Hamster UDP-N-acetylglucosamine:dolichol-P N-acetylglucosamine-1-P transferase has multiple transmembrane spans and a critical cytosolic loop.

UDP-GlcNAc:dolichol-P GlcNAc-1-P transferase (GPT) is an endoplasmic reticulum (ER) enzyme responsible for synthesis of GlcNAc-P-P-dolichol, the committed step of dolichol-P-P-oligosaccharide synthesis. The sequence of hamster GPT predicted multiple transmembrane segments (Zhu, X., and Lehrman, M. A. (1990) J. Biol. Chem. 265, 14250-14255). GPT has also been predicted to act on the cytosolic face of the ER membrane, based on topological studies of its substrates and products. In this report we test these predictions by: (i) immunofluorescence microscopy with antibodies specific for native GPT sequences or epitope tags inserted into GPT, after selective permeabilization of the plasma membrane with digitonin; (ii) insertion of Factor Xa cleavage sites; (iii) in vitro translation of GPT; and (iv) site-directed mutagenesis. The loops between the 1st and 2nd and between the 9th and 10th predicted transmembrane spans of GPT were found to be cytosolic. In contrast, the loop between the 6th and 7th transmembrane spans, as well as the carboxyl terminus, were lumenal. Thus, hamster GPT must cross the ER membrane at least three times, consistent with previous computer-assisted predictions. There was no apparent N-glycosylation or signal sequence cleavage detected by in vitro translation. The cytosolic loop between the 9th and 10th transmembrane spans is the largest hydrophilic segment in GPT and, as judged by site-directed mutagenesis, has a number of conserved residues essential for activity. Hence, these results directly support the hypothesis that dolichol-P-P-oligosaccharide assembly is initiated in the cytosol and that a downstream intermediate must translocate to the lumenal face of the ER membrane.

Many functions have been attributed to the asparaginelinked class of oligosaccharides on glycoproteins (1), such as directly mediating specific interactions with endoplasmic reticulum (ER) 1 chaperones (2,3), and there is considerable interest in the mechanisms governing their synthesis. UDP-GlcNAc: dolichol-P GlcNAc-1-P transferase (GPT) is responsible for the synthesis of GlcNAc-P-P-dolichol, the first intermediate in the synthesis of dolichol-P-P-oligosaccharides, which serve as precursors of asparagine-linked glycans (reviewed in Refs. 4 and 5). GPT is well known as a site of action for tunicamycin (Tn), although at least one other site of action exists (6). GPT is found in all eukaryotes and has been purified from bovine mammary gland (7) and cloned from hamster cells (8,9), Saccharomyces cerevisiae (the ALG7 protein) (10,11), Schizosaccharomyces pombe (12), Leishmania (13), and mouse (14). Although computer-assisted analysis of the hamster sequence predicted that GPT might contain as many as 10 transmembrane segments (8), there has been little direct information regarding the topological structure of this critical enzyme.
GPT is of interest because it is one of a small group of glycosyltransferases that diverge from the common type II model typical of glycosyltransferases found in the Golgi apparatus. Type II glycosyltransferases have small cytosolic aminoterminal domains, single transmembrane segments, lumenal stem regions, and large carboxyl-terminal lumenal domains responsible for catalysis (15). Transferases that now appear to have alternate topological arrangements include the ALG5 protein (glucose-P-dolichol synthase) (16), which may have multiple transmembrane spans, and UDP-GlcNAc:phosphatidylinositol ␣-GlcNAc transferase, which is likely to be composed of at least three genetically defined subunits (17).
Interest in the structure of GPT also stems from current hypotheses regarding the topological orientation of dolichol-P-P-oligosaccharide synthesis. The most widely accepted model for the topology of oligosaccharide assembly dictates: (i) cytosolic assembly of GlcNAc-P-P-dolichol, mannose-P-dolichol, and glucose-P-dolichol; (ii) cytosolic conversion of GlcNAc-P-Pdolichol to Man 5 GlcNAc 2 -P-P-dolichol by nucleotide sugar donors; (iii) flipping of mannose-P-dolichol, glucose-P-dolichol, and Man 5 GlcNAc 2 -P-P-dolichol to the lumenal side of the membrane; and (iv) lumenal conversion of Man 5 GlcNAc 2 -P-P-dolichol to Glc 3 Man 9 GlcNAc 2 -P-P-dolichol with mannose-P-dolichol and glucose-P-dolichol as donors (reviewed in Refs. 5, 18, and 19).
Evidence in support of cytosolic assembly of GlcNAc-P-Pdolichol includes the observations that GPT in intact microsomes is active toward exogenous dolichol-P carried by liposomes (20), that the product of the subsequent reaction, GlcNAc 2 -P-P-dolichol, is accessible to a membrane-impermeant galactosyltransferase (21), and that inhibition of an ER UDP-GlcNAc transporter does not inhibit GPT activity in intact microsomes (21). Although these approaches provide valuable topological information, such experiments are subject to the possibility that the enzyme substrates or products under analysis may have relocated during the experiment. This could be due to the action of unknown flippases or transporters or, in the case of dolichol-P, the ability of the substrate to alter membrane structure (22) to give a misleading result. For example, a careful study involving galactosyltransferase as an impermeant probe with intact microsomes concluded that GlcNAc 2 -P-P-dolichol was lumenal, not cytosolic (23). In addition, Man 1 GlcNAc 2 -P-P-dolichol and Man 2 GlcNAc 2 -P-P-dolichol clearly have access to the lumenal space (24). Hence, studies of dolichol-P-P-oligosaccharide topology that focus on enzyme substrates and products may be affected by factors that are difficult to control.
An approach that is not subject to such limitations is to define the topology of the membrane-bound enzymes involved in dolichol-P-P-oligosaccharide synthesis and to identify the topological locations of critical residues. In this regard, GPT has received considerable attention, because it forms the first intermediate. For example, the sensitivity of GPT in intact microsomes from either rat (25) or embryonic chick (20,26) liver to proteases has been studied. GPT appears to have one or more cytosolic elements that are sensitive to trypsin and Pronase (20,26), although in some cases these enzymes failed to affect GPT (25,26), suggesting lot-to-lot variation. These data suggested that the catalytic site of GPT faced the cytosol. However, as noted (25), such studies can be difficult to interpret, because an enzyme with a lumenal catalytic site could still exhibit "cytosolic" protease sensitivity if the enzyme has an essential cytosolic domain or cytosolic accessory protein.
The ability to express recombinant GPT (27) raises the possibility of other methods of topological mapping. However, in initial studies we found that many of the conventional methods used to map the topologies of proteins with multiple membrane spans failed with hamster GPT. GPT becomes unstable after truncation (28), precluding reliable analyses of fusions with marker proteins, and yields a complex pattern of products on digestion with trypsin and other common proteases. 2 Currently there are no sensitive immunoprecipitation protocols for GPT, which could be used with sequence-specific antipeptide antibodies, and insertion of N-glycosylation sequons has yielded negative results. 3 This report describes the topological mapping of hamster GPT by insertion of diagnostic epitope tags and protease recognition sites and a mutagenic analysis of conserved residues found facing the cytosol.

EXPERIMENTAL PROCEDURES
Materials-UDP-[ 3 H]GlcNAc (26 Ci/mmol) was from DuPont. Translation-grade [ 35 S]methionine (1000 Ci/mmol) and 14 C-methylated proteins were from Amersham Corp. T7 RNA polymerase, ribonucleotide triphosphates, RNAsin, nuclease-treated rabbit reticulocyte lysate, yeast ␣-mating factor mRNA (0.1 mg/ml), and canine pancreatic microsomal membranes (2 eq/l) were from Promega. Protein A-agarose and Muta-Gene 2 kits were from Bio-Rad. Endoglycosidase H was from Genzyme. Geneticin (G418 sulfate) and powdered cell culture media were from Life Technologies, Inc. Serum was from Atlanta Biologicals. Factor Xa was from New England Biolabs. Citifluor glycerol was purchased from Ted Pella, Inc. Paraformaldehyde, trypsin, and digitonin were from Sigma. High activity soybean trypsin inhibitor was from Calbiochem. Various enzymes required for manipulation of recombinant DNA were from New England Biolabs, Fisher, or Boehringer Mannheim. Mouse monoclonal antibodies were from Eastman Chemicals (anti-DYKDDDDK peptide (FLAG) IgG, catalog number IB13010), StressGen (anti-KSEKDEL peptide (carboxyl terminus of BiP) IgM, catalog number SPA-827), or Sigma (antivimentin IgM, catalog number V-5255, a gift from Dr. George Bloom, University of Texas, Southwestern Medical Center). Fluorescein isothiocyanate-labeled antibodies used for immunofluorescence microscopy were obtained from Sigma (goat anti-mouse IgG, catalog number F-5262, and goat anti-rabbit IgG, catalog number F-9887) or Zymed (goat anti-mouse IgM, a gift of Dr. George Bloom).
Cell Culture-CHO-K1 cells were normally maintained in Ham's F-12 medium buffered at pH 7.2 with 15 mM Na-HEPES with 2% fetal bovine serum and 8% calf serum as described (29). COS-6 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum.
Transient and Stable Transfection-Stable transfection of normal GPT cDNA and various mutants in plasmid pJB20 (30) was performed by the calcium phosphate procedure as described earlier (27). Stable transfectants were selected with 1 g/ml Geneticin (for vector) or with Tn (T7765, Sigma) for GPT transfectants (27); primary selections were performed with 1 g/ml Tn, and individual colonies were screened with higher concentrations to obtain the highest expressers. Transfectants were subcloned by limiting dilution.
Transient transfection of COS-6 cells by a minor modification of the DEAE-dextran method (8) was also performed with cDNA subcloned into pJB20.
Protein concentrations were determined immediately after membrane preparation (29). The membranes were stored at 4°C and used within 48 h.
GPT assays were performed with UDP-[ 3 H]GlcNAc as described (29). Determination and Interpretation of Cellular Resistance to Tunicamycin-As described earlier (29), individual transfectants were plated in media containing various concentrations of Tn, in 2-fold increments, and stained after 7 days of incubation. The highest concentration of Tn that had no visible effects on colony size or number was taken as the Tn resistance of that transfectant. For any given Tn resistance value-transfectant combination, we have found that Tn concentrations that are 2-4fold higher cause complete inhibition of the growth of that transfectant.
Tn resistance is a linear function of GPT expression (27). In addition, some GPT mutants lack enzymatic activity but are still able to confer cellular resistance to Tn (for example, see Table I and "Discussion"), indicating that such enzymes are properly folded and probably confer resistance to Tn by a buffering effect. The endogenous GPT in CHO transfectants contributes to the total GPT activity and Tn resistance. Thus, for stable transfectants of either normal or mutated GPT, the ratio of GPT enzyme activity/Tn resistance was determined by first subtracting the contribution of the endogenous GPT as determined with vector-transfected cells. This was done as indicated in the descriptions accompanying Fig. 1 and Table I.
Selective Permeabilization of the Plasma Membrane and Indirect Immunofluorescence Confocal Microscopy-Selective permeabilization of CHO cells with digitonin was performed in a manner similar to that described earlier (31,32). Unless specified, all steps were carried out at room temperature with gentle agitation. Cells grown on coverslips at 37°C for 24 -48 h were washed three times with PBS and then fixed in 2% paraformaldehyde in PBS for 30 min. Excess paraformaldehyde was neutralized with 100 mM NH 4 Cl in TBS (20 mM Tris-Cl, 150 mM NaCl, pH 7.4) for 10 min and briefly washed three times with PBS. For selective permeabilization of the plasma membrane, the cells were incubated on ice for 12 min with 10 g/ml of digitonin in 0.3 M sucrose, 0.1 M KCl, 2.5 mM MgCl 2 , 1 mM Na 3 EDTA, and 10 mM Na-PIPES, pH 6.8. For permeabilization of plasma and intracellular membranes, the cells were treated with 0.1% Triton X-100 in PBS at room temperature for 5 min. After permeabilization by either method, cells were washed three times with TBS and incubated for 1 h in blocking buffer (TBS containing 10% goat serum and 1 mM CaCl 2 ). The solution was then replaced with blocking buffer with primary antibody (anti-BiP, 0.2 g/ml; anti-FLAG, 5 g/ml; anti-loop 1/2 (DEAE-IgG fraction; Ref. 27), 15 g/ml; or antivimentin, 1:200 dilution) and incubated for 4 h. The cells were then washed four times with TBS containing 1 mM CaCl 2 and incubated for 1 h with appropriate fluorescent secondary antibodies (anti-mouse IgM, 1:200 dilution; anti-mouse IgG, 1:300; or anti-rabbit IgG, 1:200) diluted in 5% bovine serum albumin in TBS containing 1 mM CaCl 2 . After six washes in TBS containing 1 mM CaCl 2 , the coverslips were rinsed twice in H 2 O and mounted on glass slides with Citifluor glycerol. Cells were viewed and photographed with a Bio-Rad MRC 600 laser confocal microscope as described earlier (27).
Treatment of Microsomes with Factor Xa and Trypsin-Freshly prepared microsomal membranes (29) were suspended at a concentration of 250 g/ml membrane protein in 20 l of a solution of 50 mM Tris-Cl, pH 8.0, 100 mM NaCl, 2 mM CaCl 2 , 0 or 0.2% Nonidet P-40, and 0 or 75 ng/ml Factor Xa for 1.5 h at 4 or 23°C with gentle shaking. Samples were then treated with SDS-PAGE sample buffer as described (27), supplemented with 4 M urea and 10% mercaptoethanol, incubated for 20 min at 23°C, and subjected to electrophoresis on a 13% SDS-poly-acrylamide gel. GPT was detected by immunoblotting with an antipeptide antiserum directed against residues 42-56 in loop 1/2 and visualized by enhanced chemiluminescence as described (27), except that the primary and secondary antisera were used at dilutions of 1:1,000 and 1:10,000, respectively.
Microsomes were treated with 250 units/ml trypsin (T-8253, Sigma) in place of Factor Xa in a similar manner, except that the trypsin was present for only the final 20 min of incubation, after which 150 M aprotinin, 300 M leupeptin, and 940 g/ml soybean trypsin inhibitor were added. Controls in which inhibitors preceded the trypsin demonstrated their effectiveness (data not shown). Samples were mixed with SDS-PAGE sample buffer supplemented with 10% mercaptoethanol, boiled for 5 min, and subjected to electrophoresis on a 13% SDSpolyacrylamide gel and immunoblotted as described (27). Anti-KSEK-DEL reactive proteins were detected with 0.1 g/ml primary antibody, and a peroxidase-labeled goat anti-mouse IgG (Amersham) was used as second antibody at a dilution of 1:2,000.
Cell-free Transcription-Transcription was carried out essentially as described (33). The 1.6-kilobase EcoRI-PstI fragment of hamster GPT cDNA was subcloned into pBSM13ϩ or pTZ18U and then linearized with PstI to give a transcript encoding the full-length polypeptide. Transcription was carried out with T7 RNA polymerase for 2 h at 37°C.
Cell-free Translation-Translations were performed in a nucleasetreated rabbit reticulocyte system using 17.5 l of lysate, 0.5 l of 1 mM amino acids (minus methionine), 15 Ci of [ 35 S]methionine, and 1 l of mRNA (typically 100 ng) in a final volume of 25 l. Where indicated the translation reaction was supplemented with 2 eq (1 l) of canine pancreatic microsomal membranes. Unless indicated otherwise, translations were performed for 15 min at 30°C. In some cases microsomal vesicles were isolated posttranslationally by centrifugation at 150,000 ϫ g for 10 min through a 0.5 M sucrose cushion containing 50 mM Na-HEPES, pH 7.5, and 100 mM KCl.
Characterization of Cell-free Translation Products-Immunoprecipitation of translation products was carried out in 0.5 ml of buffer A (50 mM Tris-Cl, pH 7.4, containing 0.15 M NaCl, 2 mM EDTA, and 1% (w/v) Triton X-100) with 12 l of translation mixture and 5 l of preimmune serum or antiserum directed against residues 42-56 or 398 -408 of GPT (27) for 4 h at 4°C. 50 l of a 10% suspension of protein A-agarose in PBS was then added and incubated with rotation for 1 h. Following centrifugation the protein A beads were washed three times in buffer A and once in 10 mM Tris-Cl, pH 7.4, and resuspended in 5 volume of 1 ϫ SDS-PAGE sample buffer (62.5 mM Tris-Cl, pH 6.8, containing 2% (w/v) SDS, 10% (w/v) glycerol, and 0.002% bromophenol blue). Unless stated otherwise, samples were treated with 50 mM dithiothreitol, denatured at 42°C for 30 min, and subjected to SDS-PAGE through 12.5% gels as described (27). Dried gels were exposed to Kodak X-Omat AR film or analyzed with an AMBIS radioanalytic imaging system.
Earlier studies (27) indicated that GPT was detectable by immunoblotting only if denaturation in SDS sample buffer was performed at mild temperatures, typically 42-45°C. GPT was not detected by immunoblotting if denaturation was performed at 100°C. However, 35 S-GPT translated in vitro was detected equally well with denaturation at 42°C or 100°C (data not shown). Thus, rather than affect entry of GPT into the gel or its migration, the 100°C treatment is presumed to affect a subsequent event during blotting or immunostaining.
For endoglycosidase H digestions, microsomes were isolated by centrifugation (see above) and resuspended in 0.1 M Tris-Cl, pH 8.0, containing 1% SDS and 1% ␤-mercaptoethanol and boiled for 2 min. The sample was then diluted 10-fold with 150 mM sodium citrate, pH 5.5, and incubated overnight at 37°C in the absence or presence of 1 milliunit of endoglycosidase H. After precipitation with trichloroacetic acid the samples were resuspended in SDS-PAGE sample buffer and subjected to SDS-PAGE.
Sodium carbonate extraction of translation products (20 l) was performed in 0.5 ml of sodium carbonate, pH 11.5, on ice for 15 min. Samples were then centrifuged at 150,000 ϫ g for 10 min to recover the supernatant and pellet. The pellets were re-extracted with sodium carbonate, and the combined supernatants were precipitated with 10% (w/v) trichloroacetic acid. The pellet and precipitated supernatant were resuspended in SDS-PAGE sample buffer and subjected to SDS-PAGE.
Site-directed Mutagenesis-Site-directed mutagenesis of the 1.6kilobase EcoRI-PstI GPT cDNA fragment in pTZ18U was conducted with the Muta-Gene 2 kit and various mutagenic oligonucleotides ranging from 28 to 36 nucleotides in length, with an essential modification as described previously (28). The correct orientation and sequence of each mutant was confirmed by standard DNA sequencing methods. When practical, unique restriction sites were also introduced to faciliate screening.
FLAG Epitopes-A FLAG epitope was attached to the carboxyl terminus of GPT in two steps. First, a silent mutation at Asp 407 was made by replacing nucleotide 1368 (T) with C, creating a unique AatII site (GACGTC) just before the stop codon. Next, an oligonucleotide cassette encoding the FLAG epitope followed immediately by a termination codon was made by hybridizing CGACTACAAGGACGACGATGA-CAAGTGA and AGCTTCACTTGTCATCGTCGTCCTTGTAGTCGA-CGT. This cassette had a 5Ј-AatII-cohesive end and a 3Ј-HindIII-cohesive end and was inserted between the AatII site at nucleotides 1366 -1371 and a unique HindIII site in the polylinker of pJB20 (30). This resulted in a mutant with all 408 residues of GPT followed by the FLAG epitope and a termination codon.
Unique BamHI sites for FLAG epitope insertion were introduced into loops 6/7 and 9/10 by replacement of nucleotides 792-795 (TGAC) with ATTC and 1123-1128 (CTCTCT) with GGATCC, resulting in the amino acid changes Asp 216 3 Ser and Leu 326 3 Gly, respectively. A double stranded synthetic DNA cassette with BamHI ends, encoding the FLAG epitope (Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys) flanked by 5 glycine residues on each side, was cloned in frame into the BamHI sites. Insertion mutants were identified by screening for a BpmI site introduced in the cassette by replacing the BamHI-compatible sequence GATCC with GATCT at the upstream BamHI site.

RESULTS
Transmembrane Topology of GPT Assessed by Mapping Natural and Artificial Epitopes in Loops 1/2, 6/7, and 9/10, and at the C Terminus-Hamster GPT resides in the ER membrane and has a predicted molecular weight of 46,112. We combined two microscopic techniques used previously to map the topological orientations of membrane proteins: (i) microscopy of epitope-tagged plasma membrane proteins in intact or fully permeabilized cells, using antibodies specific for the epitope tag (34,35); and (ii) microscopy of ER membrane proteins, using antibodies against native epitopes, after permeabilization of both the plasma and ER membranes with Triton X-100 or only the plasma membrane with digitonin (31,32). In the latter strategy cytosolic epitopes are detected in cells permeabilized with either digitonin or Triton X-100, but lumenal ER epitopes are detected only with Triton X-100.
A FLAG epitope tag (DYKDDDDK) was attached to the carboxyl terminus of GPT, and Gly 5 -flanked FLAG epitope tags were inserted in loops 6/7 and 9/10 (the loops flanked by predicted spans 6 and 7 and 9 and 10, respectively) of GPT. The flanking glycine polymers were added to enhance antibody accessibility (36). Each construct folded correctly and had appropriate catalytic activity, as judged by enzyme assays after COS-6 transfections (data not shown), assays after stable expression in CHO cells, and the ability to confer resistance to tunicamycin (Fig. 1). Loop 9/10 FLAG had relatively low enzyme activity (Fig. 1A) and correspondingly low expression (Fig. 1B). Although not expressed as abundantly as other FLAG constructs, the loop 9/10 FLAG protein had a ratio of enzyme activity/Tn resistance that was essentially the same as for the other FLAG constructs and wild-type GPT (Fig. 1C). Since this ratio is independent of expression level for native GPT (27), the loop 9/10 FLAG protein behaved normally. This contrasted with the ratios obtained for inactivating mutations in loop 9/10 (see below). As expected, constructs containing the FLAG insertion were somewhat less mobile in SDS gels than unmutagenized enzyme (Fig. 1B).
To demonstrate that overexpression of GPT did not cause an ER permeabilization artifact, antibodies to the carboxyl-terminal peptide (KSEKDEL) of BiP, a lumenal ER protein, and vimentin, a cytoskeletal protein, were tested with cells expressing a loop 6/7 FLAG mutant (Fig. 2). Vimentin was detected after treatment with either digitonin or Triton X-100, but proteins reactive with anti-KSEKDEL were detected only with Triton X-100.
Representative micrographs of key experiments with GPT are shown in Fig. 3. Both anti-FLAG and anti-loop 1/2 antibodies were used. By the criteria discussed above, loops 1/2 and 9/10 were found to be cytosolic, whereas loop 6/7 and the carboxyl terminus were lumenal. Thus, GPT is a polytopic enzyme with a minimum of three transmembrane spans. The orientations of loops 1/2, 6/7, and 9/10 and that of the carboxyl terminus are in full agreement with the computer-assisted predictions reported earlier (8). Clearly, the actual orientations of the other predicted loops remain to be determined (see "Discussion").
Factor Xa Cleavage Site Insertions-Due to particular interest in loop 9/10 (see below and Table I), stable transfectants were generated with Factor Xa cleavage sites (IEGR) inserted into loop 9/10 and, for comparison, loop 6/7, to independently confirm the results of Fig. 3. By the same criteria described above for the FLAG constructs, the Factor Xa constructs were judged to be folded normally (Fig. 1). Consistent with the results of FLAG insertions, only the loop 9/10 Factor Xa mutant was specifically degraded, yielding a discrete product, when intact microsomes were treated with Factor Xa at 23°C in the absence of detergent (Fig. 4A, upper panel, lane 6). Fig. 4B shows that the intact microsomes did not become permeabilized during the 23°C incubation. In this experiment an antibody directed against the carboxyl terminus of BiP was used that could recognize multiple lumenal ER proteins terminated with the sequence Lys-Asp-Glu-Leu. As shown, all detectable proteins remained resistant to trypsin unless the vesicles were permeabilized with detergent. The material that remained after trypsin treatment in the presence of detergent could be digested further by the use of higher trypsin concentrations (data not shown).
Although some resistant material could be attributed to the endogeous GPT in CHO cells (Fig. 4A, upper panel, lanes 1 and  2), it appeared that only a fraction of the presumed monomeric form of the loop 9/10 mutant was degraded by Factor Xa (Fig.   FIG. 1. 5 and 6). Interestingly, a higher molecular mass form of this mutant of approximately 66 kDa, possibly an SDS-resistant oligomer, was efficiently degraded yielding a product of approximately 62 kDa. As evidenced by multiple exposures of this and other experiments (data not shown), the corresponding oligomers of the other GPTs shown in the figure were not degraded. This suggests that the conformation or accessibility of loop 9/10 may differ depending on the oligomeric state of GPT.

Properties of GPT mutants with insertions of FLAG tags and Factor
Since the loop 6/7 mutant was not cleaved by Factor Xa in the absence of detergent (Fig. 4A, upper panel, lane 4), these results were also consistent with the lumenal orientation identified for loop 6/7 (Fig. 3). To demonstrate that loop 6/7 could, in fact, be recognized by Factor Xa, microsomes were permeabilized with detergent and treated with Factor Xa at 4°C (Fig.  4A, lower panel). Both the loop 6/7 (lane 4) and loop 9/10 (lane 6) mutants yielded discrete fragments. For this experiment it was necessary to incubate at 4°C, because at 23°C it was found that all GPTs were completely degraded by a nonspecific activity present in the microsomal preparation. We could not identify a protease inhibitor that prevented this degradation without also inhibiting Factor Xa (data not shown).
Although native GPT (46 kDa calculated and 36 kDa experimental) and the loop 9/10 Factor Xa fragment (38 kDa calculated and 30 kDa experimental) both deviated significantly from their calculated sizes, the loop 6/7 Factor Xa fragment (24 kDa calculated and 25 kDa experimental) did not. This indicates that the anomalous behavior of GPT noted earlier (27) may be due to an element between residues 213 and 336, perhaps due to formation of a compact SDS-resistant structure or excessive binding of SDS.
In Vitro Translation of Hamster GPT-To test for N-glycosylation and other covalent posttranslational modifications, in vitro transcription and translation were performed. Hamster GPT has one potential N-linked glycosylation site in loop 4/5 and three in loop 9/10 (8). However, the enzyme is not retained by a column of immobilized concanavalin A. 4 Purified bovine mammary GPT was also judged to be free of N-glycans (7), although it is not known whether the bovine enzyme contains N-glycosylation sequons.
Synthetic GPT mRNA was translated in vitro with a rabbit reticulocyte lysate in the absence or presence of canine pancreatic microsomes. Control experiments (not shown) indicated that 5Ј-capping of the mRNA did not improve translation, so uncapped mRNA was used in all experiments. A single predominant polypeptide was observed with an apparent size of 35 kDa (Fig. 5A, lane 2). As discussed above, this anomalous behavior of GPT (predicted size, 46 kDa) may be due to the sequence between residues 213 and 336. A series of smaller polypeptides of approximately 20 kDa or less, possibly due to proteolytic degradation or premature translational termina-  27) or GPT with a FLAG tag inserted into loops 6/7 (L 6/7 FLAG) or 9/10 (L 9/10 FLAG) or at the carboxyl terminus (C-term FLAG) were permeabilized with digitonin (DIG) or Triton X-100 (TX) and stained with antibodies specific for the FLAG epitope (anti-FLAG) or residues 42-56 in loop 1/2 of GPT (anti-L 1/2; Ref. 27) as described under "Experimental Procedures." Weak outlines of cells are visible in panels for which there was no specific staining. Since expression was relatively low in loop 9/10 FLAG stable transfectants, they were subjected to genomic amplification in the presence of tunicamycin for several weeks (29) prior to microscopy to enhance the immunofluorescence signal. In addition to results shown here, the anti-loop 1/2 antibody also gave a cytosolic pattern with normal GPT, 9/10 FLAG, and carboxyl-terminal FLAG (data not shown).

TABLE I Effects of mutations in cytosolic loop 9/10 on GPT activity
Transient transfectants of COS-6 cells and stable transfectants of CHO-K1 cells were isolated as described under "Experimental Procedures." The normal sequences for regions 1 and 2 are underlined, and the various mutations are listed underneath. All GPT activities were determined with intact membranes. To eliminate effects of day-to-day variations of absolute activity, activities are reported as -fold changes with respect to membranes from vector-transfected cells assayed in the same experiment. For transient transfectants (at least five determinations) the relative intensity of the GPT signal on immunoblots is given along with the GPT activity. For stable transfectants (at least six determinations) GPT activity is listed with relative resistance to tunicamycin and intensity of immunoblot signal. The normalized activity ratio (A/C) is presented as (-fold change of GPT activity Ϫ 1, the contribution of endogenous enzyme)/(tunicamycin resistance in g/ml Ϫ 0.5, the contribution of endogenous enzyme).  tion, was also usually observed. The 35-kDa protein and smaller polypeptides were absent from translations performed without added mRNA (data not shown). Translation of the intact GPT polypeptide was verified by immunoprecipation of the 35 S-labeled 35-kDa polypeptide with antibodies directed against amino acid residues 42-56 and 397-408 of the hamster cDNA sequence (Fig. 5A, lanes 3-5). Translation was complete after 15 min at 30°C in the absence or presence of microsomes (data not shown). GPT contains 11 cysteine residues, but its mobility was unaffected by reduction with dithiothreitol, demonstrating that its migration through the gel did not depend on maintainance of disulfide bonds (data not shown). Furthermore, the mobility of GPT translated in vitro was not altered by addition of microsomes (Fig. 5B, lanes 1 and 2) or endoglycosidase H digestion (Fig. 5B, lanes 3 and 4), suggesting an absence of signal sequence cleavage and N-glycosylation and consistent with a cytosolic orientation for loop 9/10. Native GPT detected on immunoblots was also resistant to treatment with endoglycosidase H (data not shown).
GPT translated in the presence of microsomes behaved as an integral membrane protein, as judged by its resistance to the effects of carbonate extraction (Fig. 5C, lanes 1-3), although GPT was readily extracted with carbonate if microsomes were added 15 min after initiation of translation (data not shown). Translation of yeast ␣-mating factor, a soluble protein normally translocated into the ER lumen, in the presence of microsomes yielded a polypeptide that was sensitive to endoglycosidase H digestion and carbonate extraction, as expected (Fig. 5, B, lanes 5-8, and C, lanes 4 -6).
Mutagenic Analysis of Loop 9/10 -Cytosolic loop 9/10 is the largest hydrophilic segment in GPT, consisting of 84 amino acid residues (residues 295-378), and is 36% identical with the corresponding segment in S. cerevisiae GPT (encoded by the ALG7 gene; Fig. 6). However, loop 9/10 does not contain conserved sequences of the UDP-GlcNAc/MurNAc family, a group of prokaryotic and eukaryotic enzymes that use UDP-GlcNAc or a related sugar donor, a polyisoprenol-P acceptor, and form a pyrophosphate bond (37). Thus, the role of loop 9/10, if any, was not clear.  3 and 7) or presence (lanes 4 and 8) of endoglycosidase H. Note that sample differences due to the endoglycosidase H protocol had a small effect on the mobility of GPT (compare lanes 2 and 3). C, effects of carbonate extraction. GPT (lanes 1-3) and ␣-mating factor (lanes 4 -6) were translated in the presence of microsomes, and total translation products were either analyzed directly by SDS-PAGE (lanes 1 and 4, T) or subjected to carbonate extraction to yield supernatant (lanes 2 and 5, S) and pellet (lanes 3 and 6, P) fractions and then analyzed by SDS-PAGE.
Specific residues in loop 9/10 were altered by site-directed mutagenesis in two highly conserved 6 -7-residue segments (Fig. 6): Pro 298 -Arg 303 (region 1) and Asn 360 -Leu 366 (region 2). Each mutant was characterized after transient expression in COS-6 cells or stable expression in CHO-K1 cells. As summarized in Table I, multiple substitutions in both regions (mutations 1, 2, and 6) resulted in essentially complete loss of activity in both expression systems. However, the mutated enzymes were readily detected on immunoblots when expressed transiently or stably and conferred resistance to Tn when expressed stably. In addition, one multiple substitution in a conserved tripeptide segment of loop 9/10 (Ile 333 -Leu 334 -Lys 335 3 Ala-Ala-Gln) also inactivated GPT in transient assays (data not shown).
Further analysis of region 2 by altering residues 363-364 (mutation 7) or only residue 366 (mutation 8) yielded enzymes with activities roughly one-half of normal. This indicated that the enzyme can reasonably tolerate some mutations in region 2. In contrast, it was found that GPT could not tolerate any mutation of Arg 303 in region 1. This region was originally chosen for more detailed analysis, since it contained a cluster of positively charged residues that could potentially interact with the phosphorylated substrates. Conservative replacement of Arg 303 with Lys (mutation 5) resulted in a loss of essentially all enzymatic activity, both in transient and stable transfections, although the mutant protein could be expressed at levels comparable with those of the normal enzyme. As expected, less conservative replacement of Arg 303 to Asn or His (mutations 3 and 4) also inactivated GPT.
Membranes from stable CHO-K1 transfectants overexpressing either normal GPT or Arg 303 3 Lys (mutation 5) were also assayed after detergent solubilization and addition of saturating dolichol-P in the presence of variable concentrations of UDP-GlcNAc. After subtracting out the activities corresponding to endogenous GPT (i.e. vector-transfected CHO-K1 cells), the mutant appeared to have a small amount of activity (approximately 10% of the normal enzyme), although it was not possible to obtain accurate enzymatic measurements because of the background of endogenous enzyme. Although the precise function of Arg 303 cannot be deduced from these data, the profound effect of the conservative Arg 303 3 Lys mutation and the effects of the other mutations in loop 9/10 demonstrate that this cytosolic loop has an essential function. DISCUSSION This article reports new information regarding the topological organization of hamster GPT. Both loop 6/7 and the car- This model is based on work from this and earlier publications. The loops are drawn to roughly correspond to their relative sizes in the amino acid sequence. The topological studies of the present study are consistent with the original predictions for membrane-spanning regions numbered 1-10 (8). Essential loop 9/10 (dashed segment), Arg 303 (ૺ), and domains with either cytosolic (large C) or lumenal (large L) locations were the subjects of the current study. The potential dolichol recognition sequences (f) (38) and the Phe-Ser-Ile sequence (ࡗ) (28) were shown previously to be essential for GPT activity. Small A-F, approximate positions of six motifs of the UDP-GlcNAc/MurNAc family (37). Preliminary data suggest that these are important for GPT activity (T. Dal Nogare, N. Dan, and M. A. Lehrman, unpublished data). Inset, alternative conformation in which predicted spans 4 and 5 (black) are membrane dips. Such a conformation would place C and D on the cytosolic side of the ER membrane. boxyl terminus have a lumenal orientation, whereas loops 1/2 and 9/10 are cytosolic. Loop 9/10 has highly conserved elements that are essential for enzyme function. This work provides a clear example of a eukaryotic glycosyltransferase with multiple transmembrane spans, in contrast to the well known type II model for Golgi apparatus transferases. This is also the first direct evidence that an enzyme involved in the initial steps of dolichol-P-P-oligosaccharide synthesis has critical elements in the cytoplasm.
The results of this and prior studies are summarized as a schematic model of GPT (Fig. 7). The topological orientations of four regions have been mapped (large letters C and L), including essential loop 9/10. One residue in loop 9/10, Arg 303 , fails to tolerate even the most conservative replacements (Table I). In addition, GPT has important elements in spans 2, 7, and 10. The sequence Phe 395 -Ser 396 -Ile 397 at the carboxyl-terminal end of span 10 plays a role in the stabilization of GPT (28). Spans 2 and 7 contain potential dolichol recognition sequences. Their actual functions are not well understood, but mutation of either causes the expression of enzyme that is inactive but still able to bind Tn (38).
With the exception of loop 1/2, for which antipeptide antibodies were already available, loops 6/7 and 9/10 and the carboxyl terminus were selected for analysis because these contain sections of poorly conserved sequence that were likely to tolerate insertion of epitope tags. Parenthetically, we also prepared a Gly 5 -FLAG insertion in loop 4/5. Although this protein was expressed transiently in COS cells and was enzymatically active, it resisted long-term stable expression in CHO-K1 cells, due to a possible dominant negative effect. Thus, this construct was not considered suitable for topological analysis. Similarly, insertion of a Factor Xa cleavage site or a N-glycosylation sequon into loop 4/5 yielded negative results, and antibodies directed against a synthetic peptide from loop 4/5 failed to react with the intact protein (data not shown).
These efforts were made because loop 4/5 is flanked by elements C and D of the UDP-GlcNAc/MurNAc family (37). Six conserved sequence elements designated A-F were identified in this family. In hamster GPT these elements are found at or near predicted membrane-water interfaces (Fig. 7) and have been suggested to be essential for substrate recognition and/or product formation (37). In the structural model elements A, B, E, and F fit this idea because they are oriented toward the cytoplasm, whereas C and D do not because they appear to be oriented toward the lumenal space. However, elements C and D could be oriented toward the cytoplasm if spans 4 and 5 were actually membrane "dips," as suggested for the M2 region of the mammalian KDEL receptor (39) and the H5 regions of potassium channels (40) (Fig. 7, inset). However, as discussed above, loop 4/5 has been refractory to the topological analyses necessary to explore this idea further.
Like GPT, yeast mannose-P-dolichol synthase is membraneassociated, appears to act on the cytosolic aspect of the ER membrane (17) and uses a nucleotide sugar donor and a dolichol-P acceptor. Yet, mannose-P-dolichol synthase probably spans the membrane once (41). Why should GPT have such a different structure? Since it is responsible for the committed step of N-linked glycosylation, GPT may be subject to acute posttranslational regulation. The ability of MPD to activate mammalian GPT in a highly specific manner in vitro is well documented (42). It has also been reported that GPT can be activated by phospholipids and other molecules in vitro and be suppressed by dolichol-linked oligosaccharides in vivo (reviewed in Ref. 5). Thus, it is possible that several of the membrane spans in GPT serve to interact with various regulators, whereas others (perhaps spans 1, 4, 5, and 8, which are associated with elements of the UDP-GlcNAc/MurNAc family) are involved in catalysis. Given that the conserved sequences of the UDP-GlcNAc/MurNAc family are at or near predicted membrane-water interfaces, it is therefore plausible that spans 1, 4, 5, and 8 come together to bring these elements into close proximity and, with loop 9/10, form a functional catalytic site.