Coordination of N-Glycosylation and Protein Translocation across the Endoplasmic Reticulum Membrane by Sss1 Protein*

Secretory proteins are translocated across the endoplasmic reticulum (ER) membrane through a channel formed by three proteins, namely Sec61p, Sbh1p, and Sss1p (Johnson, A. E., and van Waes, M. A. (1999) Annu. Rev. Cell Dev. Biol. 15, 799–842). Sec61p and Sss1p are essential for translocation (Esnault, Y., Blondel, M. O., Deshaies, R. J., Schekman, R., and Kepes, F. (1993) EMBO J. 12, 4083–4093). Sec61p is a polytopic membrane protein that lines the protein translocation channel. The role of Sss1p is unknown. During import into the ER through the Sec61p channel, many proteins are N-glycosylated before translocation is completed. In addition, both the Sec61 channel and oligosaccharyl transferase (OST) copurify with ribosomes from rough ER, suggesting that OST is located in close proximity to the Sec61 channel (Gorlich, D., Prehn, S., Hartmann, E., Kalies, K.-U., and Rapoport, T. A. (1992) Cell 71, 489–503 and Wang, L., and Dobberstein, B. (1999) FEBS Lett. 457, 316–322). Here, we demonstrate a direct interaction between Sss1p and a subunit of OST, Wbp1p, using the split-ubiquitin system and co-immunoprecipitation. We generated mutants in the cytoplasmic domain of Sss1p that disturb the interaction with OST and are viable but display a translocation defect specific for proteins with glycosylation acceptor sites. Our data suggest that Sss1p coordinates translocation across the ER membrane and N-linked glycosylation of secretory proteins.

Protein translocation across the ER 1 membrane is an essential process in all eukaryotes (1). Translocation is mediated by a channel formed most likely by three copies of the heterotri-meric Sec61 complex (Sec61p, Sbh1p, and Sss1p in yeast; Sec61␣, Sec61␤, and Sec61␥ in mammals) (1,2,5). During cotranslational protein import through the Sec61 channel, the large ribosomal subunit binds to the channel, and channel proteins copurify with ribosomes upon solubilization of rough ER-derived microsomes (3,4). The signal peptide is cleaved off from secretory proteins as soon as the cleavage site is sufficiently exposed on the lumenal side (1). The enzyme complex responsible for cleavage, signal peptidase, is anchored to the Sec61 channel by Sec61␤ (6). This interaction is most likely transient, as signal cleavage is required only once early in translocation (6). The OST complex in yeast consists of eight or possibly nine subunits (Ost1p, Wbp1p, Stt3p, Swp1p, Ost2p, Ost3p, Ost5p, Ost6p, and Ost4p); the first five of the subunits listed are essential. All subunits except Ost4p, against which no antibodies are available, were shown to be assembled in a 240-kDa complex in the ER membrane (7). OST can N-glycosylate secretory proteins early during translocation into the ER and close to the lumenal exit of the Sec61 channel (7)(8)(9). Nglycan acceptor sites can occur throughout a protein and may need to be modified cotranslationally before they are buried by protein folding. Close proximity of OST to the Sec61 channel is therefore required until termination of translocation. Proximity of the two complexes has also been suggested by binding of the mammalian homologues of OST subunits, Ost1p (ribophorin I) and Swp1p (ribophorin II), to ribosomes that are associated with the Sec61 channel in the ER membrane (10,11). A direct interaction between OST and the Sec61 channel, however, has not been demonstrated to date, nor has a role for glycosylation in translocation been found.
To analyze the interaction of OST with other proteins in the ER membrane, we used the split-ubiquitin-based yeast twohybrid system, which can be used to monitor interaction between membrane proteins in situ (12,13). The N-terminal and C-terminal halves of ubiquitin (Nub and Cub), when fused to heterologous proteins, can interact and refold into a quasinative ubiquitin structure (13). A mutant form of Nub, NubG, can reconstitute ubiquitin with Cub only if both NubG and Cub are fused to proteins that are in close spatial proximity ( Fig.  1A) (13). If a transcriptional activator is fused to the C terminus of Cub, formation of quasi-native ubiquitin results in cleavage by cytosolic ubiquitin-specific proteases (Fig. 1A, UBP) carboxyl-terminally of Cub, diffusion of the transcriptional activator to the nucleus, and expression of reporter genes. Using this system we found an interaction between the OST subunit Wbp1p and the protein translocation channel subunit Sss1p. We generated mutants in Sss1p in which the interaction with Wbp1p is perturbed to investigate the functional significance of this interaction in vivo and in vitro.
Plasmids-pNub-ALG5 and pOST1-Nub are described (12). All other Nub fusions were cloned in-frame with the respective genes using PCR. Mutants in SSS1 were generated by error-prone PCR using a limiting concentration of dATP and the Nub-SSS1 plasmid as template. Inserts from plasmids that no longer supported growth of YG0673 on selective media without histidine were sequenced.
In Vitro Translocation-Microsomes were prepared from SSS1 wild type and mutant strains (15). Excess in vitro translated, [ 35 S]methionine-labeled, wild type pp␣F or a mutant form in which the glycosy-lation sites had been removed by site-directed mutagenesis (p⌬gp␣F) (10 6 cpm) was translocated into 0.5-l microsomes (A 280 ϭ 25) at 10 or 20°C for the indicated periods of time in the presence of ATP and an ATP-regenerating system (17). The amount of membranes was limiting FIG. 1. Sss1p and Wbp1p form a complex in the ER membrane. A, splitubiquitin constructs and detection of protein-protein interactions. Cub-PLV was fused to the cytosolic C terminus of Wbp1p. NubG was fused to cytosolic termini of an enzyme that catalyzes lipidlinked glucose formation, Alg5p, the OST subunit Ost1p, proteins required for misfolded protein export from the ER and degradation (Hrd1p, Der1p, Cue1p, and Ubc7p), and the translocon subunit Sss1p. Interaction of Wbp1p-Cub-PLV with Ost1p-NubG leads to release of PLV by cytosolic ubiquitin-specific proteases (UBP). B, growth of strains expressing Nub-Sss1 and Wbp1-Cub-PLV on minimal media without tryptophane and leucine (left side; LW) or without tryptophane, leucine, and histidine and with a chromogenic substrate of ␤-galactosidase (right side; LWH/X-Gal). C, lysates from strains expressing the indicated combination of proteins were separated by SDS-PAGE, and transferred proteins were detected with peroxidase-IgG. The asterisk indicates a Wbp1p-Cub-PLV derived band that occurred independently of ubiquitinspecific protease (UBP)-mediated cleavage.

FIG. 2. Subunits of OST and the protein translocation channel co-immunoprecipitate.
A, microsomes from wild type or Wbp1p-HAexpressing cells were lysed, and native immunoprecipitations were performed with the indicated antibodies. Immunoprecipitates (IP) were separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes and detected with anti-HA antibody. Proteins in the far right lane were digested with protein N-glycanase (PNGase) prior to electrophoresis. B, wild type yeast microsomes were lysed, and Sss1p and Wbp1p were immunoprecipitated in duplicate with specific polyclonal antisera as above. After electrophoresis and transfer, Sss1p was detected by immunoblotting. Blots shown in panels A, and B were also probed with an antibody against a signal peptidase subunit, Spc3p, but signal peptidase did not co-precipitate with Wbp1p or Sss1p. C and D, wild type yeast microsomes were lysed, and Sss1p and Wbp1p were immunoprecipitated in duplicate. After electrophoresis and transfer, Swp1p and Sbh1p, respectively, were detected by immunoblotting.
in the reactions. Trichloroacetic acid-precipitated samples were resolved on 18%, 4 M urea acrylamide gels, and a translocated, glycosylated, wild type 3gp␣F or a translocated, signal-cleaved, mutant ⌬gp␣F was quantified using a phosphorous imager (PerkinElmer Life Sciences). Samples were taken in duplicate. Translocated proteins were at least 90% protease-protected. Microsomes were washed with buffer containing 25 mM EDTA for 15 min and 500 mM potassium acetate for 1 h for the experiment shown in Fig. 5D as described (18).

Sss1p and Wbp1p Form a Complex in the ER Membrane-To
identify ER membrane proteins that interact with OST, we used a chimeric protein consisting of the OST subunit Wbp1p and Cub, Protein A, and the transcriptional activator LexA-VP16 (Wbp1p-Cub-PLV) as bait (Fig. 1A) (12,13). This bait has been well characterized and used successfully in the past to demonstrate the fidelity of the split-ubiquitin system to measure protein-protein interactions (12,13). Interaction of this protein with a NubG fusion protein results in cleavage of the PLV moiety by cytosolic ubiquitin-specific proteases (Fig. 1A, UBP), diffusion of PLV to the nucleus, and transcription of Lex A-VP16-controlled reporter genes in our strain HIS3 and LacZ. As shown in Fig. 1B, co-expression of Wbp1p-Cub-PLV with a NubG-tagged OST subunit, Ost1p, resulted in growth of blue colonies on selective plates without histidine (see also Ref. 12). Direct interaction between the ER lumenal domains of the mammalian homologs of Ost1p and Wbp1p (ribophorin I and OST48, respectively) has been shown previously (19). This result confirms that the split-ubiquitin system can serve to monitor interaction of ER membrane proteins. Alg5p, which generates lipid-linked glucose in the ER membrane but is not a subunit of OST, did not interact with Wbp1p in the splitubiquitin assay (Fig. 1B) (12). We generated a variety of NubGfusions to the cytoplasmic N or C termini of other reasonably abundant ER-resident proteins and a subunit of the protein translocation channel whose function has been poorly defined to date (Sss1p). Wbp1p-Cub-PLV did not interact with any other NubG-tagged ER membrane or ER-associated protein tested, with the exception of Sss1p (Fig. 1B). Cleavage of Wbp1p-Cub-PLV in the presence of Ost1p-NubG or NubG-Sss1p was confirmed directly by blotting with peroxidase-IgG (Fig. 1C). Sec61p and Ost1p did not interact in the split-ubiquitin assay, suggesting that the binding of Wbp1p to Sss1p was specific.
To ascertain that the interaction between Wbp1p and Sss1p was independent of the ubiquitin moieties in the fusion proteins, we co-immunoprecipitated Wbp1p tagged with an influenza hemagglutinin epitope (Wbp1p-HA) with a polyclonal anti Sss1p antiserum (Fig. 2A). In addition, we co-immunoprecipitated Sss1p with polyclonal anti-Wbp1p antiserum from a strain in which neither protein was tagged (Fig. 2B). Under the lysis conditions used, Sss1p antibodies also precipitated the OST subunit Swp1p (Fig. 2C), and Wbp1p antibodies precipitated Sbh1p (Fig. 2D) and Sec61p (not shown), suggesting that both the OST complex and the Sec61 complex were stable and associated with each other. Our data demonstrate a close physical interaction between the Sec61 channel and the OST complex in the ER membrane.
Mutations in Sss1p Disrupt Interaction with Wbp1p in the Split-ubiquitin System-To investigate the functional significance of the interaction of Wpb1p and Sss1p, we used errorprone PCR to generate Sss1p mutants that no longer interact with Wbp1p in the split-ubiquitin system (Fig. 3A). We isolated two mutants, sss1-1 and sss1-4, that expressed mutant Sss1p at wild type levels (not shown). In both sss1-1 and sss1-4, a charged residue replaces a conserved hydrophobic or uncharged residue within the amino-terminal cytoplasmic ␣-helical domain of Sss1p (sss1-1: Q17 to R; sss1-4: V18 to D; Fig. 3, B and C); in addition, the mutations in the double mutant sss1-4 are predicted to disrupt the ␣-helix (Fig. 3C). Despite the complete lack of interaction of both mutant Sss1 proteins with Wbp1p in the split-ubiquitin system, both proteins still coimmunoprecipitated Wbp1p (not shown). This suggests that the sss1 mutations affect the orientation of OST and the translocon toward each other rather than completely disrupting their interaction. Haploid strains in which wild type SSS1 had been replaced with either of the mutant genes were viable, and the mutant strains showed no growth phenotype over a temperature range from 16 -37°C (not shown), suggesting that the essential function of Sss1p in protein translocation was not affected.

Mutations in Sss1p Affect N-Glycosylation and Protein
Secretion-We predicted that a disturbed interaction between the translocon and OST should result in less efficient glycosylation and, hence, a reduced amount of intracellular glycoproteins after a short pulse. We labeled wild type and sss1 mutant yeast for 5 min with [ 35 S]methionine and lysed the cells either immediately or after a 5-min chase. Intracellular glycoproteins were isolated from the lysates using the mannose-specific lectin Concanavalin A. As shown in Fig. 4A, both sss1 mutants displayed a profound reduction in glycoprotein biosynthesis compared with wild type (58% of wild type after 5 min chase).
[ 35 S]methionine incorporation was comparable in wild type and sss1 mutant strains. Most secretory proteins in yeast are Nglycosylated and, for many proteins, glycosylation is important for folding. As only fully folded proteins are packaged into ER-to-Golgi transport vesicles and secreted (20), hypoglycosylation should result in reduced protein secretion. Protein secretion in the sss1 mutants, however, was delayed rather than diminished (Fig. 4B). We observed a similar delayed secretion for individual glycoproteins (not shown). Consistent with the transient nature of the glycoprotein biosynthesis defect in the sss1 mutants, these strains displayed no increased sensitivity to tunicamycin compared with wild type yeast. Our data suggest that N-linked glycosylation and protein folding were inefficient when the interaction of OST with the Sec61 channel was suboptimal but that most proteins were eventually glycosylated and secreted.
We next examined the glycosylation status of specific proteins in the ER of SSS1 wild type and mutant cells. We found that neither carboxypeptidase Y (CPY; four N-linked glycans) nor protein disulfide isomerase (PDI; five N-linked glycans) nor the translocated ER-form of the ␣-factor precursor (3gp␣F, three N-linked glycans) were under-glycosylated in the mutants (Fig. 4C, and not shown) (21-23). Translocation of the cytosolic ␣-factor precursor pp␣F into the ER lumen of sss1 mutants, however, was delayed (Fig. 4C). Note that in wild type and sss1 mutant cells, the translocated, glycosylated ␣-factor precursor 3gp␣F is transported to the Golgi with equal rapidity and proteolytically processed during the 5-min chase, suggesting that there is no defect in transport of secretory proteins in these cells (Fig. 4C, compare 0-and 5-min chase). No translocation defect was detected for the precursors of PDI, CPY, or a protein without N-glycan acceptor sites, BiP, even when the pulse time was reduced to 1 min (Fig. 4C, and not  shown). was not affected in the mutants. C, SSS1 wild type and mutant cells were labeled for 5 min and chased for the indicated periods. PDI and pp␣F were immunoprecipitated with polyclonal antisera and visualized after electrophoresis on 10% gels (PDI) or 18% 4 M urea gels (pp␣F) using a phosphorous imager (PerkinElmer Life Sciences). Note that the gel system used can resolve pp␣F and p␣F and that PDI is mildly under-glycosylated in this strain background (5gPDI, fully glycosylated).

Mutations in Sss1p Cause a Translocation Defect Specific for
Glycoproteins-The N-linked glycans of ␣-factor precursor (pp␣F) are required for protein folding (24). All three glycosylation acceptor sites of pp␣F are in the N-terminal half of the protein, and, in contrast to PDI and CPY, the first is within four amino acids of the signal cleavage site (60 amino acids, PDI; 106 amino acids, CPY) (21)(22)(23). Our data suggested that glycosylation of one or more of these sites was essential for efficient pp␣F translocation into the ER lumen. We examined this possibility in a cell-free system based on in vitro translated, radiolabeled pp␣F and microsomes prepared from SSS1 wild type and mutant cells (25,26). As shown in Fig. 5A, posttranslational pp␣F translocation into wild type microsomes was extremely efficient with a t1 ⁄2 of less than 2 min. Translocation of pp␣F into sss1 mutant microsomes was significantly slower (t1 ⁄2 , 5 min), and maximal import was reduced by ϳ20% compared with wild type. To investigate the effect of the sss1 mutants on cotranslational import, we performed experiments with a modified version of pp␣F in which the signal peptide had been replaced with a transmembrane domain (D HC ␣F) (18). Maximal import of this protein was similarly reduced in the mutants (not shown). Decreasing the assay temperature to 10°C exacerbated the translocation defect in sss1 mutant microsomes (Fig. 5B). To examine whether this difference in import kinetics was indeed related to glycosylation of the translocation substrate, we repeated the experiment with a mutant form of pp␣F in which the N-glycan acceptor sites had been removed by site-directed mutagenesis (p⌬gp␣F) (27). Translocation of p⌬gp␣F into wild type microsomes was slower than that of pp␣F (t1 ⁄2 of 5 min versus Ͻ 2 min; Fig. 5, C versus A, black squares) but proceeded with identical kinetics in wild type and sss1 mutant microsomes at 20°C (Fig. 5C). Our data suggest that the translocation defect for pp␣F in sss1 microsomes was indeed due to reduced glycosylation efficiency caused by the perturbed translocon-OST association in these membranes.
A recent paper by Nikonov et al. (28) suggests that OST may stay associated with Sec61 channels after termination of translocation. To disrupt existing translocon-OST complexes, we treated SSS1 wild type and mutant microsomes with EDTA and high salt (18). This treatment removes ribosomes from the cytoplasmic face of the microsomes, which may stabilize the interaction of OST with the Sec61 channel. The high salt/EDTA wash also disrupts electrostatic protein-protein interactions mediated by cytoplasmic domains of ER proteins. Translocation of pp␣F into salt-washed microsomes was slower than into untreated membranes (Fig. 5, D versus A, black squares) suggesting that the interaction of OST with the translocon had been affected by the EDTA/high salt treatment. Membranes from both sss1 mutants were severely deficient in the reformation of a functional OST/Sec61 channel complex in vitro and, hence, in the import of pp␣F (Fig. 5D, open symbols). DISCUSSION We have shown here that an essential subunit of the protein translocation channel in the ER membrane, Sss1p, mediates binding of the OST complex to the channel via the OST subunit Wbp1p. The role of Sss1p in OST recruitment to the protein translocation channel is independent of its essential function in stabilizing the translocation channel, as mutants in the cytoplasmic domain of Sss1p are viable and display no general defects in protein import into the ER (29, 30) (not shown; Figs. 4 and 5). Appropriate anchoring of OST to the protein translocation channel promotes the immediate glycosylation of translocating proteins and this, in turn, accelerates protein translocation into the ER of a high proportion of proteins with N-glycan acceptor sites and, hence, results in efficient biosynthesis of glycoproteins (Fig. 4A). Using a posttranslationally imported secretory glycoprotein and a mutant version without glycan acceptor sites as a model proteins, we have demonstrated that at physiological temperature, membranes containing mutant Sss1p, which do not interact appropriately with FIG. 5. Mutations in Sss1p cause a translocation defect specific for glycoproteins. Excess pp␣F or p⌬gp␣F, as indicated, was translocated at 10 or 20°C for the indicated periods of time into a limiting amount of microsomal membranes from SSS1 wild type (black squares) or mutant cells (open squares, sss1-1; open circles, sss1-4). Incubations were terminated by addition of trichloroacetic acid to 10%, and precipitated proteins were resolved by SDS-PAGE on 18% 4 M urea gels. Translocated, glycosylated 3gp␣F and translocated, signal-cleaved p␣F were quantified using a phosphorous imager. Each time point was taken in duplicate, and each experiment was repeated at least once. Variation at each time point was Ͻ2%. Membranes in panel D were treated with 25 mM EDTA and 0.5 M potassium acetate to remove ribosomes from the membranes prior to translocation.
Wbp1p, are fully competent for import into the ER of nonglycoproteins (p⌬gp␣F, Fig. 5C). Thus, the structure of the protein translocation channel itself in the sss1-1 and sss1-4 mutants is not significantly compromised. The half-time for import of a secretory glycoprotein into the ER, however, was significantly increased in these sss1 mutants (2.5ϫ; Fig. 5A), suggesting that glycosylation contributes to efficient translocation into the ER.
Ours is the first demonstration of an effect of glycosylation on protein translocation into the ER. The acceleration of protein import by glycosylation may be due to improved folding of the lumenal part of the translocation intermediate. The Nglycans of pp␣F are indeed required for its folding (24); the unglycosylated mutant form p⌬gp␣F is not transported efficiently from the ER to the Golgi complex but rather is retained in the ER and subsequently targeted for ER-associated degradation (24,31). Matlack et al. have shown that pp␣F is prone to backsliding in the translocation channel and that its import into proteoliposomes, which do not contain active OST, requires binding of bulky molecules such as BiP or pp␣F antibodies in the ER lumen (32). In intact cells, glycosylation of the N-terminal proregion of pp␣F early during translocation into the ER may fulfill a similar function. Although N-glycans themselves cannot prevent backsliding in the translocation channel (33), glycosylation-induced folding of the pp␣F proregion may interfere with movement of the translocating chain toward the cytoplasm and thus promote vectorial transport of pp␣F into the ER lumen.
The propensity for backsliding in the translocon may be a function of the signal sequence that is recognized by the protein translocation channel itself (34). We found indeed that if the pp␣F signal peptide was substituted with the transmembrane region of dipeptidyl aminopeptidase B (DPAPB) and the protein was cotranslationally imported into microsomes, glycosylation no longer increased the speed of import into the ER, although maximal import into sss1 mutant membranes was still lower than into wild type microsomes (not shown). Most studies on protein import into the ER are done with cotranslationally imported substrates, which may explain why an effect of glycosylation on import has not been observed so far. Our observation that mutations in SSS1 that disturb the OSTtranslocon interaction lead to a delay in biosynthesis of ϳ40% of glycoproteins in Saccharomyces cerevisiae suggests that glycosylation contributes to the efficient import into the ER of a large number of proteins. In a preliminary analysis of secretory glycoproteins in the S. cerevisiae genome, we found that the majority of these contain multiple N-glycan acceptor sites, which makes it likely that their modification contributes to protein folding. In addition, a significant fraction of glycoproteins contain acceptor sites close to the N terminus, where glycosylation is more likely to have an immediate effect on vectorial transport through the protein translocation channel.
At present, we cannot exclude the possibility that additional interactions between subunits of the protein translocation channel and OST subunits contribute to the association between the two complexes. Given that mutations in SSS1 alone resulted in a glycosylation-related translocation defect, however, we propose that the binding of Wbp1p to Sss1p is essential for OST-translocon interaction. Further experiments should allow us to address the mechanisms that allow the close cooperation between the Sec61 channel and OST during protein translocation into the ER.