The Saccharomyces cerevisiae Oligosaccharyltransferase Is a Protein Complex Composed of Wbplp, Swplp, and Four Additional Polypeptides*

Asparagine-linked glycosylation of proteins in the lu- men of the endoplasmic reticulum is catalyzed by the oligosaccharyltransferase. Previously, the mammalian oligosaccharyltransferase was shown to co-purify with a protein complex consisting of three integral membrane proteins: ribophorin I and ribophorin I1 and a nonglycosylated 48-kDa polypeptide designated OST48. Here, we describe the purification of the oligosaccharyltrans- ferase from Saccharomyces cerevisiae. The yeast oligosaccharyltransferase complex is composed of six subunits (a, p, y, S, E, and 0. The a subunit of the yeast oligosaccharyltransferase complex is a heterogeneously glycosylated protein with three glycoforms of 64,62, and 60 kDa that contain, respectively, four, three, and two asparagine-linked oligosaccharide chains. The p and S subunits were shown to correspond to the 45-kDa Wbpl glycoprotein and the 30-kDa Swpl protein, respectively. The Wbpl and Swpl proteins were previously shown to be essential for asparagine-linked glycosylation in vivo. The nonglycosylated y, E, and 5 subunits have apparent molecular masses of 34, 16, and 9 kDa. Homology between the yeast and mammalian oligosaccharyltrans-

* This work was supported by National Institutes of Health Grant PHs GM 43768 (to R. G . ) and was conducted during the tenure of an Established Investigatorship of the American Heart Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. nescence; HPLC, high pressure liquid chromatography; OST, oligosaccharyltransferase; OS-PP-Dol, dolichol-linked oligosaccharide; PMSF, phenylmethylsulfonyl fluoride; PVDF, polyvinylidene difluoride; 3':5'-ADP-agarose, 3':5'-bis,bi~-adenosine diphosphate agarose; PAGE, polyacrylamide gel electrophoresis. phate in a series of reactions that is initiated on the cytoplasmic face of the RER membrane and is completed within the RER lumen (Hirschberg and Snider, 1987;Kukuruzinska et al., 19871. The preferred lipid-linked oligosaccharide donor for the glycosylation reaction is the glucosylated high mannose oligosaccharide (Glc,Man,GlcNAc,-PP-Dol) (Turco et al., 1977;"rimble et al., 1980). However, incompletely assembled lipidlinked oligosaccharides can serve as the donor i n vitro and i n vivo (Sharma et al., 1981;Huffaker and Robbins, 1983;Verostek et al., 1991Verostek et al., , 1993.
Although a purification of the oligosaccharyltransferase from Saccharomyces cerevisiae has not been reported, biochemical studies of the detergent-solubilized enzyme have revealed considerable similarity between the yeast and mammalian enzymes with respect to substrate specificity and divalent metal ion requirement (Trimble et al., 1980;Sharma et al., 1981). Remarkable progress toward the identification of the yeast oligosaccharyltransferase complex has been achieved by the analysis of a protein designated as wheat germ binding protein 1 (Wbpl) (te Heesen et al., 1991(te Heesen et al., ,1992. The W s P l geneencodes a 45-kDa integral membrane protein that is essential for vegetative growth of yeast and is localized to the yeast endoplasmic reticulum ( t e Heesen et al. , 1991). Phenotypic analysis of yeast strains bearing a conditional mutation in the WBPl gene (wbpl-1 and wbpl-2) has shown that the Wbpl protein is required for asparagine-linked glycosylation of proteins i n vivo and for oligosaccharide transfer to acceptor peptides in vitro (te Heesen et al., 1992). A second gene (SWP1) encoding a 30-kDa polypeptide was identified as an allele-specific high copy suppressor of the wbpl-2 allele (te Heesen et al., 1993). Gene product depletion experiments indicate that the Swpl protein is also required for expression of oligosaccharyltransferase activity in yeast (te Heesen et al., 1993). Overexpression of the WBPl protein and the SWPl protein, either alone or in combination, did not enhance the oligosaccharyltransferase activity of yeast microsomal membrane preparations (te Heesen et al., 1992(te Heesen et al., , 1993 suggesting that an additional protein or proteins are required for oligosaccharyltransferase activity. Oligosaccharyltransferase activity from canine pancreas copurifies with a relatively abundant protein complex consisting of ribophorin I (M, = 66,0001, ribophorin I1 (M, = 63,0001, and OST48 (M, = 48,000) . The ribophorins are well characterized integral membrane glycoproteins that are restricted to the rough endoplasmic reticulum Marcantonio et al., 1984). Protein sequence analysis and protease accessibility studies indicate that ribophorin I, ribophorin 11, and OST48 are all integral membrane proteins with the bulk of each polypeptide located within the lumen of the endoplasmic reticulum (Crimaudo et al., 1987;Harnik-Ort et al., 1987;Silberstein et al., 1992). Due to the extensive conservation of the donor and acceptor substrates for the oligosac-charyltransferase, it would be surprising if the yeast and mammalian enzymes were not related despite the evolutionary distance between mammals and S. cereuisiae. Indeed, a comparison of the protein sequence of OST48 and Wbplp has revealed that these two polypeptides are 25% identical in sequence (Silberstein et al., 1992). In contrast, the sequence of the 30-kDa Swplp has been reported to be unrelated to both ribophorins I and I1 (te Heesen et al., 1993). Here, we report the purification of the yeast oligosaccharyltransferase as a complex composed of six subunits. An initial characterization of the six subunits of the yeast oligosaccharyltransferase shows that two of the subunits correspond to the Wbpl and Swpl proteins. A protein sequence comparison indicates that the yeast Swpl protein is related to the carboxyl-terminal half of mammalian ribophorin I1 with respect to amino acid sequence and organization of hydrophobic segments.
The yeast from a 7.6-liter culture was harvested by centrifugation for 5 min at 3,000 rpm in a Sorvall HG4L rotor. The cells were resuspended in 500 ml of 20 m~ Tris-C1, pH 7.4, 100 mM NaCl and centrifuged for 5 min at 3,500 rpm in a Sorvall HS-4 rotor. The washed cells (-94 g wet weight) were suspended in 150 ml ofbufferA (16% (w/w) sucrose, 20 m~ Tris-C1, pH 7.4, 5 m~ MgCl,, 1 m~ EDTA, 1 m~ EGTA, 1 m~ dithiothreitol (DTT), protease inhibitor mixture, and 0.1 m~ phenylmethylsulfonyl fluoride (PMSF)). Addition of protease inhibitor mixture yields final concentrations of 0.1 pg/ml each of pepstatin A, chymostatin, and antipain, 1.0 pg/ml aprotinin, and 5.0 pg/ml leupeptin. The cell suspension was mixed with 210 ml of 0.5-mm acid-washed glass beads, and the volume was adjusted to 400 ml with buffer A. The cells were homogenized in a Biospec Model 11079 Bead Beater (Biospec Products Inc., Bartlesville, OK) using eighteen 10-s homogenizations separated by 20-s cooling periods. The homogenate was filtered through a 41-pm nylon mesh sheet to remove the glass beads, which were then washed with a n additional 40 ml of buffer A. The 280-ml homogenate was centrifuged for 10 min at 10,000 x g in a Sorvall SS34. The supernatant was recentrifuged for 15 min using the same rotor and speed. Fresh PMSF and leupeptin were added to the supernatant (0.1 m~ PMSF and 5 pg/ml leupeptin). Aliquots (16.7 ml) of the supernatant were layered over 7 ml of buffer B (20 mM Tris-C1, pH 7.4, 1 m~ Dl", protease inhibitor mixture) containing 0.1 m~ PMSF and 36% (w/w) sucrose. After centrifugation for 1 h in a Ti50.2 rotor (Beckman Instruments) at 184,000 xgav, the supernatant and the upper 2 ml of the sucrose cushion were discarded. The remainder of the sucrose cushion and the loose pellet were resuspended in 100 ml of buffer B with a Dounce homogenizer. The volume of the microsomal membranes was adjusted to obtain an optical density of 100 at 280 nm when measured in 1% SDS, and this concentration is defined a s 2 eq/pl. A typical yield of 110 ml of microsomal membranes (i.e. 220,000 eq) was obtained from a 7.6-liter culture.
Purification of the Yeast Oligosaccharyltransferase-Yeast microsomal membranes (200,000 eq) were adjusted to 166.5 ml with buffer B and mixed with 33.5 ml of 3 M NaCl. After a 20-min incubation on ice, the suspension was centrifuged for 1 h a t 184,000 x g,, in a Ti50.2 rotor. The supernatant was removed and the loosely packed membrane pellet was resuspended and adjusted to a final volume of 100 ml with buffer B. The microsomal membranes were permeabilized by the addition of an equal volume of 20 m~ Tris-C1, pH 7.4,0.2 M NaCl, 2 m~ MgCl,, 2 m~ MnCl,, 1 m~ DTT, protease inhibitor mixture, and 0.1% Nikkol (octaethylene glycol mono-N-dodecyl ether; Nikko Chemical Co, Ltd., Tokyo, Japan). After a 20-min incubation on ice, the membranes were recovered by centrifugation for 1 h a t 184,000 x g,, i n a Ti50.2 rotor. The membrane pellet was resuspended by Dounce homogenization in buffer B, adjusted to 110 ml, and then incubated for 30 min on ice after the addition of 90 ml of 3.3% digitonin, 1.1 M NaCl, 2 m~ MgCl,, 2 m~ MnCl,, 1 m~ Dl", protease inhibitor mixture. Digitonin (Sigma) stock solutions were prepared as described previously . Digitonin-high salt solubilized membrane proteins were separated from Ti50.2 rotor. an insoluble residue by centrifugation for 30 min at 184,000 x g,, in a To minimize dissociation of concanavalin A (ConA) oligomers in digi-tonin solution, the ConA-Sepharose CL-4B was cross-linked with dimethylpimelimidate (Pierce) essentially as described for other protein affinity reagents (Schneider et al., 1982). The dimethylpimelimidatecross-linked ConA-Sepharose was equilibrated with buffer C (20 m~ Tris-C1, pH 7.4, 1 m~ MgCl,, 1 m~ MnCl,, 1 m~ D m , protease inhibitor mixture, 0.125% digitonin, 34 p~ egg phosphatidylcholine) adjusted to 500 m~ NaCl. The detergent extract was batch-incubated for 12 h with 12 ml of ConA beads by end-over-end rotation. m e r centrifugation for 5 min at 1200 x g, the supernatant was collected as the ConA flowthrough fraction. The Cod-Sepharose beads were washed by successive resuspension and centrifugation steps four times with 88 ml of buffer C containing 500 m~ NaCl and two times with 88 ml of buffer C containing 100 m~ NaC1. Glycoproteins were eluted from the beads by three sequential incubations with 48 ml of buffer C containing 0.8 M a-methyl mannopyranoside and 100 m~ NaCl. The ConA-Sepharose was incubated with the elution buffer for 12 h by end-over-end rotation, at which time the eluted proteins were separated from the beads by centrifugation as described above. To avoid overloading the next column, half the eluate from the Cod-Sepharose column (72 ml) was mixed with an equal volume of 20 m~ Tris-C1 pH 7.4, 1 m~ Dm, protease inhibitor mixture, 0.125% digitonin, 34 p~ egg phosphatidylcholine, and applied to a 1.0-ml Mono Q column (Pharmacia LKB Biotechnology Inc.) that was equilibrated with buffer C containing 25 m~ NaCl and 20% glycerol. The Mono Q column was washed with 20 volumes of equilibration buffer and then eluted with a 50-ml linear 25-500 m~ NaCl gradient in buffer C containing 20% glycerol. Fractions of 1 ml were collected and assayed for oligosaccharyltransferase activity. The activity typically eluted as a 5-ml "shaped peak centered at 145 mM NaCl. The second half of the ConA eluate was processed in an identical manner. The pooled eluate from both Mono Q columns (10 ml) was diluted with 30 ml of buffer D (20 m~ Tris-C1, pH 7 . 4 , l m~ D m , 5 pg/ml leupeptin, 0.125% digitonin, 34 p~ egg phosphatidylcholine) and applied to a 1.0-ml Mono Q column that was equilibrated with buffer D containing 50 p~ MnC1,. The Mono Q column was washed with 20 ml of buffer D containing 25 m~ NaCl and 50 p~ MnCl, and then eluted with buffer D containing 200 m~ NaCl and 50 PM MnCl,. Fractions of 1 ml were collected and assayed, and the three most active fractions were pooled as the Mono Q eluate.
A 1.0-ml column of 3':5'-bis, bis-adenosine diphosphate agarose (3':5'-ADP-agarose, A3640; Sigma) was equilibrated with 25 ml of buffer D containing 500 m~ NaCl and 5 m~ MnCl,, followed by 25 ml of 20 m~ Tris-C1, pH 7.4, 5 mM MnCl,, followed by 25 ml of buffer D containing 50 p~ MnCl,. The Mono Q eluate was diluted 10-fold with buffer D to reduce the concentration of NaCl and MnC1, to 20 m~ and 5 PM, respectively, and the diluted sample was applied to the 3':5'-ADPagarose column a t a flow rate of 0.05 ml/min. The column was washed with 25 ml of buffer D containing 25 m~ NaCl and 50 p~ MnCl. The 3':5'-ADP-agarose column was eluted a t a flow rate of 0.5 mumin using a 10-ml linear gradient of both 25-500 m~ NaCl and 50 p~ to 5 m~ MnCl, in buffer D. The enzyme eluted as a symmetrical peak of A,,, centered at 225 m~ NaCl and 2.25 mM MnCl,.
Glycerol Gradient Centrifugation-A 2.1-ml sample of the eluate from the 3':5'-ADP-agarose column was loaded onto a 37-m1%30% (v/v) glycerol gradient in buffer C containing 500 m~ NaCl. The samples were centrifuged in a Beckman VTi50 rotor for 3.2 h a t 206,000 x gav. The fourteen 2.85-m1 fractions collected using an ISCO Model 185 gradient fractionator (ISCO, Inc. Lincoln, NE) were assayed for enzyme activity. Active fractions from the glycerol gradient were concentrated by Mono Q ion exchange chromatography as described above using buffer C containing 200 m~ NaCl and 20% glycerol as the elution buffer. The molecular weight of the oligosaccharyltransferase-digitonin complex was calculated using the following equation: M = (6nNq rh s ,~,~) + (1u p), where M is the molecular weight of the oligosaccharyltransferase-digitonin complex, N is Avogadro's number, q is the viscosity of water at 20 "C, s,,,~ is the experimentally determined sedimentation coefficient of the oligosaccharyltransferase-digitonin complex, rh is the Stokes radius of the oligosaccharyltransferase-digitonin complex, 6 is the partial specific volume of the complex, and p is the density of water a t 20 "C.
Oligosaccharyltransferase Assay-The iodinated tripeptide acceptor (Na-Ac-Asn-['2511Tyr-Thr-NH,; 18,000-28,000 cpdpmol) and the lipidlinked oligosaccharide donor for the oligosaccharyltransferase assay were prepared as described previously . Oligosaccharyltransferase assays of the digitonin-solubilized enzyme were identical with that described previously  except that the digitonin concentration was reduced to 0.007%.
Peptide Isolation and Protein Sequencing-Approximately 50-100 pmol of the yeast oligosaccharyltransferase complex was resolved into subunits by SDS-polyacrylamide gel electrophoresis. electrophoretically transferred onto 0.22-pm polyvinylidene diiluoride (PVDF) memhranes (Transhlot, Bin-Rad) or 0.45-pm nitrocellulose sheets (Schleicher and Schuell), and the subunits were localized by staining with Ponceau S. Coomassie Blue. or colloidal gold (Auro Dye Forte; Amersham). The stained hands were processed for amino-terminal sequencing using standard procedures (Matsudaira. 19x7) and were sequenced by the Worcester Foundation for Experimental Biology Protein Chemistry facility. Internal tryptic peptides from the 62and 64-kDa polypeptides were resolved by narrow-hore reverse phase HPLC after in situ trypsinization as described (Aehersold rt nl., 1987). The mass of two peptides from the HPLC elution profilrs was determined by matrixassisted laser desorption mass spectrometry ( a s reviewed hy Chait and Kent (1992)) using a Finnegan Lanermat mass spectrometer. I n situ trypsinization. peptide purification. and mass spectrometric analysis of peptides was perfnrmrd by the Harvard University Microchemistry Facility.
Protrin Glwtrophorrsis, Immrrnohlots, C o d Blots. and Endog1.vcosidnsc. H Digrstions-Proteins resolved hy polyacrylamide gel clectrophoresis in SDS were stained with Coomassie Blue or with silver (Ria-Rad) or were electrophoretically transferred to a 0.22-pm PVDF membrane. PVDF membranes were probed with a polyclonal rabbit antisera that rrcognizcs the Whpl protein (te Hresen rt 01.. 1991) or with ConA-peroxidase (Sigma) using standard procedures (Harlow and Lane. 1988). Peroxidase-lahrlcd secondary antibodies or ConA-peroxidase was detected hy rnhanced chemiluminescence (ECL; Amersham) following the manufacturer's recommendations. The oligosaccharyltransferase was digested with cndoglycosidase H essentially as described (Trimhlc and Malry, 1984). Kndoglycosidase H was kindly provided by Dr. Robert Trimhle (NY State Dept. of Health).

RESULTS
Purification of the Yrast Oligosacchn~ltransfrrase-The yeast oligosaccharyltransferase was purified from microsomal membranes by combining the following purification steps: selective removal of peripheral and lumenal membrane proteins, solubilization of integral membrane proteins with digitonin, ConA affinity chromatography, Mono Q ion exchange chromatography, and 3':5'-ADP-agarose affinity chromatography. Purification of the yeast oligosaccharyltransferase was monitored by analyzing the protein composition of selected fractions by Coomassie Blue staining of SDS-polyacrylamide gels (Fig. 1). Yeast microsomal membranes (lane a ) were extracted with 0.5 M NaCl to remove peripheral membrane proteins and then permeabilized with 0.05% Nikkol, a nonionic detergent, to selectively extract lumenal content proteins. The salt-stripped, detergent-permeabilized membranes (lnnr h ) were solubilized with a combination of 1.5% digitonin and 0.5 M NaCl to obtain a detergent extract (lane c ) . The majority of the proteins in the digitonin extract do not bind to a ConA-Sepharose column but were instead recovered in a flow-through fraction (lane c l ) t h a t lacked detectable oligosaccharyltransferase activity. The enzyme activity was eluted from the ConA column with methyl a-n-mannopyranoside (lane e ) and further enriched by ion exchange chromatography using a Mono Q ion exchange column. Rechromatography of the active fractions on a second Mono Q column to concentrate the sample and reduce the divalent metal ion concentration yields the Mono Q pool shown in lane f. Rased upon the supposition that the oligosaccharyltransferase contains a binding site for Mn", as enzyme activity is strictly dependent upon Mn" (Trimble et al., 1980;Sharma et al., 1981). we applied the eluate from the Mono Q column to a 3':5'-ADP-agarose column that had been equilibrated with MnCI,. Enzyme activity bound to the affinity column and was subsequently eluted with a linear gradient of NaCl and MnCI, (lane g ) . Selective binding of oligosaccharyltransferase to the column was evident, as the majority of proteins with molecular weights greater than 70,000 did not bind and were recovered in a flow-through fraction as shown by a comparison of the load (lanr f ) and eluate (lane g ) fractions. Additional experiments indicate that the enzyme will bind to a n ATP-agarose column that was equilibrated in an identical manner (data not shown). Furthermore, the concentration of MnCI, required to elute the activity from the 3':5'-ADP-agarose column was not significantly altered when the NaCl concentration was held constant during elution (data not shown), suggesting that immohilized Mn" rather than the adenine nucleotide is the affinity ligand.
The Coomassie Blue staining intensity of a number of polypeptides in lanes r-g was proportional to the amount of enzyme activity loaded on the gel. These proteins. which are the presumed subunits of the oligosaccharyltransferase complex, include a triplet ofpolypeptides migrating bctwccn 60 and 64 kDa (a), a 45-kDa fp), a 34-kDa ( y ) , a 30-kDa (6). a 16-kDa ( E ) , and a 9-kDa ( 0 polypeptide. An additional 7-kDa polypeptide was visible when large aliquots of the preparation were analyzed by polyacrylamide gel electrophoresis in SDS. The 7-kDa polypeptide co-migrated with the bovine lung aprotinin that was included in all purification huffers as n protrase inhibitor. For this reason, we tentatively identify the 7-kDa polypeptide as aprotinin. Although the final preparation contains traces of several other polypeptides (r.g. 75 and 90 kDa). we present evidence below showing that the latter pnlypcptides do not correlate with enzyme activity.
The recovery and enrichment of the oligosaccharyltransferase were monitored by activity assays using the tripeptide N"-Ac-Asn [ ' 2 S I ]~r -T h r -N H , as the acceptor and dolichollinked oligosaccharide isolated from bovine pancreas as the donor (Table I).
Oligosaccharyltransferasc assays of intact membranes cannot be compared with assays of detergent extracts due to the low quantity of endogenous oligosaccharide donor (OS-PP-Dol). Instead, rough microsomnl membranes and the salt-stripped, Nikkol-permcabilized membranes were as- The letter in parentheses following each fraction designates the corresponding grl lane in Fig. 1.
Total activity and specific activity values for membrane fractions are enclosed in parentheses due to the large appnrrnt incrrnsr in nrtivity that occurs upon preparation of the digitonin extract (see text). The oligosaccharyltransferase activity in the Cod-Srphnrosr rluatr wns nssnvrd aftrr gel filtration chromatography to remove the n-methyl mannopyranoside. The yield of 94% for the oligosaccharyltransferase activity in the digitonin extract assumes that 6a of t h r oligosaccharyltrnnsff,rn?ic. activity was trapped within the detergent-insoluble residue (see Fig. 6).
'' The 6.2-fold fold enrichment of activity in the digitonin extract is calculated from the recovery of protein in the digitonin rxtract rrlativr to thr RM fraction corrected for the assumed 94% yield of activity (i.r. 6.2 = (1455 + 219) x 0.94).
sayed by adding suflicient digitonin to solubilize these samples prior to assays in the presence of exogenous OS-PP-Dol. Nevertheless, the enzyme activity detected in the rough microsomal fraction was only 80% of that detected in the salt-stripped Nikkol-permeabilized membrane fraction, and the latter fraction was only 60% a s active as the subsequent digitonin extract. Clearly, the initial fractionation procedures remove endogenous inhibitors of oligosaccharyltransferase activity including proteins that interfere with the assay by serving as oligosaccharide acceptors (Trimble et al., 1980). Approximately 15% of the protein present in the microsomal membrane fraction was recovered in the digitonin extract (Table I). To provide a more realistic estimate of the recovery of the oligosaccharyltransferase during the initial two purification steps, we have utilized protein immunoblot analysis (see Fig. 6). Based upon this analysis, we estimate that 94% of the oligosaccharyltransferase complex present in intact membranes was recovered in the digitonin extract. The high recovery of enzyme activity during subsequent chromatography steps permits a calculated yield of 52% relative to the initial membrane fraction. The 4070-fold enrichment of activity during the purification procedure is based upon the estimated recovery of 94% of the actual activity during the initial purification steps.
Co-purification of OliRosaccharyltransferase Activity with a Protein Complex-Experimental evidence for the existence of a n oligosaccharyltransferase complex was obtained by determining which polypeptides co-purify with enzyme activity. Yeast oligosaccharyltransferase purified as described above can be partially resolved from several less rapidly sedimenting proteins by glycerol gradient centrifugation (Fig. 2). In the experiment shown here, an additional 1.3-fold enrichment in enzyme activity was obtained. Nonspecific hydrophobic or ionic interactions that could result in artifactual co-sedimentation of proteins were minimized by including digitonin and 0.5 M NaCl in the glycerol gradient. The enzyme activity, which was recovered in fractions 8-10 of a glycerol gradient (Fig. 2B), co-sedimented precisely with the putative a-5 subunits of the oligosaccharyltransferase complex (Fig. 2.4 ). Co-sedimentation of the a-3 polypeptides on a glycerol gradient suggests that these proteins are subunits of a complex, as monomeric proteins ranging between 9 and 64 kDa would sediment less rapidly. We noted that the relative staining intensity of the subunits varied depending upon whether the polyacrylamide gels were stained with Coomassie Blue (Fig. 1) or with silver ( Fig. 2 A ) . For example, the p subunit stained most intensely with Coomassie Blue, while the fi subunit stained most intensely with silver. Therefore, the lower staining intensity of the y subunit relative to the fi subunit in Fig. 2 should not be equated with a substoichiometric yield of the y subunit in this experiment. Fraction 1 is the top of thr gradient. h'. duplicntr 1-pI nliquots from rnch of 14 gradient fractions wrre assayrd for oliRoRacchnry1tr;lnsfrrnsr nctivity. 9sr+ of the oligosaccharyltrnnsfrrns(, actlv~ty npplird to thr k~ndient was recovered. Thr srdimrntation positions of stnndnrd protrlns (cytochrome c. 1.7 S; aldolasr. 7.9 S; cntalnsr. 11.4 S; thyro~lohulin. Ifi.5 S) was determined by polyacrylamidr grl rlrctrnphorrs~s of fr;~ct~ons from a duplicate glycrrol p a d i r n t .
A sedimentation coefiicient (s2,),,, ) of 14 can hr rstimakd for the oligosaccharyltransferase complex in digitonin solution hv comparison with the sedimentation rates of protein stnndnrds (Fig. 2R ). Gel filtration chromatogrnphv of thr yrnst olipwnccharyltransferase complex on a Suprrosr 12 column (Phnrmn-cia) yielded a Stokes radius of 50 A relative to protein standards (data not shown). The molecular weight of the oligosaccharyltransferase-digitonin complex can be calculated from these data provided that the partial specific volume of the protein-detergent complex is known. Although the latter value was not experimentally determined, the partial specific volume (E) of the digitonin micelle is 0.73 cmVg (Steele et al., 1978); a value that falls within the range of partial specific volumes for typical proteins (0.71-0.74 crn:'/g). Consequently, 0.73 cm,"/g is a reasonable estimate for the L, of the oligosaccharyltransferasedigitonin complex. From the sedimentation velocity and gel filtration data, we can estimate a molecular weight of 287,000 for the oligosaccharyltransferase-digitonin complex. The combined molecular weight of the rr-< subunits of the oligosaccharyltransferase complex was estimated to be 196,000 by denaturing gel electrophoresis (Fig. l), suggesting that the protein-detergent complex contains roughly 90 kDa of digitonin. The latter value is consistent with the reported molecular mass of 70 kDa for a digitonin micelle in 0.1 M NaCl (Smith and Pickels, 1940). The size of detergent micelles increases as the ionic strength is raised, so the precise size of the digitonin micelle under the experimental conditions used here is not known. Nonetheless, the hydrodynamic data obtained here are consistent with the proposed subunit composition of the oligosaccharyltransferase complex.
Additional information concerning the subunit composition of the oligosaccharyltransferase complex was obtained by examining the elution profile of the enzyme on a Mono Q anion exchange column (Fig. 3). Two partially resolved peaks of oligosaccharyltransferase activity were obtained when enriched samples of the enzyme were eluted from a Mono Q column with a shallow NaCl gradient (Fig. 3A). For the experiment shown here, the fractions from the Mono Q column were analyzed on a 9% polyacrylamide gel to maximize the resolution of the 60-64-kDa polypeptides (Fig. 3R). As expected, the active fractions from the anion exchange column contain the presumed subunits of the oligosaccharyltransferase complex, while the inactive fractions lack these proteins. The 75-and 90-kDa polypeptides designated by diamonds (Fig. 3R) do not co-elute with either peak of enzyme activity and are therefore not subunits of the oligosaccharyltransferase. The most remarkable feature of the elution profile is the partial resolution of the 60-64-kDa polypeptides. The 64-kDa polypeptide is enriched in the first peak of enzymatic activity (fractions 13-15), while the 60-and 62-kDa polypeptides were enriched in the second peak of activity (fractions 16-20). While one interpretation of this result is that the co-purification of the 60-64-kDa polypeptides with oligosaccharyltransferase activity is fortuitous, we favor a different explanation. Instead, we suggest that the 60-64-kDa polypeptides correspond to alternate forms of a single polypeptide, and that differences in the ion exchange properties of these variants are responsible for the unanticipated separation of the enzyme into two partially resolved activity peaks.
The relative amount of the a-fi subunits in each fraction from the Mono Q column was determined by densitometric scanning of the Coomassie Blue-stained gel. The densitometric units for a given subunit were then normalized to the amount of that subunit in fraction 14 so that the relative staining intensity of each subunit could be compared with the normalized oligosaccharyltransferase activity (Fig. 3C). In the case of the 60-64-kDa polypeptides, the densitometric units for the three presumed variants of the n subunit were combined before normalization. The correlation between enzymatic activity and the staining intensity of the cr-fi subunits was within the experimental error expected for quantification of proteins by densitometry. When Mono Q elution profiles were analyzed on higher percentage polyacrylamide gels, the E and < subunits also co-eluted with enzyme activity (data not shown). However, densitometric quantification of these subunits was not possible due to the low staining intensity of these polypeptides.
The Gl.vcosylated Subunits of the O l i~o s a r c h n~l ( r a n s f~r a s~ Complex-The subunits of the yeast oligosaccharyltransferase complex were transferred to a PVDF membrane after denaturing gel electrophoresis, and the membrane was probed with ConA to determine which polypeptides contain oligosaccharides. The major glycoproteins in the preparation co-migrate precisely with the n and p subunits (Fig. 4A ). The two morc  (45,0001, carbonic anhydrase (30,000). and soybean trypsin inhihitor (22,000). The tncomplrte rrcovrry of t h r y subunit is due to acid pH-induced precipitation of this suhunit during sample preparation for the endoglycosidase H digrstion. abundant variants of the a subunit (i.e. the 64-and 62-kDa forms) were readily detected (lanes a and b ) while the less abundant 60-kDa variant was detected when greater amounts of protein were loaded (lane d ). The remaining subunits of the complex do not bind C o d , so they lack N-linked and 0-linked oligosaccharides. Several less abundant proteins were not quantitatively removed by the 3'5"ADP-agarose column (Fig.   1, compare lanes f and g), and these glycoproteins (* and + ) persist as minor contaminants in the preparation (Fig.   4A, lanes d and e). The oligosaccharyltransferase complex was digested with endoglycosidase H to determine whether the a and p subunits contain N-linked or 0-linked oligosaccharide (Fig.   4B). After an extensive digestion with endoglycosidase H, the mobility of the subunit was reduced by approximately 3 kDa consistent with the presence of either one or two N-linked oligosaccharides (compare lanes b and c). The presence of a single digestion intermediate in addition to the final product ( p ) after a brief incubation with endoglycosidase H indicates that the p subunit contains two N-linked carbohydrates (lane d ). A larger reduction in molecular weight was seen after digestion of the a subunit (lanes h and c). More importantly, a single major product (a") was derived from the three variants of the a subunit by extensive digestion with endoglycosidase H (lane c). The ladder of evenly spaced digestion intermediates (lane d ), strongly suggests that the 64-, 62-, and 60-kDa forms of the a subunit contain, respectively, four, three, and two N-linked oligosaccharides.
The formation of a single major endoglycosidase H digestion product from the 60-64-kDa polypeptides supports the hypothesis that these three polypeptides are glycosylation variants of a single protein. To solidify this conclusion, the 62-and 64-kDa polypeptides were resolved by PAGE in SDS, transferred to nitrocellulose, and localized by staining with Ponceau S. Pep-tides derived by in situ trypsin digestion of the immobilized 62and 64-kDa polypeptides were eluted and resolved on a reverse phase high pressure liquid chromatography (HPLC) column using a gradient of acetonitrile in lr5 trifluoroacetic acid (Fig.  5 ) . A comparison of the HPLC elution profiles for tryptic peptides derived from the 64-kDa (panel A ) and 62-kDa (panel n ) polypeptides revealed a common set of major absorbance peaks. Two well-resolved peptides from each elution profile were analyzed by matrix-assisted laser deso:ption mass spectrometry. The molecular mass values obtained for the two peptides drrived from the 64-kDa polypeptide were idrntical, within experimental error, with the mass values obtainrd for the corresponding peptides derived from the 62-kDa subunit. Given that the elution position as well as the molecular mass of these peptides is identical, we can conclude that the 62-and the 64-kDa polypeptides contain identical protein s e p e n t s . Amino-terminal sequencing of the 62-and 64-kDa polypeptides also yielded identical amino acid sequences (data not shown ). As the latter sequence is not present in the current releases of the protein sequence databases (SWISS-PROT. NRRF PIR. and translated GenBank), we conclude that the n suhunit of the yeast oligosaccharyltransferase complex is a previously undrscribed yeast protein. Although suficient quantities of the 60-kDa polypeptide were not available for protein sequence analysis, the endoglycosidase H digestion experiment suggests that the latter polypeptide is also a variant of the n suhunit.
Identification of the p and 6 Subunits of the Oligosacchnt$transferase Complex-Having isolated a protein complex with demonstrable oligosaccharyltransferase activity, we askrd whether any of the subunits corresponded to the WRPl or SWPl gene products. After signal sequence cleavage. t h r Wbpl protein has a calculated molecular Wright of 46.900 and has two consensus sites for N-linked glycosylation (te Hersen et 01.. 1991). Protein immunoblot experiments were conducted to determine whether the 48-kDa glycosylated p subunit of the yeast oligosaccharyltransferase complex was identical with the Wbpl protein (Fig. 6). Antibody to the Wbpl protein recognizes a 48-kDa polypeptide in the purified oligosaccharyltransferase preparation (lanes e&). More importantly, the Wbpl protein co-migrates precisely with the fl subunit of the yeast oligosaccharyltransferase complex. Early steps in the purification procedure that were designed to selectively remove peripheral and lumenal RER proteins do not reduce the membrane content of the integral membrane protein Wbplp (compare lanes a and 6 ) . Likewise, more than 9070 of the Wbpl protein was recovered in the supernatant fraction (lane c ) after digitonin solubilization of the membranes. The presence of low amounts of the Wbplp in the digitonin pellet (lane d ) is likely due to trapping, as the pellet volume accounted for 6% of the solution volume prior to centrifugation. The estimated yield (94%) and enrichment (6.2fold) of the oligosaccharyltransferase at this point in the purification are based upon these data (see Table I).
The Swpl protein has a calculated molecular weight of 31,700 and lacks consensus sites for N-linked glycosylation ( t e Heesen et al., 1993). Both the 34-kDa y subunit and the 30-kDa 6 subunit of the oligosaccharyltransferase complex lack oligosaccharides and have apparent molecular weights that are similar to the calculated molecular weight of the Swpl protein. The y and 6 subunits were resolved by PAGE in SDS and were electrophoretically transferred to a PVDF membrane for amino-terminal sequence analysis. The amino-terminal sequence of the y subunit does not correspond to any portion of the Swpl protein (data not shown), nor could this sequence be aligned with any sequence in the current releases of the protein sequence databases (SWISS-PROT. NBRF PIR, and translated GenBank). However, the amino-terminal 16 residues of the fi subunit could be aligned with the Swpl protein beginning at residue 20 of the published protein sequence (Fig. 7). Consistent with this alignment, hydropathy analysis suggests that the Swpl protein is an integral membrane protein with a cleavable amino-terminal signal sequence ( t e Heesen et al., 1993). The most probable signal sequence cleavage site for the Swpl protein was calculated using the weight matrix method of von Heijne (von Heijne, 1986) and was found to agree with that determined by N-terminal sequencing of the mature protein ( Fig. 7). Thus, the p and fi subunits of the yeast oligoaaccharyltransferase complex correspond to previously identified yeast proteins that had been shown to be required for N-linked glycosylation ( t e Heesen et al., 1992( t e Heesen et al., , 1993 Homology between Swplp and Rihophorin II-The observation that the yeast oligosaccharyltransferase is a protein complex was not unexpected, based upon analogy to the canine enzyme (Kelleheret al., 1992). When the mammalian and yeast enzymes are electrophoresed in adjacent lanes, the Wbpl protein co-migrates with the homologous mammalian subunit OST48 (data not shown). Otherwise, only the Q subunit (M, = 62,000-64,OOO) in the yeast complex has a relative molecular mass similar to either ribophorin I (M, = 66,000) or ribophorin I1 (M, = 63,000). Antibodies raised against the mammalian ribophorins recognize avian and amphibian ribophorins I and 11, but do not cross-react with any yeast microsomal membrane proteins (Crimaudo et al., 1987). The difference in subunit size and the lack of antibody cross-reactivity raises the question of whether the yeast oligosaccharyltransferase complex contains homologues for either ribophorin I or ribophorin 11. A partial answer to this question is provided by a comparison of the protein sequences of ribophorin I1 and Swplp (Fig. 8 ) 1987) and Swplp (Fig. 7). More importantly, hydropathy analysis also reveals the presence in each protein of three hydrophobic segments near the carboxyl terminus that resemble membrane spans (Crimaudo et al., 1987;te Heesen et al., 1993). Interestingly, the spacing and relative hydrophobicity of the carboxyl-terminal segments in Swplp and ribophorin I1 are remarkably similar as shown by a comparison of the published hydropathy plots (Fig. &I). Together, the Nand C-terminal hydrophobic segments specify membrane integration of these proteins into the endoplasmic reticulum with the bulk of the protein located within the RER lumen. Proteolysis of canine microsomal membranes does not reduce the apparent molecular weight of ribophorin 11, suggesting that all three carboxylterminal hydrophobic segments may be membrane-associated (Crimaudo et al., 1987).
The BLASTP protein sequence comparison algorithm ( A tschul et al., 1990) was used to search the protein sequence databases (SWISS-PROT, NBRF PIR, and translated Gen-Bank) for proteins homologous to the Swpl protein. The most closely related proteins to Swplp were rat and human ribophorin 11. Optimal sequence alignment between ribophorin I1 and Swplp was obtained by comparing the mature sequence of Swplp with the carboxyl-terminal half of human ribophorin I1 (Fig. 8B). The sequence similarity between the two proteins extends throughout the entire mature region of Swplp. The sequence identity within this region was 22%, whereas sequence similarity was estimated to be 49%.

DISCUSSION
The yeast oligosaccharyltransferase preparation contains six major polypeptides that correlate with enzyme activity. These six polypeptides were observed to co-sediment with enzyme activity on a glycerol gradient and to co-elute with enzyme activity on a n ion exchange column. We conclude that these six polypeptides correspond to the subunits of a n oligosaccharyltransferase complex. Adventitious co-purification of these six polypeptides is unlikely as the relative staining intensity of the six proteins remained constant during fractionation procedures that rely on different biochemical and biophysical properties. Two subunits of the yeast oligosaccharyltransferase were shown to be Wbplp and Swplp. These two proteins were initially proposed as subunits of the yeast oligosaccharyltransferase based upon the finding that their expression is essential for N-linked glycosylation of proteins in vivo and for glycosylation of acceptor peptides in vitro (te Heesen et al., 1992(te Heesen et al., , 1993. Furthermore, the Wbpl protein can be cross-linked t o the Swpl protein in detergent extracts of yeast microsomal membranes, suggesting that the two proteins form a complex in vivo (te Heesen et al., 1993). However, alternate mechanisms could be invoked to explain how expression of the Wbpl and Swpl proteins could be indirectly required for oligosaccharyltransferase activity. Thus, co-purification of these two proteins with the oligosaccharyltransferase activity provides direct biochemical evidence that Wbpl and Swpl are subunits of the yeast oligosaccharyltransferase complex. Sequence alignment has disclosed that two of the yeast oligosaccharyltransferase subunits (Wbplp and Swplp) are homologous to mammalian oligosaccharyltransferase subunits (OST48 and ribophorin 11). We anticipate that further characterization of the yeast subunits will reveal a homologue for ribophorin I.
An unexpected difference between the yeast and mammalian oligosaccharyltransferase complexes was first noted when the amount of yeast and canine enzyme loaded on adjacent gel lanes was equalized with respect to total activity units. The subunits of the canine oligosaccharyltransferase complex stained more intensely than the corresponding yeast subunits (data not shown). Subsequent experiments suggest that the turnover number for the purified yeast complex may be 5-10fold higher than that estimated for the canine enzyme . However, before we can conclude that the yeast enzyme is catalytically more active, we must explore several alternative explanations for this difference in apparent enzyme activity. The oligosaccharide donor that we have isolated from bovine pancreas for the oligosaccharyltransferase assay is not a chemically homogeneous compound with respect to lipid content or oligosaccharide structure. OS-PP-Dol preparations contain other lipophilic compounds that could conceivably interfere with enzyme activity. The nonglucosylated compounds (Man,GlcNAc,-PP-Dol) have been reported to be severalfold more abundant than the glucosylated lipid-linked oligosaccharides (Glc,,Man,GlcNAc,-PP-Dol) in OS-PP-Dol preparations isolated from calf pancreas (Badet and Jeanloz, 1988). Incompletely assembled lipid-linked oligosaccharides lacking the terminal glucose residues are utilized at 10-35-fold reduced rates by the mammalian (Turco et al., 1977;Spiro et al., 1979) and yeast (Trimble et al., 1980) enzymes, when tested using solubilized membranes as the source for the oligosaccharyltransferase. The less restrictive utilization of incompletely assembled oligosaccharides by the yeast enzyme may account for the greater apparent turnover number of the latter enzyme when mixed populations of lipid-linked oligosaccharides are used as the donor substrate, primarily because the effective concentration of the donor substrate may be higher. A more accurate comparison between the purified yeast and mammalian enzymes will require the isolation and characterization of more highly purified donor substrates. Although the iodinated tripeptide substrate is chemically homogeneous, we have not determined whether similar differences in specific activity are observed when larger peptides are used as the oligosaccharide acceptor. Finally, the observed differences in subunit composition between the yeast and canine enzymes may be indicative of the selective loss of several low molecular weight subunits from the mammalian oligosaccharyltransferase complex during purification. If this were the case, the previously purified heterotrimeric canine oligosaccharyltransferase complex may be a low specific activity core of a more active oligosaccharyltransferase complex that may exist in intact membranes. The latter possibility is currently being explored.
Limited information is currently available concerning the a , y, E , and 5 subunits of the yeast oligosaccharyltransferase complex. The artifactual proteolytic derivation of the y, E, and 5 polypeptides from the a, p, and 6 subunits is considered unlikely because protease inhibitors were present throughout the purification, and the microsomal membranes were isolated