Yeast GTB1 Encodes a Subunit of Glucosidase II Required for Glycoprotein Processing in the Endoplasmic Reticulum*

Glucosidase II is essential for sequential removal of two glucose residues from N-linked glycans during glycoprotein biogenesis in the endoplasmic reticulum. The enzyme is a heterodimer whose α-subunit contains the glycosyl hydrolase active site. The function of the β-subunit has yet to be defined, but mutations in the human gene have been linked to an autosomal dominant form of polycystic liver disease. Here we report the identification and characterization of a Saccharomyces cerevisiae gene, GTB1, encoding a polypeptide with 21% sequence similarity to the β-subunit of human glucosidase II. The Gtb1 protein was shown to be a soluble glycoprotein (96–102 kDa) localized to the endoplasmic reticulum lumen where it was present in a complex together with the yeast α-subunit homologue Gls2p. Surprisingly, we found that Δgtb1 mutant cells were specifically defective in the processing of monoglucosylated glycans. Thus, although Gls2p is sufficient for cleavage of the penultimate glucose residue, Gtb1p is essential for cleavage of the final glucose. Our data demonstrate that Gtb1p is required for normal glycoprotein biogenesis and reveal that the final two glucose-trimming steps in N-glycan processing are mechanistically distinct.

Protein N-glycosylation occurs in the endoplasmic reticulum (ER) 2 by en bloc transfer of a specific oligosaccharide (Glc 3 Man 9 GlcNAc 2 ; G3) from a dolichol donor to asparagine (Asn) residues in the sequon Asn-Xaa-(Ser/Thr) of nascent polypeptide chains. The transferred oligosaccharide is then processed in the ER by the sequential action of specialized trimming enzymes (see Fig. 1). Glucosidase I, a type II membrane protein with a lumenal catalytic domain (3) removes the outermost ␣1,2-linked glucose residue to yield the G2 form of the glycan (Glc 2 Man 9 GlcNAc 2 ), followed by the action of glucosidase II, which removes the middle and innermost ␣1,3-linked glucose residues, yielding the G1 and M9 forms, respectively (Glc 1 Man 9 GlcNAc 2 ; Man 9 GlcNAc 2 ). Finally, in a slower reaction, ER mannosidase I selectively removes a specific mannose residue, yielding the M8 form (Man 8 GlcNAc 2 ) before glycoproteins either exit the ER or are targeted for degradation if they have failed to fold correctly.
In higher eukaryotic cells, there is a close link between N-glycanprocessing events and ER quality control (for reviews, see Refs. 1 and 2). Cleavage of the middle glucose by glucosidase II generates monoglucosylated glycans that are required for productive interactions with the ER-resident chaperones calnexin and calreticulin (4 -6). Cleavage of the final glucose residue then prevents further association with calnexin/ calreticulin, allowing correctly folded proteins to proceed through the secretory pathway. In contrast, incorrectly folded glycoproteins can be reglucosylated by UDP-glucose:glycoprotein glucosyltransferase (Fig. 1,  GT), which acts as a folding sensor (7), permitting another cycle of calnexin/calreticulin interaction. This enables unfolded substrates to go through multiple rounds of interaction with the chaperones of the cycle until the native conformation is reached, when recognition by glucosidase II (but no longer by glucosyltransferase) allows exit from the cycle and the ER. If the native conformation is not achieved, the slower acting ER mannosidase I cleaves a specific terminal mannose, resulting in the targeting of the glycoprotein for degradation (1).
Glucosidase II has been extensively characterized from mammalian tissues as a soluble luminal enzyme composed of two tightly associated ␣ and ␤ glycoprotein chains (8,9). The ␣-subunit (GII␣) is a 107-kDa protein sharing sequence similarity with other glycosidase enzymes and that has a glycosyl hydrolase activity in vitro that is independent of the ␤-subunit (9). The function of the ␤-subunit (GII␤) is less well defined, but it has been suggested that it may be required to retain the heterodimer in the ER via its C-terminal HDEL motif (10 -12). More recently, it has been proposed that GII␤ interacts with N-glycans via a carbohydrate recognition domain to stimulate the GII␣ trimming of both the middle and innermost glucose residues (13). The importance of a physiological role for GII␤ is underlined by the finding that autosomal dominant polycystic liver disease can develop in individuals carrying mutations in the GII␤ gene (14 -17).
Glucosidase II has been characterized to a lesser extent in Saccharomyces cerevisiae. The GLS2 gene encodes a GII␣ homologue (Gls2p), and a ⌬gls2 null mutant is unable to process the G2 form of glycoproteins (8,18). Here we report that a previously uncharacterized open reading frame (ORF), YDR221w, encodes the yeast homologue of GII␤. This ORF, which we now refer to as GTB1 (glucosidase two ␤-subunit), expresses a soluble 96 -102 kDa glycoprotein that co-immunoprecipitates in a complex together with Gls2p. In ⌬gtb1 null mutant cells, we found the stability and localization of Gls2p to be unaltered; thus yeast Bgt1p is not required for retention of Gls2p in the ER. Surprisingly, we found that trimming of G2 to G1 was unaffected in ⌬gtb1 cells, demonstrating that Gls2p is sufficient for this reaction. In striking contrast, we found that the ⌬gtb1 cells accumulated monoglucosylated forms of N-linked glycoproteins. These results indicate that Gtb1p is specifically required for the final glucose-trimming event during normal glycoprotein processing. We propose a model in which the ␤-subunit is specifically required to present monoglucosylated substrates to the catalytic domain of the ␣-subunit.

EXPERIMENTAL PROCEDURES
Yeast Strains, Plasmids, Media, and Growth Conditions-Yeast strains were grown at 30°C with rotation in YP medium (1% yeast extract, 2% peptone) containing 20 mg/liter adenine and 2% glucose (YPAD) or in minimal medium (0.67% yeast nitrogen base) with 2% glucose and 10 mM ammonium chloride plus appropriate supplements for selective growth. Geneticin (Melford Labs, UK) was added to the YPAD medium at 200 g/ml for kanMX4 marker selection. Standard techniques of mating haploid strains, sporulation, and tetrad analysis were employed to construct double mutant strains. The genotypes of yeast strains are detailed in Table 1. Null mutant strains were obtained from the systematic knock-out series in strains BY4741 and BY4742 (20). pMP220 is a pRS315-based (CEN, LEU2) plasmid encoding a 42-kDa Pho8-Ura3p fusion protein consisting of the N-terminal 82 residues of Pho8p with Ura3p fused at the C terminus. 3 A single N-glycan acceptor site has been engineered in the Ura3p portion and a Myc epitope added at the C terminus.
Construction of Epitope-tagged Yeast Strains-The gtb1::HA-kanMX4 strain was made by genomic integration of a cassette designed to fuse the 3HA epitope at the extreme C terminus of the GTB1 ORF. The cassette was produced by PCR amplification of pFA6a-3HA-kanMX4 (21) with primers 5Ј-GACTTTCGAATACGAGCCCCCTAAGTTC-AATTTAAGTGAACGGATCCCCGGGTTAATTAA-3Ј and 5Ј-TT-GTTTACTTACTAAAAAAGCCTAACTACTCATTTTCCAGGAA-TTCGAGCTCGTTTAAAC-3Ј. This was used to transform strain BY4742 to geneticin resistance. Transformants were tested by genomic PCR amplification of GTB1 to confirm the integration producing the gtb1::HA-kanMX4 allele. The gls2::Myc-kanMX4 strain was made by genomic integration of a cassette designed to fuse the 13Myc epitope at the extreme C terminus of the GLS2 ORF. The cassette was produced by the PCR amplification of pFA6a-13Myc-kanMX4 (21) with primers 5Ј-CCTATCGCTTGACATAACTGAAGATTGGGAAGTTATTTTT-CGGATCCCCGGGTTAATTAA-3Ј and 5Ј-GAGCTAAAAAGATG-TAAAGATATCACTTCGTTTTTTTCCCGAATTCGAGCTCGTT-TAAAC-3Ј. This cassette was used to transform the haploid strain BY4741 to geneticin resistance. Transformants were tested by genomic PCR amplification of GLS2 to confirm correct integration. BWY616 (gtb1::HA) and BWY618 (gls2::Myc) were mated to create a gtb1::HA/gls2::Myc double mutant strain. The resulting diploid was sporulated and subjected to tetrad analysis. After scoring the geneticin resistance, colonies from tetrads showing a 2 resistant:2 sensitive segregation were immunoblotted for the presence of the expression of Gtb1p-HA and Gls2p-Myc. The isolate named BWY620 expressing both tagged proteins was thus identified.
Yeast Cell and Medium Extract Preparation-Yeast whole cell extracts were prepared from 5.0 A 600 units of exponentially growing cells by suspension in SDS-PAGE sample buffer containing 5% ␤-mercaptoethanol and 0.5-mm glass beads, followed by disruption at 6.5 M/s for 30 s (Hybaid Ribolyser) and incubation for 5 min at 95°C. Growth medium protein extracts were also prepared from exponentially growing cultures. Typically, 10.0 A 600 nm units of culture were subjected to centrifugation at 20,000 ϫ g for 10 min, the supernatant removed to a fresh tube, and the centrifugation repeated again. Trichloroacetic acid was added to the resulting cell-free medium to a final concentration of 15%, with incubation on ice for 30 min. Precipitated material was pel-   leted by centrifugation at 20,000 ϫ g for 20 min. The pellet was washed twice with ice-cold acetone and then resuspended in 1% SDS, 62 mM Tris, pH 6.8, and 1 mM EDTA at 95°C. For endoglycosidase H digestion, whole cell or growth medium extracts prepared in 1% SDS, 62 mM Tris, pH 6.8, and 1 mM EDTA were diluted 10-fold into 50 mM sodium citrate buffer, pH 5.5, containing 0.5 mM 4-(2-aminoethyl)-benzenesulfonylfluoride HCl and 10 mM dithiothreitol. Mock or endoglycosidase H (2 milliunits/A 600 nm equivalents) digestion was carried out at 37°C for 2 h. Reactions were trichloroacetic acid-precipitated as described above and resuspended in SDS-polyacrylamide gel sample buffer at 95°C.
Membrane Extraction-The membrane association of Gtb1p-HA and Gls2p-Myc were analyzed by the extraction of microsomes with reagents that discriminate between integral, peripheral, and soluble proteins. For each extraction, 2.5 A 280 nm units of microsomes prepared as previously described (25) were harvested at 17,000 ϫ g for 10 min and resuspended in 100 l of membrane storage buffer containing the extracting reagent, except for carbonate extraction, when microsomes were resuspended with vortexing directly into 100 l of 0.1 M Na 2 CO 3 (pH 11.0). Extractions were incubated on ice for 30 min before centrifugation at 100,000 ϫ g for 30 min. The supernatant fractions were subjected to trichloroacetic acid precipitation, and all fractions were heated in SDS-polyacrylamide gel sample buffer at 95°C.
Native Immunoprecipitation-For native immunoprecipitation, 2.5 A 280 nm units of microsomes, prepared as previously described (25), were harvested at 17,000 ϫ g for 15 min at 4°C and resuspended with vortexing in 250 l of solubilization buffer (20 mM Tris-HCl, pH 7.4, 5 mM magnesium(OAc) 2 , 10 g/ml leupeptin, 5 g/ml chymostatin/pepstatin, 1 mM 4-(2-aminoethyl)-benzenesulfonyl-fluoride HCl, and 12% glycerol) containing 1% Triton X-100 and 200 mM NaCl. The mixture was incubated on ice for 30 min prior to centrifugation at 10,000 ϫ g for 10 min. For anti-HA precipitation, 25 l of anti-HA affinity matrix (rat monoclonal antibody, clone 3F10, Roche Applied Science) washed in solubilization buffer was added to the supernatant followed by incubation at 4°C for 2 h. Antigens were dissociated from the beads by the addition of SDS-PAGE sample buffer containing 5% ␤-mercaptoethanol and heating at 95°C for 5 min.
Radiolabeling and Denaturing Immunoprecipitation-For [ 35 S]methionine labeling, cells were grown in minimal medium, and cell extracts were prepared as previously described (24). 5.0 A 600 nm equivalents of cell extract were added to saturating amounts of antiserum followed by rotation for 2 h at room temperature. Immune complexes were collected on protein A-Sepharose CL4B beads (Sigma) at room temperature for 2 h. The beads were pelleted in a microcentrifuge and washed three times with 1 ml of immunoprecipitation buffer. Antigens were dissociated from beads by the addition of SDS-PAGE sample buffer containing 5% ␤-mercaptoethanol and heating to 95°C for 5 min. Radiolabeled samples were processed by 7.5% SDS-PAGE and visualized by phosphorimaging (Fuji FLA 3000) before exposure to x-ray film (Kodak). For Lhs1p and DPAP B immunoprecipitations, 5.0 A 600 nm equivalents were loaded per lane and 2.0 equivalents for CPY.
Analysis of N-Glycans-N-glycans labeled with [ 3 H]mannose for 4 min were prepared from yeast strains and analyzed as described by Jakob et al. (18). 20 A 600 nm equivalents of [ 3 H]mannose-labeled N-glycan were run on a Supelcosil LC-NH 2 (240 ϫ 4.6 mm) column in acetonitrile:water (70:30, v:v), at a flow rate of 1 ml/min using an Akta purifier (Amersham Biosciences). Samples (1.0 ml) were collected and analyzed by scintillation counting.
Immunofluorescence-BWY625 cells transformed with pAC54 (SEC61-GFP) were prepared for immunofluorescence analysis essentially as described by Craven et al. (26). Briefly, cultures grown in minimal medium to A 600 nm of 0.4 with selection for pAC54 were harvested and fixed in 5% formaldehyde at room temperature. The cells were spheroplasted by treatment with yeast lytic enzyme (ICN), and anti-Myc antibody (1:400; Sigma) staining was carried out for 1 h before visualization with Cy3-conjugated antibody (1:400; Sigma). Stained cells were incubated with 0.1 mg/ml 4Ј,6-diamidino-2-phenylindole for 5 min. Images were captured with a Zeiss Axiophot microscope at 2500ϫ magnification. Plasmid pAC54 expresses the N-terminal 475 residues of Sec61p fused to GFP. A 2.5-kb HindIII fragment containing the SEC61 gene with a unique BamHI site, as described in the construction of a plasmid expressing the N-terminal 475 residues of Sec61p fused to invertase (27), was cloned into HindIII-digested pRS425 (2 m, LEU2). The GFP gene was derived from pS65T as an EcoRI fragment cloned into pUC118 and then ligated as a BamHI fragment into the SEC61 vector to create pAC54. This plasmid complements the lethal ⌬sec61 mutation (data not shown).

GTB1 Encodes the Yeast Homologue of GII␤-There is a single
uncharacterized yeast open reading frame, YDR221w, sharing significant (21%) sequence identity with human GII␤ (11) (Fig. 2). This ORF, which we now refer to as GTB1, would encode a 702-residue polypeptide with a predicted N-terminal signal sequence and six consensus sites for N-glycosylation. However, it lacks a C-terminal ER retrieval motif (-XDEL COOH ), which is a general feature of other known GII␤ sequences. The highest degree of sequence similarity is found in the N-terminal region corresponding to residues Gln 64 -Cys 84 of the S. cerevisiae protein (Fig. 2). The equivalent region of the human protein has been shown to be involved in binding GII␣ and has significant similarity to a calcium-dependent EF-hand (28). Another conserved region, Asn 100 -Cys 148 , has homology with C-type lectin domains (Fig. 2), suggesting a role in calcium-dependent carbohydrate recognition. In addition, there is another region toward the C terminus of the molecule with 27% identity to the mannose-6-phosphate receptor homology domain, including three completely conserved residues that have been implicated in mannose binding (Fig. 2) (29).
Gtb1p Is Required for N-Glycan Processing in Vivo-To examine a possible role for Gtb1p in N-glycan trimming, we analyzed the biogenesis of glycoproteins in null mutant cells. Wild-type, ⌬gtb1, and ⌬gls2 cells were labeled for 4 min with [ 35 S]methionine/cysteine in minimal medium at 30°C. Cell extracts were subjected to immunoprecipitation with antisera directed against three different glycoproteins, namely Lhs1p, CPY, and DPAP B. As expected, ⌬gls2 mutant cells accumulated glycoproteins that migrated more slowly than those from wild-type cells due to their failure to remove two glucose residues from N-glycans (Fig.  3A, lanes 1 and 5). Glycoproteins from ⌬gtb1 mutant cells also migrated more slowly than those from wild-type cells but faster than those from the ⌬gls2 mutant (Fig. 3A, lane 3). This indicated a partial defect in glucose trimming in the ⌬gtb1 mutant but not the complete block in glucosidase II function seen with the ⌬gls2 mutant. Treatment with The Function of ER Glucosidase II ␤-Subunit MARCH 10, 2006 • VOLUME 281 • NUMBER 10 tunicamycin led to the accumulation of unglycosylated species of the same size in all of the strains, thus confirming that the decreased gel mobility of glycoproteins in the mutants was because of a defect in N-glycan processing. The most obvious explanation for the phenotype of ⌬gtb1 mutant cells is that the middle glucose is trimmed normally but that there is a defect in the trimming of the final glucose residue.
The trimming of glucose residues was analyzed further using a genetic approach to manipulate the glucosylation state of N-glycans. Cells carrying the ⌬alg8 mutation synthesize an incomplete core sugar that is equivalent to the G1 monoglucosylated glycan (30). This structure would normally be trimmed by glucosidase II but accumulates in ⌬gls2/⌬alg8 double mutant cells (31). It follows that, if GTB1 is required for the trimming of the final glucose, then a ⌬gtb1/⌬alg8 strain should have the same glucosylation phenotype as the ⌬gls2/⌬alg8 mutant. The double mutant strains were constructed and glucose trimming analyzed by radiolabeling cells and immunoprecipitation with antisera against CPY. As previously reported, the transfer of the incomplete cores onto proteins in ⌬alg8 cells is inefficient, leading to detection of a range of glycoforms with fewer N-glycans (30). This is evident in the two faster migrating forms of CPY detected in ⌬alg8 cells compared with wild type (Fig. 3B, lanes 1 and 2). As expected, all glycoforms of CPY were larger in the ⌬gls2/⌬alg8 double mutant compared with ⌬alg8 alone, indicating the accumulation of monoglucosylated glycans (Fig. 3B, lane 3). Significantly, the forms of CPY detected in the ⌬gtb1/⌬alg8 strain (Fig. 3B,  lane 4) were identical in size to the ⌬gls2/⌬alg8 strain, indicating that both Gls2p and Gtb1p are required to trim the final glucose in N-linked glycans.
Our interpretation of the trimming defect in ⌬gtb1 cells has thus far considered only the trimming of glucose residues. However, a change in the rate of mannosidase I trimming must also be considered. This is particularly relevant when one considers the possibility that the gel migration assays above might not discriminate between Glc 1 Man 9 and Glc 2 Man 8 glycoforms. To address this issue, we used the ⌬mns1 mutation to block mannose trimming, thus ensuring that all N-glycans retain a full complement of nine mannose residues (18). The ⌬mns1 allele was combined with either ⌬gtb1 or ⌬gls2 and then glycoproteins analyzed as done previously. We observed a small but reproducible increase in the relative molecular weight of Lhs1p from ⌬mns1 cells compared with wild type, consistent with the expected defect in mannose trimming (Fig. 3C). The ⌬gls2/⌬mns1 double mutant accumulates Glc 2 Man 9 glycans (18), resulting in a substantially reduced gel mobility compared with the glucose-trimmed Man 9 (M9) control from ⌬mns1 (Fig. 3C,  compare lanes 2 and 4). Crucially, Lhs1p from the ⌬gtb1/⌬mns1 double mutant has an intermediate mobility, from which we must conclude that it represents monoglucosylated glycans. Similar patterns of migration were observed for other glycoproteins tested, namely DPAP B and CPY (not shown). These findings were further confirmed by HPLC analysis of isolated oligosaccharides prepared from total glycoprotein extracts (Fig. 3D). In this analysis, column retention time increases with the size of the N-glycan structure (18). The [ 3 H]mannose-labeled N-glycans derived from ⌬gtb1/⌬mns1 cells were found to be retained for a shorter time compared with those from ⌬gls2/⌬mns1 cells (Fig. 3D). From these results, we can conclude that ⌬bgt1 mutant cells accumulate monoglucosylated N-linked glycans. The primary protein sequence predicted from the S. cerevisiae (Sc) gene GTB1 (YDR221w) was aligned with the GII␤ sequences from Homo sapiens (Hs) and Schizosaccharomyces pombe (Sp). Alignments were performed using the ClustalX program, with black shading representing 100% amino acid conservation and gray representing 67% conservation. The boxed regions indicate conserved regions with specific motifs. Residues Gln 64 -Cys 84 of the S. cerevisiae sequence shares the highest degree of homology and corresponds to a putative EF-hand. The other boxed regions correspond to a C-type lectin domain (Sc, residues Asn 100 -Cys 147 ) and a mannose-6-phosphate receptor homology domain (Sc, residues Ser 539 -Glu 696 ). Three conserved residues present in this domain, which are implicated in mannose binding, are indicated by filled circles. Overall, the human primary sequence shares 21 and 22% identity with the S. cerevisiae and S. pombe sequences, respectively.
In these experiments, short pulse labeling times were used; therefore, we next examined the fate of labeled proteins following an extended chase period. Wild-type and ⌬gtb1 cells were pulse-labeled and samples analyzed during a 40-min chase period (Fig. 3E). It was found that untrimmed Lhs1p did not chase to the completely trimmed form observed in wild-type cells. These data suggest a complete block in glucose trimming.
It has been reported that more than one glycan must be added to polypeptides to promote efficient glucose trimming by mammalian glucosidase II (13). We therefore tested whether this might also be the case in yeast. To do this, we examined the processing of a Pho8-Ura3p fusion protein that is targeted to the ER by the Pho8p signal anchor domain and that has a single N-glycan acceptor site. 3 A low copy plasmid expressing this protein was transformed into the wild-type, ⌬gtb1, and ⌬gls2 strains and then the fusion protein analyzed after a short pulse labeling as before (Fig. 3F). The protein expressed in wild-type cells migrated more rapidly than that from either of the mutants, demonstrating that a monoglycosylated protein is trimmed efficiently by glucosidase II in the yeast ER.
Expression of Gtb1p and Gls2p-We next examined the expression of both Gls2p and Gtb1p by engineering epitope-tagged genomic alleles. In the absence of any obvious C-terminal ER retrieval sequence, three contiguous copies of the hemagglutinin epitope were added after the FIGURE 3. Gtb1p is required for the completion of glycoprotein glucose trimming. A, wild-type (WT) and deletion mutant strains were grown in minimal medium at 30°C and labeled with [ 35 S]methionine/cysteine for 4 min. Tunicamycin was added to 10 mg/ml for 1 h before labeling. Immunoprecipitations were carried out on cell extracts using DPAP B-, CPY-, or Lhs1p-specific antisera. B, wild-type and ⌬alg8 single and double mutant strains were grown and radiolabeled as described for A, and cell extracts were subjected to immunoprecipitation with CPY-specific antiserum. The bands indicated by stars in lane 2 (⌬alg8) correspond to p1CPY species containing (slowest to fastest migrating) 4, 3, and 2 glucose-trimmed N-glycans, respectively. The bands indicated by stars in lane 3 (⌬alg8/⌬gls2) correspond to species with the same number of glycans but each larger due to the failure to trim the final glucose, a pattern also evident in lane 4 (⌬alg8/⌬gtb1). C, single and double mutant strains were grown and radiolabeled as described for A, and cell extracts were immunoprecipitated with Lhs1p-specific antiserum. D, N-glycans prepared from ⌬bgt1/⌬mns1 (empty circles) and ⌬gls2/⌬mns1 (filled circles) after a 4-min labeling with [ 3 H]mannose were examined by HPLC (see "Experimental Procedures"). The positions of N-glycan signals are indicated. The signal at 7 min corresponds to monosaccharide and is probably due to the presence of free [ 3 H]mannose carried over from the initial labeling. E, wild-type (BY4742) and ⌬gtb1 strains were grown in minimal medium at 30°C and labeled with [ 35 S]methionine/cysteine for 4 min. A cold chase was initiated by the addition of 2 mM each methionine and cysteine, and 5.0 A 600 nm equivalents of cells were removed at 10-min time points. Extracts were prepared and subjected to immunoprecipitation with Lhs1p-specific antiserum. F, trimming of a protein containing a single N-linked glycan. Wild-type, ⌬bgt1, and ⌬gls2 cells transformed with a low copy plasmid expressing a 42-kDa Pho8-Ura3p protein containing a single N-glycan acceptor site were grown at 30°C in minimal medium with plasmid selection and labeled with [ 35 S]methionine/cysteine for 4 min. The fusion protein was immunoprecipitated from cell extracts with Myc-specific antibodies, and 2.0 A 600 nm equivalents were analyzed by 10% SDS-PAGE. The expression of a small amount of non-glycosylated protein is indicated by an asterisk.
C-terminal residue (Glu 702 ) of Gtb1p (see "Experimental Procedures"). Expression of Gtb1p-HA was confirmed in total cell extracts from one transformant (BWY616) by immunoblot analysis with 12CA5 monoclonal antibodies. A diffuse band with a relative molecular mass of 96 -102 kDa was detected (Fig. 4A, lane 1), which shifted to a discrete band of 89 kDa upon digestion with EndoH (Fig. 4A, lane 2). These results indicate that Gtb1p-HA is expressed under normal growth conditions and that it is itself N-glycosylated.
A similar strategy was employed to generate a Myc-tagged allele of GLS2. Once again, the epitope was added at the extreme C terminus (Phe 954 ) of the genomic locus (see "Experimental Procedures"). This produced strain BWY618 in which Gls2p-Myc expression was confirmed by immunoblot analysis of a total cell protein extract using 9E10 antibodies. This identified a band with a relative molecular mass of 175 kDa (Fig. 4A, lane 3), which shifted to 148 kDa upon treatment with EndoH (Fig. 4A, lane 4), indicating that Gls2-Myc is also N-glycosylated. Gls2p contains eleven potential sites for N-glycosylation, and the substantial EndoH shift suggests that most, if not all, of these sites are modified in the ER. The primary sequence of Gls2p predicts a 110-kDa polypeptide chain, but after taking signal sequence cleavage (ϳ2 kDa) and the addition of the Myc tag (ϩ21 kDa) into account, Gls2p-Myc is expected to have a molecular mass of ϳ130 kDa before post-translational modification. The discrepancy between this and the detected endoglycosidase H deglycosylated mass of 148 kDa could be the result of O-glycosylation, which has not been examined here.
The glucose trimming activity of strains bearing tagged alleles of GTB1 or GLS2 were found to be indistinguishable from wild-type cells (Fig. 4B). Thus, we conclude that the C-terminal epitope tags did not interfere with the glucose trimming function of either protein.
Gtb1p and Gls2p Are in a Soluble ER Complex-The sequences of both Gtb1p and Gls2p predicted an N-terminal signal sequence whose cleavage would lead to an ER lumenal disposition. This was confirmed in membrane fractionation studies using microsomes prepared from tagged strains. Neither Gtb1p-HA (Fig. 5A) nor Gls2p-Myc (Fig. 5B) were released from membranes by treatment with either 0.5 M NaCl or 2.5 M urea, but both were solubilized by treatment with 0.1 M Na 2 CO 3 (pH 11). This profile mirrored that of the ER lumenal protein Kar2p (Fig.  5, A and B) (32), and therefore we conclude that both Gtb1p-HA and Gls2p-Myc are themselves lumenal proteins. The integral membrane protein Sec63p (33) also served as a control in this analysis.
The ␣and ␤-subunits of mammalian glucosidase II have been shown to form a 1:1 heterodimer (9). To test any association between the yeast counterparts, we made the haploid strain BWY620, which combined the tagged alleles of gtb1-HA and gls2-myc (see "Experimental Procedures"). Microsomes prepared from BWY620 or from singly tagged control strains were then solubilized in 1% Triton X-100 prior to immu-  1 and 2) and expression of Gls2p-Myc detected with 9E10 anti-Myc antibodies (lanes 3  and 4). B, wild-type (WT), ⌬gls2, ⌬gtb1, BWY616 (gtb1-HA), and BWY618 (gls2-myc) cells were grown in minimal medium and radiolabeled with [ 35 S]methionine/cysteine for 4 min, and cell extracts were immunoprecipitated with DPAP B-specific antiserum. FIGURE 5. Gtb1p and Gls2p are soluble ER proteins that can be co-precipitated. A, microsomes prepared from strain BWY616 (Gtb1p-HA) were treated with membrane storage buffer, 0.1 M Na 2 CO 3 , or buffer containing 0.5 M NaCl, 2.5 M urea, or 1% Triton X-100, as described under "Experimental Procedures." After 7.5% SDS-PAGE of supernatant (S) and pellet (P) fractions, immunoblotting was carried with antibodies against the HA epitope, the soluble luminal ER protein, Kar2p, and the integral ER membrane protein Sec63p. B, microsomes prepared from strain BWY618 (Gls2p-Myc) were processed and analyzed as described for A, except with immunoblotting for the Myc epitope. C, microsomes prepared from BWY616 (Gtb1p-HA, lane 2), BWY618 (Gls2p-Myc, lane 3), and BWY620 (Gtb1p-HA/Gls2p-Myc, lane 4) were solubilized in 1% Triton X-100 and incubated with an anti-HA affinity resin, as described under "Experimental Procedures." Immunoprecipitates were analyzed by 7.5% SDS-PAGE and immunoblotted with 12CA5 and 9E10 antibodies to detect Gtb1p-HA (upper panel) and Gls2p-Myc (lower panel), respectively. The total (lane 1) was provided by an equivalent loading of solubilized BWY620 microsomes.
noprecipitation with an anti-HA matrix. Immunoprecipitates were then analyzed by SDS-PAGE and immunoblotting with either anti-Myc (9E10) or anti-HA (12CA5) antibodies. As expected, Gtb1p-HA was immunoprecipitated from extracts derived from Gtb1p-HA-and Gtb1p-HA/Gls2p-Myc-expressing strains but not from the control strain expressing tagged Gls2p-Myc alone (Fig. 5C, upper panel). Immunoblotting with 9E10 antibodies revealed that Gls2p-Myc co-immunoprecipitated from extracts containing Gtb1p-HA but not from control extracts (Fig. 5C, lower panel, lane 4). In the reciprocal experiment, Gtb1p-HA was also co-immunoprecipitated from BWY620 membranes with 9E10 antibodies (data not shown). These results demonstrated that Gtb1p-HA and Gls2p-Myc form a stable protein complex.
Gls2p Localization Is Unaffected in ⌬gtb1 Cells-It has been suggested that GII␤ may contribute to the stability and/or ER retention of GII␣ in mammalian cells (11,12). We therefore further examined the expression of Gls2p in ⌬gtb1 cells. Defects in the retention of other ER luminal proteins results in their transport to the Golgi and subsequent secretion to the medium (34). Therefore, total cell and growth media extracts were prepared from BWY619 (gls2-myc) and BWY625 (⌬gtb1/ gls2-myc) and analyzed by immunoblotting. Because yeast glycoproteins may become extensively modified in transit through the Golgi (35), we also treated samples with EndoH to collapse any heterogeneous species. First, the level of Gls2p-Myc detected in cell extracts was not reduced in the absence of Gtb1p (Fig. 6A compare lanes 1 and 2 with 5 and 6), and we found no evidence of Gls2p-Myc in the culture medium from either strain (Fig. 6A). Moreover, EndoH treatment of whole cell extracts produced no evidence of any significant pool of Golgi-modified Gls2p-Myc in either wild-type or ⌬gtb1 cells (Fig. 6A, compare lanes 1  and 2 and lanes 5 and 6). Finally, the localization of Gls2p-Myc was confirmed by immunofluorescence. The Myc-tagged protein exhibited strong perinuclear staining with further staining at or close to the cell periphery (Fig. 6B). This same pattern was also seen with the direct fluorescence of Sec61-GFP in the same cells (Fig. 6B). These results demonstrated that Gls2p-Myc remains ER-localized independently of Gtb1p. A misfolded protein in the ER lumen would be expected to become subject to ER-associated degradation (36). We therefore examined whether Gls2p-Myc might be more rapidly degraded in the absence of Gtb1p. Our data indicated no significant difference in the turnover of Gls2p-Myc in ⌬gtb1 mutant cells (Fig. 6, C and D). Taken together, these data demonstrate that Gtb1p is not required for either the stability or ER retention of Gls2p-Myc. All of these data are consistent with the efficient trimming of G2 glycans observed in ⌬gtb1 cells (Fig. 3A), which provides compelling evidence of a functional Gls2p. . Gls2p does not require Gtb1p for its stable expression and its ER retention. A, analysis of Gls2p secretion into the growth medium. Wild-type or ⌬gtb1 cultures were harvested as indicated, and then whole cell protein extracts (lanes 1, 2, 5, and 6) and trichloroacetic acid-precipitated growth medium (lanes 3, 4, 7 and 8) were subjected to either endoglycosidase H digestion or mock treatment, as indicated. 1.0 A 600 nm equivalent of whole cell extracts and 5.0 A 600 nm equivalents of growth media were analyzed by 7.5% SDS-PAGE and immunoblotting with 9E10 antibodies to detect Gls2p-Myc. B, the localization of Gls2p in the absence of Gtb1p. BWY625 cells expressing Sec61-GFP as an ER-localized control were stained with 9E10 antibodies followed by Cy3-conjugated secondary antibodies and 4Ј,6-diamidino-2-phenylindole staining. Cy3 (anti-Myc), GFP, and 4Ј,6-diamidino-2-phenylindole (DAPI) fluorescence are shown in separate panels. C, the stability of Gls2p was examined by pulse-chase analysis of strains BWY619 (gls2-myc) and BWY625 (⌬gtb1/ gls-myc). Cells were labeled with [ 35 S]methionine/cysteine for 10 min with a chase initiated by the addition of methionine and cysteine to a concentration of 2 mM. Four A 600 nm equivalents of cells were harvested at each of the indicated time points and cell extracts prepared and subjected to anti-Myc denaturing immunoprecipitation. Immunoprecipitates were analyzed by 7.5% SDS-PAGE. D, The decay of Gls2p-Myc as visualized in B above was quantified and plotted against time. WT, wild-type.