of YLL031c-depleted cells as revealed by CFW hypersensitivity is a common feature of all gpi mutants (33)(34)(35)(36). Indeed, many GPI proteins of yeast are cell wall proteins or plasma membrane proteins participating in the building of the cell wall (37,38). The appearance of abnormal GPI lipid intermediates and the specific accumulation or disappearance of mature and immature forms of GPI proteins equally point to a GPI anchoring defect.
This defect seems to be due to a defect in making GPI lipids but not in the attachment process itself. This can be concluded, since YLL031c-depleted microsomes still make some CP2 in vitro, but YLL031c-depleted cells do not accumulate CP2 in vivo as is observed in mutants that are deficient in the GPI attachment process itself (6,7). The difficulty of YLL031cdepleted cells seems to reside in the attachment of the strategically important EtN-P to Man3, the EtN-P that will have to make the link to the protein moiety (Fig. 1). Earlier studies indicated that GPI7 is required for the addition of an HFsensitive substituent onto Man2, but not Man1 and Man3, and that PIG-N, the human homologue of MCD4, is required for the addition of EtN-P onto Man1 but not Man3. Together with the data presented here, it therefore appears reasonable to assume, as a working hypothesis, that MCD4, GPI7, and YLL031c are required for the addition of EtN-P to Man1, Man2, and Man3, respectively. Moreover, their homology with phosphodiesterases suggests that these three genes encode the cor-responding EtN-P-transferases. (These transferases may be expected to bear homology with phosphodiesterases, since previous studies by the group of Anant Menon (39,40) demonstrate that the attachment of the EtN-P to Man3 occurs by transesterification of EtN-P from phosphatidylethanolamine onto Man). A protein such as YLL031c would seem to be well suited to carry out this transfer inasmuch as its hydrophobic C-terminal domain may bind or even flip phosphatidylethanolamine, and the hydrophilic phosphodiesterase domain could operate the actual transfer reaction. Also, the subcellular localization of YLL031c in the ER is compatible with its functioning as an EtN-P-transferase, since it has been reported recently that, at least up to the stage of [NH 2 -(CH 2 ) 2 -PO 4 3] Man␣1-4GlcN␣1-6[acyl3]Ins-PO 4 -lipid, GPI lipids are made in the ER (41).
The structure of the abnormal GPI lipids accumulating in various gpi mutant cells may not tell us in what order these transfer reactions take place in WT cells, but, if we assume that the observed GPI structures represent biosynthetic intermediates, they may tell us in what order the physiological pathway does not work. The six theoretical possibilities for the order of EtN-P additions to Man1, Man2, and Man3 starting from a Man 4 -GlcN-Ins core structure are represented in Fig. 9. Pathways starting with reaction 1 (i.e. the addition of EtN-P by YLL031c) may not be possible, since we then would expect that MCD4-depleted cells or mcd4 mutants would accumulate the theoretical structure H1 or M4*, which is not the case (Fig. 4A, lane 6) (10). Pathways using reaction 3Љ, leading from 031b to H2, may also be impracticable, since YLL031c-depleted cells accumulate the substrate for reaction 3Љ, but H2 is not observed. On the other hand, the appearance of 031bЈ in these cells suggests that reaction 3 is possible albeit not favored. Since reactions 1 and 3Љ seem unfeasible and 3 is not favored, we may predict that the physiological pathway in normal cells proceeds mainly via the only remaining way, which is 2-1Љ-3ٞ. Thus, the addition of EtN-P to Man1 may represent a prerequisite for the addition of further EtN-P residues. The fact that MCD4 is essential is in accordance with a pivotal role of the EtN-P transfer to Man1. EtN-P is already present on M2* of gpi10 -1 having the structure Man␣1-6[NH 2 -(CH 2 ) 2 -PO 4 3] Man␣1-4GlcN␣1-6[acyl3]Ins-PO 4 -lipid. Also, several studies in mammalian cells document that EtN-P may be added before any other residues are attached to Man1 (11,23,42,43). Nevertheless, the abundance of 031a in YLL031c-depleted cells argues that the addition of EtN-P to Man1 is not a prerequisite for the addition of mannoses (Man2, Man3, and Man4, Fig. 1).
Since YLL031c is an essential gene, we may conclude that lipid 031b cannot be used as an anchor device (i.e. that at least some essential GPI proteins cannot be attached to the EtN-P on Man1 or that this process, if it occurs, does not yield a functional GPI protein). A similar conclusion has been reached previously in that the gpi10 -1 mutant does not seem to add the accumulating Man␣1-6[NH 2 -(CH 2 ) 2 -PO 4 3]Man␣1-4GlcN␣1-6[acyl3]Ins-PO 4 -lipid to proteins (4).
Genes other than YLL031c may be required for the transfer of EtN-P onto Man3. In mammals this step is dependent on PIG-F, a gene encoding a very hydrophobic protein. Its exact function has not yet been elaborated (44).
Further studies will be required to prove the validity of the model proposed in Fig. 9 and to show that the MCD4/GPI7/ YLL031c gene family indeed encodes EtN-P-transferases.
Acknowledgments-We are grateful to Drs Yves Bourbonnais, Hitoshi Shimoi, Dr. Y. Jigami, and Dr. M. Aebi for the generous gift of reagents. We thank Dr. Peter Orlean for sharing data before publication. We also thank Christine Vionnet and Anne Schneider for invaluable technical assistance.
FIG. 9. Biosynthesis of CP2: Are ethanolaminephosphates added in random order? All theoretically possible intermediates in the biosynthesis of CP2 from 031a are placed at the corners of a cube. The six nearest pathways along the edges leading from 031a to CP2 represent all of the possible orders of the EtN-P side chain addition to the carbohydrate core. GPI structures that have actually been identified are filled: 031a, 031b, and 031bЈ in this report; M2*, M4*, M4, and CP2 in previous publications (4,5,8). H1 and H2 indicate hypothetical structures, which have not been identified. Reactions 1, 2, and 3 are proposed to be carried out by YLL031c, Mcd4p, and Gpi7p, respectively.