Ubiquitination is required for the retro-translocation of a short-lived luminal endoplasmic reticulum glycoprotein to the cytosol for degradation by the proteasome.

In the endoplasmic reticulum (ER), an efficient "quality control system" operates to ensure that mutated and incorrectly folded proteins are selectively degraded. We are studying ER-associated degradation using a truncated variant of the rough ER-specific type I transmembrane glycoprotein, ribophorin I. The truncated polypeptide (RI332) consists of only the 332 amino-terminal amino acids of the protein corresponding to most of its luminal domain and, in contrast to the long-lived endogenous ribophorin I, is rapidly degraded. Here we show that the ubiquitin-proteasome pathway is involved in the destruction of the truncated ribophorin I. Thus, when RI332 that itself appears to be a substrate for ubiquitination was expressed in a mutant hamster cell line harboring a temperature-sensitive mutation in the ubiquitin-activating enzyme E1 affecting ubiquitin-dependent proteolysis, the protein is dramatically stabilized at the restrictive temperature. Moreover, inhibitors of proteasome function effectively block the degradation of RI332. Cell fractionation experiments indicate that RI332 accumulates in the cytosol when degradation is prevented by proteasome inhibitors but remains associated with the lumen of the ER under ubiquitination-deficient conditions, suggesting that the release of the protein into the cytosol is ubiquitination-dependent. Accordingly, when ubiquitination is impaired, a considerable amount of RI332 binds to the ER chaperone calnexin and to the Sec61 complex that could effect retro-translocation of the polypeptide to the cytosol. Before proteolysis of RI332, its N-linked oligosaccharide is cleaved in two distinct steps, the first of which might occur when the protein is still associated with the ER, as the trimmed glycoprotein intermediate efficiently interacts with calnexin and Sec61. From our data we conclude that the steps that lead a newly synthesized luminal ER glycoprotein to degradation by the proteasome are tightly coupled and that especially ubiquitination plays a crucial role in the retro-translocation of the substrate protein for proteolysis to the cytosol.

ity control system" operates to ensure that mutated and incorrectly folded proteins are selectively degraded. We are studying ER-associated degradation using a truncated variant of the rough ER-specific type I transmembrane glycoprotein, ribophorin I. The truncated polypeptide (RI 332 ) consists of only the 332 amino-terminal amino acids of the protein corresponding to most of its luminal domain and, in contrast to the long-lived endogenous ribophorin I, is rapidly degraded.
Here we show that the ubiquitin-proteasome pathway is involved in the destruction of the truncated ribophorin I. Thus, when RI 332 that itself appears to be a substrate for ubiquitination was expressed in a mutant hamster cell line harboring a temperature-sensitive mutation in the ubiquitin-activating enzyme E1 affecting ubiquitin-dependent proteolysis, the protein is dramatically stabilized at the restrictive temperature. Moreover, inhibitors of proteasome function effectively block the degradation of RI 332 . Cell fractionation experiments indicate that RI 332 accumulates in the cytosol when degradation is prevented by proteasome inhibitors but remains associated with the lumen of the ER under ubiquitination-deficient conditions, suggesting that the release of the protein into the cytosol is ubiquitinationdependent. Accordingly, when ubiquitination is impaired, a considerable amount of RI 332 binds to the ER chaperone calnexin and to the Sec61 complex that could effect retro-translocation of the polypeptide to the cytosol. Before proteolysis of RI 332 , its N-linked oligosaccharide is cleaved in two distinct steps, the first of which might occur when the protein is still associated with the ER, as the trimmed glycoprotein intermediate efficiently interacts with calnexin and Sec61.
From our data we conclude that the steps that lead a newly synthesized luminal ER glycoprotein to degradation by the proteasome are tightly coupled and that especially ubiquitination plays a crucial role in the retro-translocation of the substrate protein for proteolysis to the cytosol.
Most proteins of the endomembrane system as well as plasma membrane and secretory proteins are synthesized on polysomes bound to the membrane of the rough endoplasmic reticulum. During and shortly after their synthesis, the ectodomains of these polypeptides assume their three-dimensional conformation in the lumen of the ER, 1 and the proteins may then also become part of oligomeric complexes. The ER houses an efficient "quality control system" to ensure that transport out of this organelle is limited to properly folded and assembled proteins (1,2). Proteins that fail to assume their correct final conformation in the lumen of the ER, in most cases, do not stably remain in this compartment, rather they are degraded by a proteolytic system (3,4).
An increasing number of diseases characterized by an "ER storage phenotype" results from impaired quality control of the ER (5). For instance, it has been observed that in most cases of cystic fibrosis, mutated forms of the transmembrane conductance regulator (CFTR) are not expressed at the cell membrane but are retained and degraded in the ER or a related compartment (6,7). Similarly, ␣ 1 -antitrypsin (␣ 1 -AT) deficiency patients with the Z mutation in ␣ 1 -AT accumulate the mutant protein in the ER of hepatocytes (8,9). In addition, some cases of familial hypercholesterinemia (10,11) and Tay-Sachs disease (12) are also related to impaired transport out of the ER. The process, and possibly the mechanism(s) involved, of ERassociated degradation appears to be highly conserved in eukaryotes, as this phenomenon has also been observed in yeast (13)(14)(15).
In several cases it has been shown that substrate proteins for ER-associated degradation interact with chaperones present in the ER, which thus might be involved in the quality control process that leads to targeting of those proteins for degradation. For instance, the binding of CFTR (16) and of the PiZ variant of ␣ 1 -AT (9,17) to calnexin, a chaperone that recognizes glycoproteins in their mono-glucosylated forms in the ER (18,19), has clearly been demonstrated. Furthermore, BiP that has been shown to bind to a variety of folding intermediates in the lumen of the ER could also represent a candidate protein that interacts with misfolded polypeptides that eventually are delivered for proteolysis. In fact, it appears that the time of interaction of a substrate protein with BiP correlates with its half-life (20,21).
The major pathways of protein degradation in the eukaryotic cell include lysosomal proteolysis, ubiquitin-dependent lysosomal proteolysis, ubiquitin-independent proteasomal proteolysis, and ubiquitin-dependent proteasomal proteolysis (22)(23)(24). The latter plays a pivotal role in the rapid turnover of abnormal proteins and in the regulation of the steady state of a variety of proteins that include cyclins, kinases, tumor suppressors, and transcriptional regulators (23)(24)(25). In this case, ubiquitin, a small polypeptide of 76 amino acids, is activated by a ubiquitinactivating enzyme (E1) in a reaction that requires ATP hydrolysis. The activated ubiquitin molecule is then transferred to a ubiquitin-conjugating enzyme (E2) that catalyzes the formation of an isopeptide bond between the COOH-terminal glycine of ubiquitin and the ⑀-amino group of a lysine residue on target proteins. The mono-ubiquitinated substrates then undergo further ubiquitinations via the lysine residue at position 48 of ubiquitin, leading to the formation of multi-ubiquitin chains that target proteins to degradation by the 26 S proteasome (23,24).
In recent years it has become clear that soluble and integral membrane proteins that have been targeted to the ER are, in fact, degraded by the ubiquitin-proteasome pathway. By using a genetic approach, Hiller et al. (26) provided evidence that the degradation of a variant of carboxypeptidase Y (CPY*) that is retained in the ER of yeast depends on the activity of Ubc6p, an ER-bound ubiquitin-conjugating enzyme, as well as functional proteasomes. Furthermore, it was demonstrated that retrograde transport of CPY* from the ER to the cytosol depends on ubiquitination, in which a complex is involved that consists of the two ubiquitin-conjugating enzymes, Ubc6p and Ubc7p, and Cue1p, an ER transmembrane protein required for the recruitment of Ubc7p to the ER membrane (27). Moreover, Wiertz et al. (28) observed an interaction between MHC class I heavy chain molecules and Sec 61, suggesting that retrograde transport of substrate proteins to the cytosol may occur through the translocation channel. This view was strengthened by the recent finding that certain mutant yeast Sec61 alleles are defective in the export out of the ER of substrate proteins for degradation (29,30).
We are studying ER-associated degradation using a COOHterminally truncated variant of ribophorin I, a type I ER transmembrane glycoprotein that is a component of the oligosaccharyltransferase complex (31)(32)(33). When the mutant protein, RI 332 , that contains only the NH 2 -terminal 332 amino acids of the luminal domain of ribophorin I is expressed in permanent transformants of HeLa cells, it is rapidly degraded by a nonlysosomal pathway with biphasic kinetics. The first phase of degradation is characterized by a half-life of about 1 h and is followed, after approximately 45 min, by a second phase of 3-fold accelerated degradation. In contrast, endogenous ribophorin I is very stable and has a half-life of more than 24 h (34,35). Here we show that the ubiquitin-proteasome pathway is involved in the degradation of RI 332 and that, in fact, release to the cytosol of the substrate protein for degradation and ubiquitination are tightly coupled.

Reagents-
The mammalian expression vector pCI-neo was purchased from Promega (Madison, WI); maleimide-activated keyhole limpet hemocyanin was from Pierce, and protein A-Sepharose CL-4B beads were from Pharmacia (Uppsala, Sweden). Geneticin (G418 sulfate), ␣-minimal essential medium, methionine-free RPMI 1640, other cell culture components, and Lipofectin were from Life Technologies, Inc. Trypsin from bovine pancreas, BFA, aprotinin, leupeptin, L-leucyl-Lleucyl-L-leucine, PMSF, NEM, TLCK, TPCK, Tricine, and CHAPS were purchased from Sigma (Deisenhofen, Germany), and ZLLL and ZLL- Antibodies-The polyclonal rabbit antibody against rat liver ribophorin I was a generous gift from Dr. Gert Kreibich (New York University School of Medicine) and has been described previously (34,36,37). The polyclonal rabbit anti-Sec61␤ antibodies (38) and anti-PDI antibodies were kindly provided by Dr. Tom A. Rapoport (Harvard Medical School, Boston) and Dr. David A. Gordon (Bristol-Myers Squibb, Princeton, NJ), respectively. The monoclonal mouse CTR433 antibody, a marker for the medial Golgi cisternae (39), was a gift from Dr. Michel Bornens (Institut Curie, Paris, France). The polyclonal anti-calnexin antibody is directed against the COOH-terminal peptide of calnexin (amino acids 555-573 of the mature dog protein) (Ref. 40). The peptide that contains an additional cysteine residue at the NH 2 -terminal side was coupled to maleimide-activated keyhole limpet hemocyanin and injected into a rabbit for antibody production (see also Ref. 41). A polyclonal rabbit anti-ubiquitin antibody was purchased from Stress-Gen Biotechnologies (Victoria, Canada). Affinity purified, Texas Redconjugated goat anti-mouse F(abЈ) 2 -IgG was obtained from Accurate Chemicals (Westbury, NY).
Cell Culture and Transfections-E36 and ts20 cells were kindly provided by Dr. Alan L. Schwartz (Washington University School of Medicine) (42). The cells were grown at 30°C in ␣-minimal essential medium, supplemented with glucose (4.5 g/liter), 10% fetal calf serum, penicillin G (100 IU/ml), streptomycin sulfate (100 g/ml), and amphotericin B (250 ng/ml). The generation of the cDNA coding for the 332 NH 2 -terminal amino acids of rat ribophorin I (RI 332 ) and its cloning into the mammalian expression vector pCI-neo will be described elsewhere (see also Refs. 32 and 34). The cells were transfected with the expression construct by the Lipofectin method according to the manufacturer's instructions, using 1 g of DNA and 10 l of Lipofectin reagent on cells cultured in a 6-cm dish and an incubation time of 18 h. Permanent transformants of E36 and ts20 cells expressing RI 332 (designated E36-RI 332 and ts20-RI 332 cells) were obtained after selection for growth in the presence of geneticin (600 mg/liter). Single clones of highly expressing cells were selected, cultured in the continued presence of geneticin (300 mg/liter), and used further for the experiments performed during this study.
Treatment of Cells with Proteasome Inhibitors, Temperature Conditions, Cell Labeling, and Immunoprecipitations-The transfected E36-RI 332 and ts20-RI 332 cell cultures were grown in 35-mm dishes near to confluence (5-8 ϫ 10 5 cells per dish). For pulse-chase experiments at the restrictive temperature, the cells were preincubated at 41°C for 2 h. For treatments with proteasome inhibitors, the cells were pretreated at 30 or at 41°C with ALLN (80 M), ZLLNva (40 M), or ZLLL (50 M) for 90 min in complete medium and for another 30-min period in serumand methionine-free RPMI 1640 medium. Cells were labeled in serumand methionine-free medium containing [ 35 S]methionine (250 Ci/ml) for 10 min at the indicated temperatures. Subsequent chase incubations were carried out in complete medium supplemented with methionine (5 mM). Cells were lysed with 300 l of an SDS-containing buffer (25 mM Tris⅐HCl, pH 7.4, 95 mM NaCl, 3 mM EDTA, 2% SDS, and a mixture of the following protease inhibitors: 1.7 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml L-leucyl-L-leucyl-L-leucine, 5 mM PMSF), and after sonication and boiling of the cell lysate, 1 ml of wash buffer (25 mM Tris⅐HCl, pH 7.4, 95 mM NaCl, 3 mM EDTA, 0.2% SDS, 1.25% Triton X-100, protease inhibitors as above) was added. Immunoprecipitations were performed with anti-ribophorin I antiserum (4 l/ml lysate) and analyzed by SDS-PAGE using 8% gels, unless noted otherwise, and fluorography, as reported previously (34). When necessary, immunoprecipitates were treated with endo H, as described (43). Quantitations of immunoprecipitations were performed by scanning densitometry using the ImageQuant software (Molecular Dynamics, Sunnyvale, CA).
Indirect Immunofluorescence-ts20-RI 332 cells were grown on coverslips for 36 h at 30°C. After pretreatment of the cells at the appropriate temperature conditions, followed by an incubation in the absence or presence of BFA (5 g/ml) for 30 min, the cells were fixed with 100% methanol at Ϫ20°C for 3.5 min and then subjected to immunofluorescence staining, as described previously (44). The monoclonal CTR433 mouse antibody was used as a marker for the medial Golgi cisternae (39) and applied at a dilution of 1:2 in blocking medium (1% non-fat dry milk in phosphate-buffered saline). The secondary affinity purified, Texas Red-conjugated goat anti-mouse F(abЈ) 2 antibody fragments were used at a dilution of 1:40 in the same medium. After mounting, the staining was visualized on a Zeiss Axiovert 135 photomicroscope (Carl Zeiss, Oberkochen, Germany) equipped with epifluorescence optics and photographed using Kodak TMAX-400 ASA film.
Sequential Immunoprecipitations with Anti-ribophorin I and Antiubiquitin Antibodies-E36-RI 332 cells were grown in a 6-cm dish near to confluence (1.5-2 ϫ 10 6 cells). The cells were pretreated and further incubated during the experiment with ZLLL, and then pulse labeling was done for 30 min, followed by a chase incubation of 10 min under the conditions described above. A cell lysate was prepared with 600 l of lysis buffer (25 mM Tris⅐HCl, pH 7.4, 95 mM NaCl, 3 mM EDTA, 0.2% SDS, protease inhibitors) in the presence of NEM (5 mM), which was also included during all subsequent manipulations to inhibit potent isopeptidase activities that may affect the detection of multi-ubiquitinated proteins (45). After sonication and boiling of the lysate, 2 ml of wash buffer (25 mM Tris⅐HCl, pH 7.4, 95 mM NaCl, 3 mM EDTA, 0.2% SDS, 1.25% Triton X-100, protease inhibitors) were added. Two 20% aliquots of the total lysate were used for anti-ribophorin I immunoprecipitations under the conditions described above, except that the SDS concentration was maintained at 0.2%. Two aliquots, 10 and 50% of the total lysate, were used for anti-ubiquitin immunoprecipitations (10 l antiserum/ml lysate). Immunocomplexes were recovered as described previously (34). Immunoprecipitates from one of the 20% aliquots of the anti-ribophorin I samples and the 10% aliquot of the anti-ubiquitin samples were analyzed directly by SDS-PAGE. For the others, the material was eluted from the protein A-Sepharose beads by boiling for 5 min in a buffer containing 0.2% SDS, 25 mM Tris⅐HCl, pH 7.4, 95 mM NaCl, 3 mM EDTA, protease inhibitors, and 5 mM NEM. These eluates were subjected to a second round of immunoprecipitations in the presence of 0.2% SDS, 1% Triton X-100, protease inhibitors, and 5 mM NEM using anti-ubiquitin and anti-ribophorin I antisera, respectively, before analysis as described (34).
Sequential Immunoprecipitations with Anti-ribophorin I and Anticalnexin Antibodies-E36-RI 332 and ts20-RI 332 cells were plated in 6-cm dishes. The cells, treated at 30 and 41°C, respectively, in the absence or presence of ZLLL, were subjected to metabolic labeling for 2 h followed by chase incubations. The preparation of the cell lysates (2.6 ml total volume) and the immunoprecipitations with the anti-calnexin (10 l/ 350 l of lysate) and anti-ribophorin I antibodies were performed in the presence of 1% CHAPS, as described previously (41). Immunocomplexes were either analyzed directly by SDS-PAGE as described (34) or eluted from the protein A-Sepharose beads by boiling for 5 min in a buffer containing 2% SDS, 25 mM Tris⅐HCl, pH 7.4, 95 mM NaCl, 3 mM EDTA, and protease inhibitors. The eluates were subjected to a second round of immunoprecipitations using anti-ribophorin I and anti-calnexin antibodies, respectively, before analysis as described (34).
Sequential Immunoprecipitations with Anti-Sec61␤ and Anti-ribophorin I Antibodies-ts20-RI 332 cells, grown in 6-cm dishes, were left untreated or treated with ZLLL at 30°C or incubated at 41°C. The cells were labeled for 30 min, incubated in chase medium, and lysed (2.6 ml total volume). The co-immunoprecipitations with anti-Sec61␤ (2.5 l/ 850 l of lysate) and anti-ribophorin I antibodies were carried out as described above for the anti-calnexin immunoprecipitations, except that no NaCl was included in the lysis and wash buffers and that the immunoprecipitates were analyzed on Tricine gels as described (46).
Cell Fractionation and Separation of Membranes and Aggregates-E36-RI 332 cells were grown in 10-cm dishes near to confluence (approximately 5 ϫ 10 6 cells per dish). The cells were left untreated or pretreated with ZLLL at 30°C for 90 min in complete medium and for another 30-min period in serum-and methionine-free medium in the absence or presence of the inhibitor. Similarly, ts20-RI 332 cells were plated in 10-cm dishes and incubated at 30 or 41°C. Cells were labeled with [ 35 S]methionine (250 Ci/ml) for 30 min and then incubated in chase medium for 10 min or 1 h in the continued absence or presence of the proteasome inhibitor and at the appropriate temperature.
Cells were lysed in a Dounce homogenizer in the presence of an isotonic buffer (25 mM Tris⅐HCl, pH 7.4, 0.25 M NaCl, 1 mM EDTA, protease inhibitors). The lysates were centrifuged at 1000 ϫ g for 5 min to remove unbroken cells and cell debris. The protein concentrations of the 1000 ϫ g supernatants were determined (47), and equal amounts of protein in each supernatant were subjected to a 100,000 ϫ g ultracentrifugation. The 100,000 ϫ g supernatants were adjusted to a final concentration of 25 mM Tris⅐HCl, pH 7.4, 0.6% SDS, 1% Triton X-100, 95 mM NaCl, 3 mM EDTA, and protease inhibitors as above and used for anti-ribophorin I and anti-PDI (6 l/sample) immunoprecipitations, respectively. The pellets were resuspended in lysis buffer (25 mM Tris⅐HCl, pH 7.4, 2% SDS, 95 mM NaCl, 3 mM EDTA, protease inhibitors). Two equal aliquots of the resuspended pellets, corresponding to the same amount of cellular material used for the analysis of the supernatants, were subjected to anti-ribophorin I and anti-PDI immunoprecipitations.
Alternatively, cells were lysed by Dounce homogenization in an isotonic buffer (25 mM Tris⅐HCl, pH 7.6, 0.25 M sucrose, protease inhibitors). Nuclei, cell debris, and unbroken cells were sedimented at 1000 ϫ g for 8 min. The supernatants were loaded on a 2 M sucrose cushion containing 25 mM Tris⅐HCl, pH 7.6, 5 mM EDTA, and protease inhibitors and then overlaid with isotonic buffer described above. The samples were centrifuged in a Beckman SW60 rotor at 110,000 ϫ g for 16 h. The interfaces between the 2 M and the 0.25 M sucrose solutions were recovered and subjected to anti-ribophorin I immunoprecipitations under stringent conditions (34). The pellets were resuspended in lysis buffer containing 2% SDS (see above) and also used for immunoprecipitations. The immunoprecipitates were analyzed by SDS-PAGE and fluorography as described (34).
Protease Protection-ts20-RI 332 cells, left at 30°C or preincubated at 41°C, were labeled with [ 35 S]methionine (250 Ci/ml) for 30 min and then incubated in chase medium for 10 min or 1 h at the appropriate temperature. Cells were lysed by Dounce homogenization in an isotonic buffer (25 mM Tris⅐HCl, pH 7.5, 0.25 M NaCl), and the lysates were centrifuged at 1000 ϫ g for 5 min. The supernatants were divided into equal aliquots, one of each was left untreated, whereas the others were incubated with trypsin (10 to 50 g/ml) in the absence or presence of Triton X-100 (0.5%) for 30 min at 30°C. Then the protease activity was blocked by the addition of TPCK, TLCK (500 g/ml each), and PMSF (5 mM). After transfer to 4°C, the samples were processed for immunoprecipitation using the anti-ribophorin I antibody, as described previously (34). The immunoprecipitates were analyzed by electrophoresis on 15% SDS-polyacrylamide gels, followed by fluorography.

RESULTS
To characterize the pathway involved in the degradation of RI 332 , we established permanent transformants of CHO-E36 and CHO-ts20 cells that express the protein. E36 is the wild type cell line, and ts20 is the corresponding mutant that expresses a thermolabile ubiquitin-activating enzyme E1 (42). At temperatures above 40°C, the ubiquitin system and consequently its protein-ubiquitin conjugating capacity are inactivated to less than 10%. We compared the life cycle of RI 332 at the permissive temperature (30°C) and at the non-permissive temperature (41°C) in both cell lines (Fig. 1).
RI 332 was rapidly degraded at the permissive temperature in E36-RI 332 cells so that after 60 min of chase no band was detectable on the gel, whereas in ts20-RI 332 cells the protein was degraded somewhat more slowly, but also in this cell line only a very small amount of RI 332 was recovered after 60 min of chase (Fig. 1A).
As to the degradation at 41°C in both cell lines, it is noteworthy that heat treatment of many cells including E36 leads to stress-induced degradation of many cellular proteolysis substrates, but heat treatment of ts20 cells inactivates the ubiquitin-conjugating system (42). At the non-permissive temperature, the inactivation of ubiquitination strongly prevented the degradation of RI 332 in ts20-RI 332 cells, as a substantial fraction of the original amount of the protein persisted even after 3 h and 30 min of chase (Fig. 1C). As expected, RI 332 was rapidly degraded in E36-RI 332 cells at 41°C (Fig. 1B).
It is interesting to mention that RI 332 molecules synthesized at 41°C in ts20-RI 332 cells remain in a state susceptible to degradation, which occurs as soon as the incubation temperature is changed to 30°C (Fig. 1D). This observation may be due to the fact that the cells rapidly regain their ubiquitin-conjugating capacity under these conditions, thus allowing for efficient ubiquitin-dependent degradation of the protein.
Many proteins that are substrates for ubiquitin-dependent proteolysis pathways accumulate in ts20 cells at the non-permissive temperature; in addition, ubiquitination is required for a variety of different cellular processes. Therefore, a concern with the use of ts20 cells was the degree to which cellular functions are affected at 41°C. A process to date not known to depend on ubiquitination is BFA-mediated retrograde Golgi to ER transport (48); hence, the ability of ts20-RI 332 cells to support BFA-induced relocation of a Golgi protein to the ER at 41°C was investigated. Indirect immunofluorescence labeling using the CTR433 antibody as a marker for the medial cisternae of the Golgi apparatus (39) on ts20-RI 332 cells at both 30 and 41°C resulted in a typical Golgi staining (Fig. 2, A and C). Conversely, after treatment of the cells with BFA for 30 min, a fluorescence pattern characteristic for the ER was obtained at both temperatures (Fig. 2, B and D). This result indicates that BFA-induced retrograde Golgi to ER transport remains functional in ts20-RI 332 cells at the non-permissive temperature. Furthermore, it is apparent that RI 332 must have been essentially fully translocated into the lumen of the ER, as it receives its N-linked oligosaccharide at Asn 275 (see below, Fig. 4). These observations, together with the result described in Fig. 1D, indicate that several cellular processes related to the endomembrane system are not affected to any significant extent by incubation of the ts20-RI 332 cell mutant at 41°C under the conditions used in this study.
To investigate if the proteasome is involved in the rapid turnover of RI 332 , we observed the effect of different proteasome inhibitors such as ALLN (MG101), ZLLNva (MG115), and ZLLL (MG132) on the degradation of RI 332 in E36-RI 332 cells at 30°C (Fig. 3A) and at 41°C (B) as well as of ts20-RI 332 cells at 30°C (C). All three of these proteasome inhibitors markedly blocked RI 332 degradation, since a consistent amount of the protein was recovered after 2 h of chase.
Interestingly, two additional forms migrating approximately 1.5 to 2 kDa below the band corresponding to RI 332 appeared in the immunoprecipitates of E36-RI 332 cells incubated with proteasome inhibitors after 45 min of chase. Since it has been shown that several glycoproteins undergo deglycosylation prior to degradation, such as the heavy chain of the MHC class I molecules (28), we speculated that the newly arising band might correspond to deglycosylated forms of RI 332 . Ribophorin I as well as RI 332 contain three potential N-glycosylation sites, but only one of these is used by oligosaccharyltransferase. It has been shown in earlier work that RI 332 remains endo Hsensitive throughout its lifetime (43). Taking this premise into account, we performed an endo H digestion on anti-ribophorin I immunoprecipitates from E36-RI 332 and ts20-RI 332 cells incubated under various conditions (Fig. 4). As expected, after endo H digestion only a single band was detectable (RI 332 *; lanes b, e, h, and k), migrating at the same position as the lowest form of RI 332 recovered from the undigested immuno- precipitates (lanes a, d, g, and j). This finding suggests that the additional forms of RI 332 appearing after proteasome inhibitor treatment represent the fully deglycosylated protein (RI 332 *) as well as a species where the single N-glycan moiety has been trimmed. The latter form should, therefore, correspond to a trimmed glycoprotein intermediate (RI 332,i ). Although we did not establish the sugar composition of the oligosaccharide structure of this intermediate, these results indicate that endo H is capable to act on truncated N-glycans. In fact, it has been shown previously that endo H cleaves trimmed N-glycans that retain an ␣1,6-linked core oligosaccharide, whereas the truncated glycans detected on glycoproteins in a CHO cell mutant bearing a glycosylation and temperature-sensitive secretion defect, for example, were found to be endo H-resistant (49). It is noticeable that the trimmed glycoprotein intermediate of RI 332 accumulates over time when proteasomal degradation is inhibited by ZLLL (Fig. 3, lanes j-l). On the other hand, the fact that the lower band of the doublet migrating faster than RI 332 is observed at 10 min of chase (lane j) could be explained by newly synthesized RI 332 molecules that have never been N-glycosylated by oligosaccharyltransferase during the labeling period.
To determine directly whether RI 332 itself is a substrate for ubiquitination, co-immunoprecipitation experiments of ribophorin I and ubiquitin from a lysate of E36-RI 332 cells were performed. The proteasome inhibitor ZLLL was included during the pulse-chase incubations, and NEM, an isopeptidase inhibitor, was included in the immunoprecipitation buffers to accumulate proteasome substrates and to maintain them in their poly-ubiquitinated state. The cell lysate was used for a co-immunoprecipitation experiment as follows. Material immunoprecipitated by antibodies directed against ribophorin I was reprecipitated by anti-ubiquitin antibodies (Fig. 5, lane d) or vice versa (lane c). When RI 332 was immunoprecipitated from cell lysates of ZLLL-treated cells in the presence of 0.2% SDS and 1% Triton X-100, and in the absence of NEM, only bands between 38 and 36 kDa were detectable (lane a; see also Fig. 3). However, in subsequent immunoprecipitations with anti-ribophorin I and anti-ubiquitin antibodies (Fig. 5, lanes c and d), or even in the anti-ribophorin I immunoprecipitations alone (lane b), all under the same conditions but in the presence of NEM, higher molecular weight bands became evident representing ubiquitinated forms of RI 332 . In fact, it appears that the majority of the RI 332 molecules is ubiquitinated under the experimental conditions used, as in the presence of NEM the band corresponding to unubiquitinated RI 332 almost disappears (lane b). As expected, from the immunoprecipitation with anti-ubiquitin antibody alone, a number of higher molecular weight bands became discernible (lane e).
Considering that the ubiquitin-proteasome pathway is located in the cytoplasm, we wanted to investigate the intracellular distribution of RI 332 molecules when their degradation is inhibited. To this effect, a cell fractionation experiment was performed in which ribophorin I and RI 332 were immunoprecipitated from the membrane and the cytosolic fractions of E36-RI 332 cells (Fig. 6A) and ts20-RI 332 cells (Fig. 6B) labeled for 30 min and incubated in chase medium under different conditions. Only a small portion (12%) of RI 332 was detected in the cytosol of untreated E36-RI 332 cells at 10 min of chase (Fig.  6A, lanes a and b). In the presence of ZLLL, the total amount of Cells were harvested and lysed in the presence of SDS (0.2%) and NEM (5 mM). All subsequent immunoprecipitations were performed in the presence of 0.2% SDS and 1% Triton X-100. Two 20% aliquots of the total lysate were used for anti-ribophorin I immunoprecipitations, the first of which was analyzed directly (lane b), whereas the material recovered from the other one was reprecipitated with anti-ubiquitin antibodies (lane d). Another co-immunoprecipitation with anti-ubiquitin antibodies first and with anti-ribophorin I antibodies in the second round was performed on a 50% aliquot of the lysate (lane c). The remaining 10% of the lysate was subjected to an immunoprecipitation with anti-ubiquitin antibodies (lane e). As a control, ribophorin I and RI 332 were immunoprecipitated from a cell lysate under the same conditions (0.2% SDS, 1% Triton X-100) but in the absence of NEM (lane a). The immunoprecipitates were analyzed by SDS-PAGE followed by fluorography. (Exposure times of lanes b-d was  15 days and of lanes a and e was 1 day.) The arrow points to distinct bands corresponding to ubiquitinated forms of RI 332 . RI 332 recovered from E36-RI 332 cells at 10 min of chase increased slightly (to 17%) when compared with untreated cells, as the protein becomes stabilized after inhibition of the proteasome (compare lanes e and f with a and b). In ts20-RI 332 cells, a portion of RI 332 (16 and 19%, respectively) was recovered from the cytosol after 10 min of chase at 30 and 41°C (Fig. 6B,  lanes a and b as well as e and f).
Strikingly, after extension of the chase time to 1 h, a significantly increased amount (61%) of RI 332 was recovered from the cytosol of ZLLL-treated E36-RI 332 cells (Fig. 6A, lanes g and h), indicating that the retro-translocated protein accumulates in the cytosol when the function of the proteasome is compromised. In contrast, RI 332 was not detectable in the cytosol of ts20-RI 332 cells at 41°C after 1 h of chase (Fig. 6B, lane h) indicating that the small portion recovered at the 10-min chase time point must have been degraded, most likely due to the residual ubiquitin-activating capacity present in ts20 cells at the restrictive temperature. In both control E36-RI 332 and ts20-RI 332 cells at 30°C, the truncated ribophorin I was essentially degraded after 1 h of chase (lanes c and d). In addition, it became apparent that the fully deglycosylated RI 332 (RI 332 *) was only detectable in the cytosol of ZLLL-treated E36-RI 332 cells (Fig. 6A, lane h), whereas the trimmed RI 332 intermediate (RI 332,i ; Fig. 6A, lanes e and g; see also Fig. 4) was observed in both the pellet and supernatant fractions. This intermediate was also recovered from the membrane fraction of ts20-RI 332 cells at 41°C (Fig. 6B, lanes e and g). These findings suggest that the deglycosylation process may be initiated in the lumen of the ER and completed in the cytosol.
As a control for the integrity of the microsomes prepared, the luminal ER protein PDI (50) was immunoprecipitated from all membrane and cytosolic fractions used for detection of the truncated ribophorin I. Some PDI ( Fig. 6; 8% in E36-RI 332 cells; 29% in ts20-RI 332 cells at 30°C; and 26% in ts20-RI 332 cells at 41°C) was consistently found in the cytosolic fractions indicating some breakage of the microsomes during cell fractionation. Even though this observation indicates that some RI 332 assigned to cytosolic fractions may be due to its leakage from microsomes, it is undoubtedly evident that the protein accumulates in the cytosol of ZLLL-treated E36-RI 332 cells (Fig. 6A,  lanes e-h).
To ascertain that RI 332 indeed remains associated with the ER of ts20-RI 332 cells at 41°C and does not form aggregates in the cytoplasm, a cell fractionation experiment was performed in which membrane vesicles and high molecular weight protein complexes and aggregates were separated. To this aim, total membrane fractions prepared from radiolabeled ts20-RI 332 cells grown at 30°C or preincubated at 41°C were subjected to ultracentrifugation over a 2 M sucrose cushion for 16 h. Under the conditions used, membranes float on top of this cushion, whereas protein complexes and also aggregates are sedimented (51). Ribophorin I and RI 332 were immunoprecipitated from the interface above the 2 M sucrose cushion and from the pellet (Fig. 7). It is clearly evident that the majority (more than 95%) of RI 332 is present in the membrane fraction at 41°C and after 10 min (lane c) and 60 min (lane e) of chase, indicating that under ubiquitination-deficient conditions the protein remains membrane-associated for extended periods. At 30°C, RI 332 is also recovered from the membrane fraction shortly after its synthesis (lane a). These findings strongly suggest that RI 332 does not form cytosolic aggregates when ubiquitination is impaired.
Furthermore, it was of interest to determine if RI 332 , when associated with microsomes, is contained within their lumen or facing the cytosolic side of the membrane. For this purpose, the accessibility of RI 332 to exogenously added protease was assessed. Total membrane fractions prepared from radiolabeled ts20-RI 332 cells grown at 30°C (Fig. 8A) or preincubated at 41°C (Fig. 8B) were treated with increasing concentrations of trypsin in the absence or presence of Triton X-100 for 30 min at 30°C. When the detergent was omitted, in all cases RI 332 remained protected to a large extent from exogenously added FIG. 6. RI 332 remains membrane-associated when ubiquitination is blocked, whereas the protein accumulates in the cytosol when the proteasome activity is inhibited. E36-RI 332 cells (A) were pulse-labeled for 30 min and chased for 10 min or 1 h at 30°C in the absence or presence of ZLLL (50 M). ts20-RI 332 cells (B) were pulselabeled for 30 min and chased for 10 min or 1 h at 30 or 41°C. Cells were lysed in isotonic buffer by Dounce homogenization, and from equal amounts of protein, membrane (P) and soluble (S) fractions were separated by ultracentrifugation at 100,000 ϫ g. Ribophorin I and RI 332 or PDI were immunoprecipitated from the cell fractions obtained. The immunoprecipitates were analyzed by SDS-PAGE followed by fluorography. Note that the deglycosylated form of RI 332 detected here in the soluble fraction of ZLLL-treated E36-RI 332 cells (A, lane h) does not appear in the experiment shown in Fig. 3, which is probably due to the different labeling conditions used. FIG. 7. RI 332 is recovered with microsomal membrane fractions and does not form cytosolic aggregates in ts20-RI 332 cells at the non-permissive temperature. ts20-RI 332 cells were pulselabeled for 30 min and chased for 10 min (lanes a-d) or 60 min (lanes e  and f) at 30°C (lanes a and b) or 41°C (lanes c-f). Cells were lysed in an isotonic buffer by Dounce homogenization. Microsomal fractions were recovered from the interface between a 2 M sucrose cushion and a 0.25 M sucrose solution after ultracentrifugation of the lysates at 110,000 ϫ g for 16 h. The microsomal fractions (M, lanes a, c, and e) and the pellet (P) fractions containing proteinaceous high molecular weight constituents (P, lanes b, d, and f) were processed for immunoprecipitation using the anti-ribophorin I antibody. The immunoprecipitates were analyzed by SDS-PAGE followed by fluorography. protease (lanes b-d, bЈ-dЈ, and gЈ-iЈ). On the other hand, when Triton X-100 was included in the incubations with the lowest concentration of trypsin, RI 332 was readily degraded (lanes e, eЈ, and jЈ). These results show that newly synthesized RI 332 is restricted to the lumen of the ER. Moreover, when the protein is stabilized due to impaired ubiquitin-conjugating capacity of ts20-RI 332 cells at 41°C, RI 332 remains segregated to the ER lumen, even after 1 h of chase.
Summarizing the results from these experiments, it appears that RI 332 remains mostly confined to the lumen of the ER when ubiquitination is affected, as is the case in ts20-RI 332 cells at 41°C, whereas the protein accumulates in the cytosol of E36-RI 332 cells when the proteasome is inhibited. Thus, it may be concluded that ubiquitination is a prerequisite to trigger the release of RI 332 into the cytoplasm for degradation by the proteasome.
To analyze the role of calnexin in the quality control of the glycoprotein RI 332 , we performed pulse-chase and co-immunoprecipitation experiments using anti-ribophorin I and anticalnexin antibodies under different conditions that block RI 332 degradation (Fig. 9). In Fig. 9A, an experiment is shown to visualize the patterns obtained with ts20-RI 332 cells at 41°C. Ribophorin I and RI 332 (lane a) as well as calnexin (lane f) were detected as distinct bands under stringent conditions (in the presence of 0.6% SDS and 1% Triton X-100). A variety of bands covering a wide range of apparent molecular weights was recovered from both the anti-ribophorin I (lane b) and the anticalnexin (lane e) immunoprecipitates under non-stringent conditions, when the mild detergent CHAPS was used for cell solubilization. As these bands were not observed when preimmune sera were used in the immunoprecipitations instead of the specific antisera (data not shown), they presumably repre-sent proteins interacting with ribophorin I and/or RI 332 and calnexin, respectively. For the sequential immunoprecipitations, anti-ribophorin I or anti-calnexin antibodies were ap-FIG. 8. RI 332 remains protease-protected in microsomes prepared from ts20-RI 332 cells grown at the non-permissive temperature. ts20-RI 332 cells were pulse-labeled for 30 min and chased for 10 min (A and B, lanes aЈ-eЈ) or 60 min (B, lanes fЈ-jЈ) at 30°C (A) or 41°C (B). Cells were lysed in an isotonic buffer by Dounce homogenization, and total membrane fractions were divided into 5 equal aliquots. One aliquot was kept as an untreated control (lanes a, aЈ, and fЈ), 3 aliquots were incubated with trypsin (10, 20, or 50 g/ml) in the absence of detergent (lanes b-d, bЈ-dЈ, and gЈ-iЈ), and 1 aliquot was incubated with trypsin (10 g/ml) in the presence of 0.5% Triton X-100 (lanes e, eЈ, and jЈ) for 30 min at 30°C. After addition of TPCK, TLCK, and PMSF, all samples were processed for immunoprecipitation using the anti-ribophorin I antibody. The immunoprecipitates were analyzed on 15% SDSpolyacrylamide gels and fluorography.
FIG. 9. RI 332 interacts with calnexin. ts20-RI 332 cells were pulselabeled for 2 h and chased for 10 or 20 min or 1 h at 30 and 41°C (A and B). E36-RI 332 cells were pulse-labeled for 2 h and chased for 10 or 20 min or 1 h at 30°C in the absence or presence of 50 M ZLLL (C). Cells were lysed in the presence of CHAPS (2%), and anti-calnexin or antiribophorin I immunoprecipitations were performed under non-stringent conditions (n) in the presence of CHAPS (1%), whereas the second steps of co-immunoprecipitations were performed under stringent conditions (s) in the presence of SDS (0.6%) and Triton X-100 (1%). As a control, ribophorin I and RI 332 or calnexin were immunoprecipitated under stringent conditions from the cell lysates. Immunoprecipitates were analyzed by SDS-PAGE and fluorography. A, an experiment is shown for ts20-RI 332 cells at 41°C and at the 10-min chase time point: lanes a and f, anti-ribophorin I and anti-calnexin immunoprecipitations, respectively, under stringent conditions; lanes b and e, antiribophorin I and anti-calnexin immunoprecipitations, respectively, under non-stringent conditions; lanes c and d, co-immunoprecipitations in which first calnexin and ribophorin I, respectively, were immunoprecipitated under non-stringent conditions, followed by anti-ribophorin I and anti-calnexin immunoprecipitations, respectively, under stringent conditions. Analogous experiments were performed for each cell line plied under non-stringent conditions in a first step, followed by precipitations using, respectively, anti-calnexin (lane d) and anti-ribophorin I (lane c) antibodies under stringent conditions. In Fig. 9, B and C, only the ribophorin I immunoprecipitates that were obtained from anti-calnexin immunoprecipitations are shown. In both cell lines at 30°C, RI 332 was hardly detectable (lanes aЈ-cЈ and aЉ-cЉ), even at 10 min of chase, suggesting that only a minute fraction of RI 332 is bound to calnexin under normal conditions. The same is true for ZLLL-treated E36-RI 332 cells, when the proteasome-mediated degradation of the protein is inhibited (lanes dЉ-fЉ). It should be noted, however, that RI 332 was clearly detectable under all the conditions when the protein was precipitated from the lysates with anti-ribophorin I antibodies directly. In contrast, the amount of RI 332 interacting with calnexin is significantly higher in ts20-RI 332 cells at 41°C than in control or ZLLL-treated E36-RI 332 cells and remains essentially constant over the chase period (lanes dЈ-fЈ). It should be pointed out that not only the fully glycosylated form but also the trimmed glycoprotein intermediate of RI 332 (RI 332,i ) is capable of binding to calnexin. In all cases, the endogenous ribophorin I was recovered from the anti-calnexin immunoprecipitates indicating that this ER-resident glycoprotein is recognized by the chaperone.
Recently, Ploegh and co-workers (28) showed that prior to degradation by the ubiquitin-proteasome pathway the heavy chain of MHC class I molecules interacts with the ␤ subunit of the Sec61 complex, one of the major constituents of the translocation apparatus. These data suggest that components of the translocation apparatus are involved in the retro-translocation from the ER lumen to the cytosol of proteins destined to be degraded by the ubiquitin-proteasome pathway. We wished, therefore, to investigate if also RI 332 is able to interact with the ␤ subunit of the Sec61 complex. To this purpose, ts20-RI 332 cells were labeled at 30°C for 30 min and chased for 10 min or 1 h in the presence or absence of ZLLL, or at 41°C only in the absence of ZLLL. Immunoprecipitations with anti-Sec61␤ antibodies, under non-stringent conditions, and reprecipitations with anti-ribophorin I antibodies under stringent conditions were performed (Fig. 10). A considerable amount of RI 332 interacting with Sec61 was recovered from ts20-RI 332 cells incubated at 41°C (lanes e and f), whereas only weak RI 332 bands were observed under the other conditions (lanes a-d). It appears that both the glycosylated and the partially deglycosylated forms of RI 332 interact with Sec61␤. Furthermore, a protein of 65 kDa that corresponds to the endogenous ribophorin I was co-immunoprecipitated with the Sec61␤ subunit. Most likely, this is due to the fact that ribophorin I is localized in close proximity to the translocation apparatus. DISCUSSION Proteasome-mediated and, in most cases, also ubiquitin-dependent degradation has been implicated in the ER-associated proteolysis of several transmembrane proteins, such as CFTR (6,7), hydroxymethylglutaryl-CoA reductase (52), connexin-43 (53), MHC class I heavy chains (28), and the T cell antigen receptor (TCR) ␣ subunit (54,55), but also of luminal proteins, like the PiZ variant of ␣ 1 -AT (9,56), CPY* in yeast (26,27), and apolipoprotein B (57). Prior to degradation, the N-linked oligosaccharide of glycoproteins is removed, as has been observed for MHC class I heavy chains and TCR␣ chains (28,54). In this paper, we demonstrate that RI 332 , a mutant luminal ER glycoprotein, follows a similar degradation pathway, as it is delivered to the ubiquitin-proteasome pathway for proteolysis. Moreover, two different intracellular fates of RI 332 were distinguishable when the degradation of the protein was inhibited. RI 332 molecules remained segregated to the lumen of microsomes as fully glycosylated and partially trimmed forms when ubiquitination was inhibited, whereas the protein accumulated in the cytosol partially or fully deglycosylated after inhibition of the proteasome. Therefore, ubiquitination appears to play an important role in the release of the ER protein into the cytoplasm for degradation.
The glycosylation status of RI 332 deserves further considerations. As RI 332 is efficiently glycosylated and its N-linked glycosylation site occurs only 58 amino acids from the carboxyl terminus, it is very likely that the protein is essentially completely translocated into the lumen of the ER, where it has to undergo quality control processes prior to re-export to the cytosol, deglycosylation, and degradation. Our data indicate that the deglycosylation of RI 332 occurs in at least two distinct steps, since a defined trimmed glycoprotein intermediate was observed. Present evidence suggests that the Nglycanase effecting the deglycosylation reaction is a cytosolic activity (58). Accordingly, it has been reported that glycosylated MHC class I heavy chains as well as TCR␣ chains undergo retro-translocation to the cytoplasm, where they are deglycosylated prior to degradation (28,54). From our cell fractionation experiments it became evident that a small part of the fully glycosylated form of RI 332 is recovered from cytosolic fractions from control and proteasome inhibitor-treated cells, indicating that the protein indeed may exit the ER when it is still glycosylated. On the other hand, completely deglycosylated RI 332 was only detectable in the cytosolic fraction where it accumulates over time but not in the membrane fraction, when the proteasome activity was impaired. The trimmed glycoprotein intermediate of RI 332 , however, was clearly recovered in both fractions and accumulated in the cytoplasm during its lifetime, suggesting that the breakdown of the N-linked oligosaccharide may be initiated when the protein is still associated with the ER membrane. In fact, this conclusion is supported by the observation that the trimming intermediate of RI 332 persists in the microsomal fraction and is found in association with calnexin (see below) when ubiquitination is inhibited in ts20-RI 332 cells at the restrictive temperature. As already pointed out, release of RI 332 into the cytoplasm is compromised under these conditions, reinforcing the concept that partial trimming of the N-linked oligosaccharide of RI 332 may already occur in the lumen of the ER. It is conceivable that this process could be attributed to the activity of the ␣-mannosidases present in the ER that have been shown to trim N-linked oligosaccharides to Man 5 structures (59, 60). FIG. 10. RI 332 interacts with Sec61. ts20-RI 332 cells were pulselabeled for 30 min and chased for 10 min (lanes a and c) or 1 h (lanes b  and d) at 30°C in the absence or presence of ZLLL (50 M) and at 41°C in the absence of ZLLL (lanes e and f). For each condition and chase point a co-immunoprecipitation was performed in which Sec61␤ was immunoprecipitated first under non-stringent conditions in the presence of CHAPS (1%), and in a second round ribophorin I and RI 332 were immunoprecipitated under stringent conditions in the presence of SDS (0.6%) and Triton X-100 (1%). For comparison, ribophorin I and RI 332 were immunoprecipitated under stringent conditions from a cell lysate of ts20-RI 332 cells incubated at 41°C (lane g). Immunoprecipitates were analyzed on Tricine gels followed by fluorography. Only the ribophorin I and RI 332 recovered from Sec61␤ immunoprecipitations are shown. (Exposure times of lanes a-d, 10 days; lanes e-g, 1 day.) Since ribophorin I and RI 332 are N-glycosylated proteins, an interaction of these proteins with the ER chaperone calnexin is expected to occur during an early step in the quality control process to which these proteins are subjected. In fact, such an interaction has been demonstrated for several substrate proteins of ER-associated degradation, such as MHC class I heavy chains (61), CFTR (16), and the PiZ variant of ␣ 1 -AT (9). In our experiments, an association of calnexin with RI 332 was clearly detectable only when the latter one was stabilized. Strikingly, however, large amounts of RI 332 that remained essentially constant over time were found in interaction with the chaperone solely under ubiquitination-deficient conditions. These results are plausible considering that under these conditions retro-translocation is compromised, so that the polypeptide may accumulate in the lumen of the ER where its N-glycan stays accessible for prolonged binding to and release from calnexin. During this time of retention in the ER, UDP-glucose: glycoprotein glucosyltransferase may be involved in the monitoring of the progress the glycoprotein has made in its folding process, reglucosylate its N-linked oligosaccharide as soon as it has lost its remaining glucose residue due to the action of glucosidase II, and thus allow for several rounds of re-association of the glycoprotein with calnexin (62,63).
As to the glycosylation status of RI 332 recovered in complexes with calnexin, it is interesting to note that not only the completely glycosylated protein but also the partially deglycosylated form interacts with the chaperone. This finding is in support of the view discussed above that deglycosylation may be initiated within the ER, whereas the alternative possibility that the trimming intermediate is exposed at the cytoplasmic side of the ER membrane while still interacting with calnexin would be difficult to conceptualize. It should be stressed that the first step of deglycosylation appears to be ubiquitinationindependent, as the partially deglycosylated form of RI 332 is efficiently recovered from ts20-RI 332 cells under restrictive temperature conditions. The interaction of this intermediate form of RI 332 with calnexin could be explained if the partially trimmed oligosaccharide is capable of binding to the chaperone in the lumen of the ER. Partially trimmed high mannose oligosaccharides have indeed been found to be recognized by calnexin and its soluble homolog in the ER lumen, calreticulin (64,65). Recently, it has been proposed that the post-translational trimming of N-linked oligosaccharides on glycoproteins that could be effected by ER mannosidases precedes the release of these proteins from calnexin and their subsequent intracellular degradation (66). A decrease in the degradation rate of CPY* has also been observed in a yeast strain in which the gene encoding the ER-associated ␣1,2-mannosidase has been disrupted (67). Alternatively, it could also be possible that calnexin binds to protein determinants on RI 332 , at least during a later phase of their interaction. Direct recognition of nonglycosylated domains of proteins by calnexin, although previously suggested, has recently been less favored, as it has been shown that the chaperone is able to bind to both folded and unfolded forms of N-glycosylated ribonucleases, which has been taken as strong evidence that calnexin acts exclusively as a lectin (68,69).
Although ubiquitination of RI 332 has been observed during this study, it remains to be determined whether additional factors involved in ER-associated degradation require ubiquitination. In this context, the ubiquitination of calnexin has been implicated in the degradative pathway of the PiZ variant of ␣ 1 -AT (9). From our observation that inhibition of ubiquitination results in prolonged interaction of RI 332 and calnexin as well as impaired retro-translocation to the cytosol, ubiquitination of calnexin could provide a mechanism to trigger the re-lease of the protein from the ER membrane.
During the passage of RI 332 to its site of degradation, it is plausible that the protein is in close contact with the Sec61 complex, one of the major components of the translocation apparatus (70 -72). This step in the degradation pathway of RI 332 may be predicted, as it has been demonstrated that MHC class I heavy chain molecules are co-immunoprecipitated with antibodies directed against the ␤ subunit of Sec61 (28). Further support for a role of Sec61 in the retro-translocation to the cytosol has been recently obtained, when it was shown that misfolded secretory proteins accumulate in the ER of yeast cells that express certain conditional sec61 alleles (29,30). In agreement with these findings, we detected an association of RI 332 with Sec61 in ts20-RI 332 cells. A large portion of RI 332 was recovered from Sec61 immunoprecipitates only when ubiquitination is blocked, whereas the interaction of RI 332 with Sec61 was weakly detectable in control and proteasome inhibitor-treated cells. It appears that both completely glycosylated and partially deglycosylated forms of RI 332 are detected in association with the translocation channel. In accordance with the cell fractionation experiments, these observations indicate that the integrity of the ubiquitination pathway may play a crucial role in the export of proteins from the ER lumen to the cytoplasm. At present, it is not clear, however, whether truly cytoplasmic proteasomes affect the degradation of ER protein substrates in vivo or if proteasome particles associated with the ER membrane that have been detected by immunocytochemical means (73) perform this task. From our observation that RI 332 accumulates in the cytosol of proteasome inhibitortreated cells, the former possibility may seem more likely.
Taken together, our data are in support of the following model for ER-associated degradation of aberrant luminal glycoproteins. After translocation into the ER, signal peptide cleavage, and N-glycosylation, the carbohydrate moiety of the protein is recognized in its monoglucosylated form by the ER chaperone calnexin and possibly its soluble counterpart, calreticulin, which participate in the quality control process the protein is subjected to in the ER lumen. Upon completion of the quality control attempts that may involve several cycles of binding to and release from calnexin/calreticulin due to the cleavage of the remaining glucose residue by glucosidase II and reglucosylation of the N-linked oligosaccharide by UDP-glucose:glycoprotein glucosyltransferase, the protein is retrotranslocated to the cytoplasm via the Sec61 channel of the translocation apparatus. The efficiency of the latter step, i.e. the release of the polypeptide into the cytoplasmic space is strongly dependent on a functional ubiquitination pathway, as in the absence of ubiquitination the protein remains restricted to the lumen of the ER. Once in the cytoplasm, the protein is eventually delivered for degradation by the proteasome. During its metabolic fate, the protein is deglycosylated in two discernible steps, the first of which may occur within the ER lumen leaving the interaction of the protein with calnexin intact, whereas the second step may be accomplished by a cytosolic N-glycanase. drawn from this work. It was reported there that CPY*, a substrate for ER-associated degradation, accumulates in the ER lumen of yeast when Ubc6p-and Ubc7p-mediated ubiquitin conjugation is abolished, thus demonstrating directly the requirement of ubiquitination at the ER membrane for retro-translocation of CPY* to the yeast cytoplasm.