Expression Cloning of PIG-L, a CandidateN-Acetylglucosaminyl-phosphatidylinositol Deacetylase*

Many eukaryotic cell surface proteins are bound to the cell membrane by a glycosylphosphatidylinositol (GPI) anchor. Several genes involved in GPI anchor biosynthesis have been cloned using complementation of mutant mammalian cell lines and yeasts that are defective in its biosynthesis pathway. However, the gene involved in the second step of this pathway, in whichN-acetylglucosaminyl-phosphatidylinositol (GlcNAc-PI) isN-deacetylated to form glucosaminyl (GlcN)-PI, has not been cloned. In this study, we established a GPI anchor-deficient mutant of Chinese hamster ovary (CHO) cells defective in the second step. Complementation analysis with the known GPI anchor mutant cells demonstrated that it belonged to the same complementation group as the CHO cell mutant G9PLAP.85. Using the new mutant, we cloned a rat gene termed PIG-L (forphosphatidylinositol glycan classL) that is involved in this step. PIG-L encodes a 252-amino acid, endoplasmic reticulum membrane protein, most of which is in the cytoplasmic side. This orientation of PIG-L protein is consistent with the notion that the second step of GPI anchor biosynthesis occurs on the cytoplasmic side of the endoplasmic reticulum.

Many eukaryotic cell surface proteins are bound to the cell membrane by a glycosylphosphatidylinositol (GPI) anchor. Several genes involved in GPI anchor biosynthesis have been cloned using complementation of mutant mammalian cell lines and yeasts that are defective in its biosynthesis pathway. However, the gene involved in the second step of this pathway, in which N-acetylglucosaminyl-phosphatidylinositol (GlcNAc-PI) is N-deacetylated to form glucosaminyl (GlcN)-PI, has not been cloned. In this study, we established a GPI anchor-deficient mutant of Chinese hamster ovary (CHO) cells defective in the second step. Complementation analysis with the known GPI anchor mutant cells demonstrated that it belonged to the same complementation group as the CHO cell mutant G9PLAP.85. Using the new mutant, we cloned a rat gene termed PIG-L (for phosphatidylinositol glycan class L) that is involved in this step. PIG-L encodes a 252-amino acid, endoplasmic reticulum membrane protein, most of which is in the cytoplasmic side. This orientation of PIG-L protein is consistent with the notion that the second step of GPI anchor biosynthesis occurs on the cytoplasmic side of the endoplasmic reticulum.
Many eukaryotic cell surface proteins are bound to the membrane by a carboxyl-terminal linkage to a glycosylphosphatidylinositol (GPI) 1 anchor (1,2). The conserved core of this glycolipid consists of a lipid containing inositol (usually phosphatidylinositol (PI)), a glucosamine, three mannose residues and phosphoethanolamine. At least 50 GPI-anchored proteins with a wide variety of functions have been identified in mammals, including cell-surface hydrolytic enzymes, receptors, adhesion molecules, complement inhibitors, and antigens of unknown functions. GPI anchors are also frequently used in many protozoa and yeasts (3).
GPI anchor synthesis occurs in the endoplasmic reticulum (ER) and essentially consists of the sequential addition of sugar components and phosphoethanolamine to PI. Several mutant mammalian cell lines defective in this biosynthesis pathway have been very useful in studying that of mammals. Three complementing mutants (classes A, C, and H) are known in the first reaction, which is the transfer of N-acetylglucosamine (GlcNAc) from UDP-GlcNAc to PI to form GlcNAc-PI (4,5). Genes for these mutants have been characterized: PIG-A (6), PIG-H (7), and PIG-C (8) for class A, H and C, respectively. In the second step, GlcNAc-PI is N-deacetylated to form GlcN-PI. Although two mutant cell lines are known in this step (9,10), the genes involved have not been cloned. Following inositol acylation of GlcN-PI (11), three mannoses are added to the GPI core. All of these mannoses are derived from dolichol phosphomannose (12)(13)(14). The class B mutant is defective in the third mannosylation (5,15), and the responsible gene, PIG-B, has been cloned (16). In the final step, phosphoethanolamine is transferred to the core, which contains three mannose residues (17). Class F is defective in this step (5,18), and PIG-F (19,20), which complements the class F mutant, has been cloned.
Mature GPI anchor precursors are post-translationally linked to proteins in the ER. Proteins that are to be GPIanchored have at their carboxyl terminus a signal sequence that directs GPI anchor addition (2). The signal peptide is exchanged with the GPI anchor by transamidation, forming an amide linkage between the new carboxyl terminus and ethanolamine (21). Class K mutant cells synthesize the mature GPI anchor precursors but are defective in the transamidation step (9,22). The gene for the class K mutation has not been cloned. The yeast Saccharomyces cerevisiae mutants gaa1 (23) and gpi8 (24) have similar defects to those of the mammalian class K mutant, indicating that at least two genes are involved in this step.
Most mammalian genes involved in GPI anchor biosynthesis have been cloned by means of expression cloning using mutant cell lines. To further elucidate the pathway, it is necessary to create new mutants. In this study, we established a mutant from Chinese hamster ovary (CHO) cells that were stably transfected with human decay-accelerating factor (DAF) and CD59 cDNAs to introduce known GPI-anchored proteins. This mutant was of the same complementation group as the mutant G9PLAP.85 (10), which is defective in the second step of the pathway, the deacetylation of GlcNAc-PI. The new mutant had no background expression of GPI-anchored protein and allowed us to clone a gene involved in this step.

EXPERIMENTAL PROCEDURES
Cells and Plasmids-The CHO-K1 cell mutant G9PLAP.85, defective in the second step of GPI anchor biosynthesis, and its wild-type * This work was supported by grants from the Ministry of Education, Science, Sports and Culture of Japan and the Uehara Memorial Foundation (Tokyo), and by Grant GM51419 from the National Institutes of Health.The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) D88364.
A rat cDNA library prepared from the rat C 6 glioma in an expression vector pMEPyori (20) was a gift from Dr. Y. Maeda. This library was prepared from oligo(dT)-primed, size-selected (Ͼ1.5 kb) cDNA and consisted of about 6 ϫ 10 5 independent clones.
A fragment containing PyT but lacking the polyoma origin of replication was obtained from pdl3027 (29) by digestion with BamHI, subcloned into the BamHI site of pBluescript II (pBSII) vector (Stratagene, La Jolla, CA), and then removed by EcoRV/SpeI digestion. The fragment was ligated into NruI/AvrII site of pcDNA I vector (Invitrogen, NV Leek, The Netherlands) to prepare pcDNA-PyT(ori Ϫ ).
Complementation Analysis-Complementation analysis with somatic cell fusion and immunofluorescence staining proceeded as described (31).
Cloning of PIG-L cDNA-A sample of 440 g of the cDNA library plasmids was cotransfected with 440 g of pcDNA-PyT(ori Ϫ ) plasmids into M2S2 cells (see "Results") (total 1.4 ϫ 10 8 cells) by electroporation. Two days later, transfected cells were double-stained with anti-CD59 monoclonal antibody 5H8 (a gift from Drs. M. Tomita and Y. Sugita, Showa University, Tokyo, Japan) in combination with fluorescein-conjugated sheep anti-mouse IgG (Organon Teknika Corp, West Chester, PA) and biotinylated anti-DAF monoclonal antibody IA10 (32) in combination with phycoerythrin-conjugated streptavidin (Biomeda, Foster City, CA), and sorted using FACS-Vantage (Becton Dickinson, Mountain View, CA). From the 707 DAF-and CD59-positive cells sorted, 1350 independent plasmid clones were recovered as described by Hirt (33) and by transformation into Escherichia coli MC1061. Pooled plasmids were transfected again into M2S2 cells, and the cells were analyzed for DAF and CD59 expression. DAF and CD59 expression was restored in about 10% of the cells. These pooled plasmids were transfected into M2S2 cells and selected again. Plasmids obtained from the sorted cells were cloned into 96-well plates and screened as described (16) to identify those that restore DAF/CD59 expression on M2S2 cells.
In Vitro Biosynthesis and Analysis of GPI Precursors-The 3-kb SmaI-SmaI fragment containing SR␣ promoter and PIG-L cDNA was removed from pMEPyori-PIG-L and ligated into the SmaI site of pBSII-PGK-Hyg (34) to make pBSII-PGK-Hyg-PIG-L. This construct was transfected into M2S2 and G9PLAP.85 cells, which were selected with 200 g/ml hygromycin. Finally, clones that permanently restored GPI anchor biosynthesis were analyzed.
Northern Blot Analysis of PIG-L-A rat PIG-L cDNA probe, the 0.7-kb SalI/XbaI fragment of pMEEB-FLAG-PIG-L (see below), was labeled with 32 P by the random primer labeling method using High Prime (Boehringer Mannheim, Mannheim, Germany). To prepare a hamster PIG-L cDNA probe, we synthesized degenerated primers each corresponding to nucleotides 541-563 and complementary to nucleotides 820 -842 of rat PIG-L cDNA and amplified a specific fragment of predicted size (302 bp) from reverse-transcribed CHO cell cDNA that was synthesized from total RNA of IIIB2A cells using oligo(dT) primer. This fragment was cloned, sequenced, and radiolabeled.
Expression and Analysis of PIG-L and Its FLAG-tagged Proteins-We fused FLAG (Eastman Kodak Co.) as a tag to the amino terminus of PIG-L as follows. We amplified the PIG-L gene from pMEPyori-PIG-L by PCR using a 5Ј primer (5Ј-TGTCGACGAAGTGGT-GGGTCTCTTG-3Ј) containing a SalI site instead of the start codon, and a 3Ј primer (5Ј-AGCAGGTCAGTCAGCCCAGGTCT-3Ј). The PCR product was subcloned into an EcoRV site of pBSII vector, digested with SalI/XbaI to obtain a fragment containing PIG-L, which was then ligated into a SalI/XbaI site of pMEEB-FLAG-PIG-A (26) to exchange PIG-A for PIG-L. We termed this construct pMEEB-FLAG-PIG-L.
To fuse FLAG to the carboxyl terminus of PIG-L, we amplified PIG-L gene from pMEPyori-PIG-L by PCR using a 5Ј primer (5Ј-CAGTT-TCAGCGGGTTCAGTTTTC-3Ј) and a 3Ј primer (5Ј-TTTACTTGTCA-TCGTCGTCCTTGTAGTCTCGCGAGAGCAACTGCAGCGAGTT-3Ј), with which we inserted an NruI site and FLAG sequences between the PIG-L carboxyl terminus and the stop codon. The PCR product was subcloned into the EcoRV site of pBSII vector, digested with ClaI/XbaI to obtain the fragment containing PIG-L-FLAG, which was then ligated into a ClaI/XbaI site of pMEEB (26). We termed this construct pMEEB-PIG-L-FLAG.
These constructs were transfected into M2S2 cells, which were selected with hygromycin to establish stable transfectants. FLAG-tagged PIG-L extracted from cells with 1% Nonidet P-40 were Western-blotted against biotinylated anti-FLAG (M2) antibody (Kodak) and alkaline phosphatase-conjugated streptavidin (Life Technologies, Inc.), then developed using the Western-Light chemiluminescent detection system (Tropix, Bedford, MA). Protein-disulfide isomerase (PDI) was detected by Western blotting with rabbit anti-PDI antibodies (a gift from Drs. R. Masaki and A. Yamamoto, Kansai Medical University, Osaka, Japan) and alkaline phosphatase-conjugated goat anti-rabbit IgG antibodies (Organon Teknika Corp). Blots were quantified using a model GS-525 Molecular Imager system (Bio-Rad).
Immunofluorescence Microscopic Analysis of PIG-L-M2S2 cells stably transfected with pMEEB-PIG-L-FLAG or pMEEB-FLAG-PIG-L were cultured on 14-mm diameter glass coverslips in 24-well plates for 24 h, washed twice with phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde in PBS for 30 min at room temperature, and permeabilized with 0.1% Triton X-100 in PBS for 1 h. Thereafter, they were incubated with Block Ace (Dainippon Pharmaceutical, Osaka, Japan) for 1 h, then stained with anti-FLAG antibody (Kodak), Rhodamineconjugated donkey anti-mouse IgG antibodies (Chemicon International, Temecula, CA), rabbit anti-PDI antibodies, and fluorescein isothiocyanate-conjugated donkey anti-rabbit IgG antibodies (Chemicon International). They were mounted on glass slides with Gel/Mount (Biomeda) and studied under a fluorescence microscope (BX50-FLA; Olympus Corp., Tokyo, Japan).
Proteinase K Protection Assay-Microsomes of M2S2 cells (equivalent to 5 ϫ 10 7 cells), stably expressing PIG-L-FLAG or FLAG-PIG-L were prepared by centrifugation of the post-nuclear supernatants at 10 5 ϫ g for 1 h at 4°C, followed by resuspension of the pellets in 1 ml of buffer containing 0.25 M sucrose and 10 mM HEPES-NaOH, pH 7.5. These microsomes were incubated with 500 g/ml proteinase K at 4°C for 30 min. The proteinase K was inactivated by phenylmethylsulfonyl fluoride (10 mM), and then the microsomes were solubilized with 1% Nonidet P-40. To confirm the susceptibility of the substrate proteins to the enzyme, the membranes were solubilized with Nonidet P-40 and incubated with proteinase K in parallel. After centrifugation at 10 5 ϫ g for 10 min, supernatants were Western blotted with anti-FLAG (M2) antibody, rabbit anti-PIG-L peptide antibodies (against amino acids 111-125) (Peptide Institute, Inc., Osaka, Japan), and anti-PDI antibodies.

Establishment of a GPI Anchor-deficient CHO Cell Line-To
establish new mutant cells, we first transfected a plasmid containing both DAF and CD59 cDNA (pDNC) into CHO-K1 cells. After selection in 600 g/ml G418, cells stably expressing DAF and CD59 were isolated by sorting twice with a cell sorter and cloned by limiting dilution. One colony, designated IIIB2A, expressed high levels of both DAF and CD59 (Fig. 1A), so we selected it for further study.
We mutagenized IIIB2A cells with ethyl methanesulfonate (100 g/ml) for 24 h. After a 48-h recovery period, cells that no longer expressed DAF and CD59 on their surface were selected using a cell sorter. This sorting was repeated once more before cloning by limiting dilution. Forty-six clones were recovered and 13 were classified by fusion with GPI anchor-deficient murine lymphoma cells of various complementation groups and mutant CHO clone G9PLAP.85. All 13 clones were of the same complementation group as G9PLAP.85, the mutant with the GPI biosynthetic defect in the deacetylation of GlcNAc-PI. The rest of the mutant clones also seemed to be of the same group, because fusion of the bulk mutant cells with G9PLAP.85 did not complement GPI biosynthesis. We termed this complementation group class L (see "Discussion"). One of the clones, designated M2S2, was selected for subsequent study because it had no residual DAF and CD59 staining when analyzed by FACS (Fig. 1B).
Expression Cloning of PIG-L cDNA-We obtained a cDNA that complements the deficiency of M2S2 cells by expression cloning. A rat C 6 glioma cDNA library constructed in a vector bearing a polyoma origin of replication was cotransfected with pcDNA-PyT, which lacks this origin of replication (pcDNA-PyT(ori Ϫ )) and screened for clones that complement DAF and CD59 expression on the M2S2 cells. We obtained three clones of essentially the same lengths and restriction profiles, which were thought to be derived from the same clone. We chose one for further studies. This clone restored the surface expression of DAF and CD59 on M2S2 cells (Fig. 1, A-C) and human placental alkaline phosphatase (PLAP) expression on G9PLAP.85 cells (Fig. 1, D-F).
To confirm that cloned cDNA complements the deficient deacetylation of GlcNAc-PI, M2S2 and G9PLAP.85 cells stably expressing the cDNA were established and the synthesis of GlcNAc-PI, GlcN-PI, and GlcN-PI(acyl) from UDP-[6-3 H]Glc-NAc was measured in lysates prepared from these cells. As shown in Fig. 2, GlcNAc-PI, GlcN-PI, and GlcN-PI(acyl) were synthesized in the wild-type cell lines, G9PLAP and IIIB2A (lanes 1 and 4, respectively), while only GlcNAc-PI was synthe-sized in the mutant type cell lines, G9PLAP.85 and M2S2 (lanes 2 and 5, respectively), indicating a deficiency in Glc-NAc-PI deacetylation. Transfectants expressing the cDNA synthesized GlcN-PI and GlcN-PI(acyl) as well as GlcNAc-PI similarly to the wild-type cells (lanes 3 and 6). Therefore, deacetylation of GlcNAc-PI was restored by expressing the cloned cDNA. Extent of deacetylation in the transfectants (lanes 3 and 6) were higher than those in the wild type cells (lanes 1 and 4), indicating that overexpression of PIG-L induced deacetylation activity higher than normal.
We thus cloned the target cDNA and termed the gene PIG-L for phosphatidylinositol glycan of complementation class L.
Characteristics of PIG-L cDNA and mRNA-The PIG-L cDNA consisted of 1903 base pairs and the longest open reading frame spanning nucleotides 391-1149, encoded a predicted protein of 252 amino acid residues (Fig. 3A) (DDBJ/EMBL/ GenBank accession No. D88364). The sequence around the initiation codon agrees well with the Kozak consensus sequence. A hydrophobicity plot revealed an amino-terminal hydrophobic sequence (Fig. 3B). This may be used as a signal peptide or alternatively, to attach to the membrane (see below).
The Northern blot analysis of PIG-L mRNA in C 6 glioma cells demonstrated a major 1.9 -1.7-kb message and a minor 4.9-kb message (Fig. 4, lane 1). The cloned PIG-L cDNA may correspond to the major message. To analyze PIG-L mRNA in wild-type and mutant CHO cells, we amplified and cloned a fragment of hamster PIG-L cDNA of 302 bp that encoded a part of hamster PIG-L protein having 88% amino acid identity with rat PIG-L protein. Using this cDNA as a probe, we analyzed wild-type IIIB2A and class L M2S2 CHO cells with Northern blotting (Fig. 4, lanes 2 and 3). The wild-type cells had 3.6-, 2.4-, 1.9-, and 1.3-kb messages (lane 2). The class L mutant lacked 3.6-, 1.9-, and 1.3-kb messages and expressed only a small amount of 2.4-kb messages (lane 3). Therefore, defective deacetylation in the mutant is due to PIG-L mRNA abnormality. Since four messages are simultaneously affected in the mutant, it is likely that they are the products of single PIG-L gene. For both C 6 glioma and CHO cells, total RNA samples did not give a signal, indicating that PIG-L mRNAs are minor components.
Comparison of PIG-L Sequence with Other Known Nucleotide and Protein Sequences-There was no overall homology between PIG-L and known deacetylases such as acetylornithine deacetylase, N-acetylglucosamine-6-phosphate deacetylase, and glucosaminoglycan N-acetylglucosaminyl N-deacetylase/ N-sulfotransferase (35)(36)(37)(38)(39)(40)(41)(42). A Basic Local Alignment Search Tool (BLAST) search (43) revealed that an S. cerevisiae open reading frame YM8021.07 (GenBank accession number: Z49704) (44) and a human EST W93799 (45) gave high homology scores with PIG-L amino acid sequence (Fig. 5). The S. cerevisiae open reading frame YM8021.07 encoded a putative 304-amino acid protein with 24% amino acid identity with PIG-L. Human EST W93799 gave a sequence of 87 amino acids that had 52% amino acid identity with a segment of PIG-L near the amino terminus. When the most similar portions were compared (56 amino acids), the identity increased to 73%. They may be yeast and human homologues of the PIG-L gene, respectively.
ER Localization and Membrane Orientation of PIG-L Protein-To localize intracellular expression site of PIG-L, its amino or carboxyl terminus was tagged by FLAG and the fusion constructs, FLAG-PIG-L and PIG-L-FLAG, respectively, were stably expressed in M2S2 cells. The fusion proteins were active as shown by complementation of the surface expression of DAF and CD59 (data not shown). The cells transfected with FLAG-PIG-L were fixed, permeabilized with Triton X-100, probed with anti-FLAG antibody, and observed using indirect immunofluorescence. Cell surface and nuclei were not stained, whereas the intracellular compartment was reticularly or uniformly stained (Fig. 6A), suggesting that the ER was stained. The staining profile of FLAG-PIG-L in red almost completely coincided with that of the ER protein PDI in green (Fig. 6B). Therefore, PIG-L protein is localized in the ER. The cells transfected with PIG-L-FLAG gave a similar result (data not shown).
To further confirm this, cells expressing PIG-L-FLAG were disrupted and fractionated into the ER, Golgi, plasma membranes, and cytoplasm by sucrose density gradient centrifugation. The fusion proteins were detected as a 30-kDa band by Western blotting against anti-FLAG antibody mainly in fractions 3 and 4, which contained the ER, indicating that PIG-L protein is an ER membrane protein (Fig. 6, C and D).
We then investigated the orientation of PIG-L protein in the  2 and 3). Samples of hamster RNA were also blotted against elongation factor 1-␣ cDNA (EF-1␣) to assess amounts of RNA used. Positions of 28 and 18 S ribosomal RNA were indicated. membrane. Cells expressing FLAG-PIG-L and those expressing PIG-L-FLAG were disrupted to obtain membrane vesicles bearing PIG-L-FLAG or FLAG-PIG-L. These vesicles were digested with proteinase K, followed by Western blotting using anti-FLAG antibody or anti-PIG-L peptide antibodies against amino acids 111-125. As shown in Fig. 7, PIG-L-FLAG and FLAG-PIG-L were not protected against proteinase K digestion when visualized using an anti-FLAG antibody (lanes 2 and 4 in upper panel), whereas ER lumenal protein PDI was protected (bottom panel). The results were similar with the anti-PIG-L peptide antibodies (middle panel). Therefore, most of the PIG-L protein is located on the cytoplasmic side of the ER.
There was no difference in size between PIG-L-FLAG and FLAG-PIG-L when visualized using an anti-FLAG antibody (Fig. 7, upper panel). This indicated that the amino-terminal hydrophobic portion was not cleaved as a signal peptide, and is rather used for attachment to the membrane. DISCUSSION In the present study, we cloned and characterized a novel gene that is involved in the second step in GPI anchor biosynthesis. Since expression cloning approach has been useful for cloning mammalian GPI anchor biosynthesis genes (46), we prepared a new GPI anchor-deficient mutant line of CHO cells that is defective in the second step and cloned a rat cDNA that complemented the defect. This mutant belonged to the same complementation group as the CHO mutant G9PLAP.85 that is also defective in the second step (10). We termed this complementation group class L and the cloned gene PIG-L.
A human K562 erythroleukemic cell line that is defective in the second step in GPI anchor synthesis has been termed a class J mutant (9). A gene corresponding to class J has not been cloned. We assigned a new complementation class for our mutants that are defective in the same step as the class J mutant. This is because class J mutant is now not available for comparison with our class L mutants. Therefore, the assignment of two complementation groups at the same step does not mean that two genes are involved in that step.
The second step in GPI anchor biosynthesis is deacetylation of the first intermediate GlcNAc-PI to generate GlcN-PI. Microsomes prepared from class L mutants accumulated the first intermediate without generating GlcN-PI, while those prepared from PIG-L-transfected mutants generated the second intermediate. Therefore, PIG-L cDNA transfection induced Gl-cNAc-PI deacetylase activity in the microsomes and the activity induced by the overexpression of PIG-L was higher than normal. Moreover, PIG-L protein was expressed on the ER membrane. Thus, PIG-L protein may be GlcNAc-PI deacetylase, although enzymatic activity is yet to be demonstrated with purified PIG-L protein.
The deacetylation of GlcNAc-PI in mammalian cell-free sys-tems has been shown to be stimulated by GTP hydrolysis (47). Defects in this reaction could, in principle, be due to mutations in either the deacetylase itself or the mediator of the GTP regulation. The PIG-L protein does not contain the main sequence elements of GTP-binding motifs as defined by Valencia et al. (48), suggesting that it does not bind this nucleotide. Microsomes from class L mutants transfected with PIG-L were capable of GlcNAc-PI deacetylation in the absence of GTP, and this activity was stimulated by this nucleotide (data not shown). While this similarity in the properties of this reaction in transfectants and wild-type cells does not conclusively exclude the possibility that PIG-L is the mediator of the GTP stimulation, it seems highly unlikely. Furthermore, if PIG-L is the deacetylase, the fact that it does not appear to bind GTP indicates that the enzyme and regulatory component are separate proteins. Two groups of deacetylases that act on N-acetylglucosamine moiety, N-acetylglucosamine-6-phosphate deacetylases (38 -40) and glucosaminoglycan N-acetylglucosaminyl N-deacetylase/N-sulfotransferase (41,42), have been cloned. PIG-L has no sequence homology to these deacetylases or to acetylornithine deacetylase (35)(36)(37). Additionally, PIG-L has no homology to proteins with known functions. Therefore, the function of PIG-L cannot be predicted from its primary structure.
PIG-L encodes a predicted protein of 252 amino acids whose molecular mass is about 30 kDa. The GlcNAc-PI deacetylase of African trypanosomes has been partially purified (49). The most highly purified preparation contained several bands in the molecular mass range 45-60 kDa. The relationship between the trypanosomal proteins and PIG-L is not clear.
The topography of GPI anchor biosynthesis on the ER membrane is not completely understood. GlcNAc-PI and GlcN-PI are oriented toward the cytoplasm on the ER because they were sensitive to PI-specific phospholipase C when intact microsomes were exposed to the enzyme (50). Two proteins involved in synthesis of GlcNAc-PI, PIG-A and PIG-H, are also oriented toward the cytoplasm (26). Thus orientation of both biosynthesis proteins and the glycolipid intermediate indicates that the first step occurs on the cytoplasmic side of the ER. Consistent with orientation of the second intermediate, most of the PIG-L protein was located on the cytoplasmic side of the ER. The second reaction, therefore, also occurs on the cytoplasmic side.
The transfer of mature GPI anchor precursors to proteins is thought to occur on the lumenal side of the ER. Therefore, some intermediate or mature GPI anchor precursors should flip into the lumenal side from the cytoplasmic side. Vidugiriene and Menon (51) analyzed the orientation of later intermediates in microsomes prepared from Trypanosoma brucei and found that they are cytoplasmically oriented; hence, mature GPI precursors may flip. In contrast, analysis of the PIG-B protein that is involved in transferring the third mannose from dolichol phosphate-mannose to an intermediate bearing two mannoses demonstrated that its functional site resides on the lumenal side (16). This is consistent with the notion that dolichol phosphatemannose serves as a mannosyl donor on the lumenal side for N-glycans and for yeast O-linked mannoses (52,53). Another observation that supports the lumenal usage of dolichol phosphate-mannose is that the Lec 35 mutant of CHO cells synthesizes dolichol phosphate-mannose but does not use it for either N-glycan or the GPI anchor (54,55). Although the exact step at which flipping occurs is yet to be determined, our present results further support the notion that it proceeds after the second step.
The GPI mannosyltransferases present in the lysates of blood stage Trypanosoma brucei transferred mannoses to GlcN-PI to generate later GPI intermediates when GlcN-PI was added as an exogenous substrate (56). However, the upstream intermediate GlcNAc-PI was a better substrate than GlcN-PI, suggesting a degree of substrate channeling via the GlcNAc-PI deacetylase enzyme to the mannosyltransferase (56). We metabolically labeled PIG-L-transfected and mocktransfected wild-type CHO cells with radioactive mannose to see if overexpression of PIG-L has any effect on later biosynthesis steps. Both cells generated similar amounts of mannosecontaining GPI intermediates. 2 It is not clear, therefore, if a similar substrate channeling exists in mammalian cells.
For expression cloning PIG-L using the CHO cell mutant, we cotransfected a cDNA library containing Pyori with pcDNA-PyT(ori Ϫ ) rather than with pcDNA-PyT(ori ϩ ) (19) for the following reasons. The PyT vector containing Pyori replicates in the cells and may interfere with replication of the cDNA plasmids. 3 In addition, when PyT(ori ϩ ) was applied, rearranged plasmids containing both the ␤-lactamase gene and the PyT gene appeared and they overcame the cDNA plasmids after repeated sorting (19). We used two cycles of sorting in PyT(ori Ϫ ) system, but we did not detect rearranged plasmids.  anti-PDI antibodies in green (B). The two proteins in the ER are almost identically co-localized. C and D, M2S2 cells stably transfected with PIG-L-FLAG were disrupted by nitrogen cavitation and centrifuged at 10 5 ϫ g for 15 min. The supernatant was fractionated further by sucrose gradient centrifugation. C, each fraction was characterized by assaying for protein content and organelle-specific marker enzyme activities: alkaline phosphodiesterase I for plasma membrane, ␣-mannosidase II for Golgi apparatus, and dolichol phosphate-mannose synthase for ER. D, detection of PIG-L-FLAG in the fraction shown in C by Western blotting with anti-FLAG antibody.  1 and 4), or proteinase K in the absence (lanes 2 and 5) or presence (lanes 3 and 6) of 1% Nonidet P-40. FLAG/ PIG-L fusion proteins were Western blotted with anti-FLAG antibody or anti-PIG-L peptide antibodies (against amino acids 111-125). PDI was detected with anti-PDI antibodies. Molecular size markers are on the right.
Thus, when CHO cells are used for cloning, we recommend origin-less polyoma large T DNA for cotransfection with a cDNA library.
CHO cells have been useful for preparing various mutants. For example, mutants corresponding to various biosynthesis steps of N-glycan have been established and used to clone responsible genes (57). There are several reaction steps in GPI anchor synthesis at which mutants have not been obtained. We used CHO cells to prepare GPI anchor-deficient mutants for these steps. It appeared, however, that only class L mutants were generated after ethyl methane sulfonate treatment. During establishment of G9PLAP.85, several mutants of class L were obtained (10), suggesting that CHO cells have a "hot spot" of mutation caused by ethyl methane sulfonate within the PIG-L gene. We plan to stably transfect CHO cells with PIG-L cDNA before mutagenesis to generate mutants for other reaction steps.