INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Secretory proteins fold in the lumen of the endoplasmic reticulum (ER; Ellgaard et al., 1999). Only fully folded proteins are packaged into ER-to-Golgi transport vesicles and transported through the secretory pathway (Ellgaard et al., 1999). Proteins that fail to acquire their native conformation are retained in the ER, at least initially by interaction with ER-resident chaperones (Ellgaard et al., 1999; Römisch, 1999). Many misfolded proteins are subsequently transported back across the ER membrane to the cytosol where they are degraded by proteasomes (Ellgaard et al., 1999; Römisch, 1999). It remains unclear how the cell distinguishes between chaperone-associated folding intermediates and misfolded proteins and at which point the decision is made to target an aberrant protein to export and degradation (Römisch, 1999; Parodi, 2000). For N-glycosylated misfolded proteins, the "timer" that triggers retrograde transport may be trimming of the N-glycan mannoses (Ellgaard et al., 1999; Cabral et al., 2000; Parodi, 2000). In the cases studied, only mannose-trimmed glycoproteins were subject to retrograde transport and degradation; inhibition of mannose-trimming resulted in the retention of the misfolded glycoproteins in the ER lumen, suggesting that a mannose-specific lectin recognizes export substrates and targets them to degradation (reviewed by Parodi, 2000).

Retrograde protein transport across the ER membrane to the cytosol is mediated by a channel that is formed by the same core component as the protein translocation channel that mediates secretory protein import into the ER lumen (Römisch, 1999). Specific mutations in this protein, Sec61p, interfere differentially with protein import and protein export, and sec61 mutants that specifically block retrograde protein transport have recently been isolated by Zhou and Schekman (Pilon et al., 1997, 1998; Plemper et al., 1997; Zhou and Schekman, 1999). The location of these point mutations suggests that the mutated sites are interaction sites of Sec61p with ER luminal or transmembrane proteins that may be involved in targeting of substrates to the export channel (Römisch, 1999; Zhou and Schekman, 1999). Specific interaction partners of Sec61p during retrograde protein transport across the ER membrane are still unknown.

The heterotrimeric Sec61 complex in the ER membrane consists of Sec61p, a 53-kDa protein with 10 transmembrane domains, and two smaller tail-anchored proteins, Sbh1p, and Ssh1p (Johnson and van Waes, 1999). The Sec61 channel is formed by four or five Sec61 complexes that assemble in the ER membrane for protein import into the ER in response to the presence of a functional signal peptide or in the presence of the Sec63 complex, which is required for posttranslational protein import into the yeast ER (Hanein et al., 1996). The Sec61 channel is sealed at both ends to maintain the permeability barrier across the ER membrane (Hamman et al., 1998). Channel opening for protein import is triggered by functional signal peptides that are recognized by the Sec61 channel itself (Jungnickel and Rapoport, 1995). The signal peptide of most misfolded secretory proteins, however, is cleaved off before retrograde transport is initiated. Channel opening from the ER lumen must therefore be triggered by a fundamentally different mechanism (Römisch, 1999).

To identify a targeting signal for retrograde protein transport, we asked whether mutant secretory proteins are covalently modified in the ER lumen before export. As a substrate we used a mutant form of the yeast pheromone precursor prepro alpha-factor, which had its N-glycosylation acceptor sites removed by site-directed mutagenesis (Mayinger and Meyer, 1993). The N-glycosylation site mutant alpha-factor precursor protein (gpf) is subject to ER-associated degradation in vivo and in a cell-free assay based on yeast microsomes and cytosol (McCracken and Brodsky, 1996). We found that a fraction of gpf, but not the wild-type precursor, was covalently modified in microsomes and in intact yeast cells. This modification was significantly enhanced in vitro under ER export conditions, i.e., in the presence of ATP and cytosol, and required the presence of export-competent Sec61p in the membranes. We identified the modification as O-mannosylation and Pmt2p in the ER as the responsible enzyme. Deletion of the corresponding gene, PMT2, however, did not abrogate gpf degradation but rather increased its turnover both in vitro and in vivo. In wild-type microsomes, the O-mannosylated fraction of gpf was stable, protease protected, and accumulated in the ER lumen over the time course of a degradation reaction. A fraction of mgpf could be cross-linked to Sec61p. Our data suggest that O-mannosylation of gpf interferes with its retrograde transport across the ER membrane at a posttargeting stage.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Strains and Growth Conditions

The following strains were used: RSY255 (MAT leu2-3,-112 ura3-52; (Stirling et al., 1992), RSY1293 (MAT can1-100 his3-11,-15 leu2-3,-112 trp1-1 ura3-1 ade2-1 sec61::HIS3 pDQ1[sec61-his6] or pDQ1 expressing the indicated sec61 mutants (Pilon et al., 1997), WCG4a (MATa leu2-3,-112 ura3 his3-11,-15 (Hiller et al., 1996), WCG4-2 (MATa leu2-3,-112 ura3 his3-11,-15 pre1 pre2 [Hiller et al., 1996]); RSY281 (MAT sec23-1 ura3-52 his4-619 [Hicke et al., 1992]); SEY6210 (MAT ura3-52 leu2-3,-112 his3-200 trp1-901 lys2-801, suc2-9 [Gentzsch and Tanner, 1996]); SEY6211 (MAT a ura3-52 leu2-3,-112 his3-200 trp1-901 ade2-101, suc2-9 [Gentzsch and Tanner, 1996]); SEY6210 pmt5::URA3, SEY6210 pmt6::URA3, SEY6210 pmt2::LEU2; all pmt double and triple mutants used were progeny from crosses of SEY6210 pmt3::HIS3 pmt4::TRP1 with SEY6211 pmt1::URA3 pmt2::LEU2 (Gentzsch and Tanner, 1996). TF1.8 (MATa ura3 leu2 his3 GAL1-PMT1-LEU2 (Gentzsch et al., 1995) transformed with YEp352 or YEp352[PMT2]. Yeast were grown in YPD (1% yeast extract, 2% peptone [Difco, Detroit, MI], 2% dextrose) or synthetic media with the appropriate additions (Sherman, 1991).

Pulse-Chase Experiments

Cells were radiolabeled and lysed, and proteins were immunoprecipitated or precipitated with concanavalin A (ConA)-Sepharose as in described by Gillece et al. (1999). Lectin-precipitated samples were eluted with 1 M -methyl mannoside in 20 mM Tris-HCl, pH7.5, 150 mM NaCl, 2 mM EDTA for 1 h at 30°C. Where indicated, samples were treated with 2 mU peptide N-glycosidase (PNGase) F (Roche Diagnostics Ltd., Lewes, United Kingdom) for 2 h at 37°C before electrophoresis.

[³H]Mannose Labeling

The method used was modified from Orlean et al. (1991). Cells were grown to OD₆₀₀ = 0.5 in full or selective medium with 0.6% sucrose as a carbon source, washed once in fresh medium labeled with 1 mCi of [³H]mannose (Amersham, Arlington Heights, IL) per 1.5 OD₆₀₀ of cells in 500 µl of medium for 90 min at the indicated temperature. Cells were lysed, and proteins were immunoprecipitated as above, resolved on 16% or 7.5% SDS-PAGE, and detected by fluorography.

Cell Fractionation

Microsomes for in vitro degradation assays were prepared from cells grown to OD₆₀₀ = 1 as described by Pilon et al. (1997). Microsomes from PMT wild-type and mutant strains for the experiment shown in Figure 3A were prepared using the same method, but the sucrose gradient purification was omitted. Cytosol was prepared from WCG4a by liquid nitrogen lysis as described by Pilon et al. (1997).

Translocation and ConA Precipitation

In vitro translated, [³⁵S]methionine-labeled nonglycosylated alpha-factor precursor (pgpf) or wild-type precursor (ppf) were translocated into wild-type or pmt mutant microsomes for the indicated periods of time in the presence of ATP and an ATP-regenerating system (40 mM creatine phosphate, 0.2 mg/ml creatine phosphokinase, 1 mM ATP, 50 µM GDP-mannose) in B88 (20 mM HEPES, pH 6.8, 150 mM potassium acetate, 5 mM magnesium acetate, 250 mM sorbitol) at 24°C. At the end of the incubation period, SDS was added to 1%, samples were heated to 95°C for 5 min, and mannosylated signal-cleaved gpf precipitated with ConA-Sepharose (Pharmacia, Piscataway, NJ). Wild-type pf was immunoprecipitated, deN-glycosylated with PNGase F (Boehringer), and subsequently precipitated with ConA-Sepharose.

ER Degradation Assay and Cross-Linking

ER degradation of gpf was assayed at 24°C as described by Pilon et al. (1997). Briefly, 20-µl translocation reactions contained 2 µl of microsomes of OD₂₈₀ = 30, B88, ATP, a regenerating system, and 2 µl of in vitro translated, ³⁵S-labeled pgpf (500,000 cpm). Translocation reactions were incubated for 50 min at 24°C, and the membranes were washed twice in B88. Membranes containing gpf were resuspended in B88 with ATP and the regenerating system, and degradation reactions were started by adding cytosol to 6 mg/ml final concentration in a 20-µl/reaction final volume. Degradation reactions were incubated at 24°C for the indicated periods of time. At the end of the incubation, samples were precipitated with trichloroacetic acid (TCA) and analyzed after electrophoresis on 18% polyacrylamide, 4 M urea SDS gels with a Cyclone phosphorimager (Hewlett-Packard, Bracknell, United Kingdom). The gpf and mannosylated gpf (mgpf) bands were quantified and expressed as percentages of gpaf at the beginning of the reaction (time 0). Cross-linking was performed with DSP as described by Pilon et al. (1997). Individual samples were 10× scaled up reactions as described above. Cross-linking was initiated after 10 min of incubation in the presence of ATP, the regenerating system, and 6 mg/ml yeast cytosol at 24°C.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Mutant Alpha-Factor Precursor Is O-Mannosylated in the ER

Misfolded secretory proteins are exported from the ER to the cytosol through the Sec61 channel after their signal sequences have been cleaved off (Römisch, 1999). We therefore asked whether misfolded proteins in the ER acquire a covalent modification that serves as an export signal during prolonged residence in the ER lumen. As substrate proteins we used [³⁵S]methionine-labeled, in vitro translated wild-type alpha-factor precursor (ppf) or a mutant counterpart in which the three N-glycosylation acceptor sites in the proregion had been destroyed by site-directed mutagenesis (pgpf) (Mayinger and Meyer, 1993; McCracken and Brodsky, 1996). In the presence of ATP and an ATP-regenerating system, both proteins can be translocated into yeast microsomes posttranslationally, resulting in signal cleaved mutant alpha-factor precursor (gpf) and signal-cleaved, triply N-glycosylated wild-type alpha-factor precursor (3gpf), respectively. Posttranslational protein import into the ER is essentially complete after a 10-min incubation at 24°C; nevertheless, we had observed previously that retrograde transport of a mutant secretory protein from the ER in vitro was more efficient after extended import reactions (K.R., unpublished data). We therefore suspected that the export substrate might be modified during its prolonged residence in the ER lumen. When we incubated yeast microsomes containing gpf for 50 min at 24°C, we detected a small increase in molecular weight of a fraction of gpf (mgpf, Figure 1A, compare lanes 1, 2 and 5, 6). Concomitantly with the molecular weight shift, this form of gpf acquired an affinity for the mannose-specific lectin ConA (mgpf, Figure 1A, compare lanes 3, 4 and 7, 8). Because gpf no longer contains any N-glycosylation sites, these data suggest that the protein might be O-mannosylated. In yeast, protein O-mannosylation is initiated in the ER lumen by the transfer of a single mannosyl residue from dolichol-P-mannose to specific serine and threonine residues of the substrate (Strahl-Bolsinger et al., 1999). So far, O-mannosylation of alpha-factor precursor has not been reported. To investigate whether the wild-type precursor acquired the same modification, after incubation of yeast microsomes containing wild-type 3gpf for 50 min at 24°C, we removed the N-glycans of the wild-type precursor with PNGase F before precipitation with ConA-Sepharose (Figure 1A, lanes 9-12). We found that none of de-N-glycosylated wild-type precursor (pf) bound to the lectin (Figure 1A, lanes 11, 12 versus 13, 14). We conclude that the mutant, but not the wild-type precursor, is likely to be O-mannosylated after prolonged residence in the ER lumen in vitro.

View larger version (41K): [in this window] [in a new window]   Figure 1.   Mutant alpha-factor precursor is O-mannosylated in the ER. (A) Prolonged ER residence of gpf in vitro results in O-mannosylation. In vitro translated, [35S]methionine-labeled pgpf (top) or ppf (bottom) was translocated into wild-type yeast microsomes at 24°C in the presence of ATP and an ATP-regenerating system for the indicated periods of time. Duplicate samples were either TCA precipitated or membranes lysed and proteins precipitated directly with ConA-Sepharose (lanes 3, 4, 7, and 8) or first treated with PNGase and then precipitated with ConA-Sepharose as indicated (lanes 13 and 14). Proteins were analyzed by SDS-PAGE on 18% 4 M urea gels and autoradiography. Note that the nonspecific binding of ppf to ConA-Sepharose is higher than that of pgpf; this is an intrinsic property of ppf. (B) gpf is O-mannosylated in vivo. RSY281 (sec23 ts) was labeled with [3H]mannose for 90 min at the permissive (24°C) or at the restrictive (37°C) temperature. Cells were lysed, and CPY and alpha-factor precursor were immunoprecipitated. Alpha-factor immunoprecipitates shown in lanes 2, 4, and 6 were PNGase digested before SDS-PAGE and autoradiography. Note that a small amount of 1gpf in lane 4 was refractory to PNGase digestion.

We next asked whether gpf was also modified in intact yeast cells and whether again the modification was specific to the mutant precursor. To this end, we labeled intact yeast cells with [³H]mannose for 90 min. Steady-state labeling is required to be able to detect mannose incorporation into secretory proteins (Orlean et al., 1991). We found, however, that at steady state there was no detectable amount of wild-type alpha-factor precursor present inside the cells, even if the protein was overexpressed from a 2µ plasmid, because of its efficient processing and secretion. To compare wild-type and mutant alpha-factor precursor in vivo, we therefore performed the experiment in a strain that carries a temperature-sensitive mutation in Sec23p, a protein essential for ER-to-Golgi transport vesicle budding from the ER (RSY281; Hicke et al., 1992). After [³H]mannose labeling of RSY281-overexpressing ppf at the permissive temperature (24°C), as in wild-type cells we were unable to detect intracellular wild-type alpha-factor precursor because of its rapid transport to the cell surface (Figure 1B, lanes 1 and 2). In contrast, after [³H]mannose labeling of RSY281-expressing pgpf at 24°C, we were able to immunoprecipitate a mannose-labeled form of the mutant precursor that was refractory to PNGase digestion and that migrated in the appropriate position for mgpf (Figure 1B, lanes 5 and 6). To investigate whether the wild-type precursor could be O-mannosylated in vivo if its residence time in the ER was increased, we performed the [³H]mannose labeling of RSY281 overexpressing ppf at the restrictive temperature for vesicle budding from the ER (37°C). At 37°C, ER-to-Golgi transport in RSY281 ceases, as evident from the accumulation of the ER-specific form of carboxypeptidase Y in the cells (p1CPY, Figure 1B, bottom, compare lanes 24 and 37). Under these conditions we were indeed able to detect intracellular, mannose-labeled forms of wild-type alpha-factor precursor (Figure 1B, lane 3). The majority of the incorporated [³H]mannose, however, could be removed from the wild-type precursor by PNGase F digestion, suggesting that the label had been primarily incorporated into the N-glycans of singly, doubly, and triply N-glycosylated wild-type precursor (1gpf, 2gpf, 3gpf, Figure 1B, compare lanes 3 and 4). The small fraction of wild-type precursor whose position in the gel suggested that it had not been N-glycosylated after translocation into the ER, however, incorporated [³H]mannose in a PNGase-resistant manner like the mutant precursor (mgpf, Figure 1B, lanes 3 and 4). Our data suggest that in the absence of N-glycans alpha-factor precursor is O-mannosylated during prolonged residence in the ER.

O-Mannosylation of Mutant Alpha-Factor Precursor In Vitro Is Stimulated by ATP and Cytosol and Requires Export-Proficient Sec61p

Export of misfolded secretory proteins from the ER through the Sec61 channel and their degradation by proteasomes can be reconstituted in a cell-free system based on yeast microsomes and cytosol (McCracken and Brodsky, 1996). Incubation of wild-type yeast microsomes containing gpf in the presence of ATP, an ATP-regenerating system, and 6 mg/ml wild-type yeast cytosol results in a disappearance of the gpf band with a half life of ~12 min (Figure 2, left; McCracken and Brodsky, 1996). Concomitantly with the decrease in gpf, we also observed a further molecular weight increase of a fraction of gpf (mgpf, Figure 2, left). This molecular weight shift was due to an increased mannosylation of gpf under export conditions, as demonstrated by ConA precipitation of the newly appearing bands (mgpf, Figure 4, top). Like gpf export and degradation, increased O-mannosylation was dependent on the presence of ATP and cytosol (Figure 2, left). In contrast to gpf degradation, however, O-mannosylation did not require functional proteasomes in the cytosol (Figure 2, right). Surprisingly, we found that O-mannosylation of gpf in the ER was dramatically reduced in microsomes derived from export-deficient sec61 mutant strains, suggesting that the O-mannosylation of this substrate required at least the initiation of its export to the cytosol (Figure 2, middle, compare SEC61 to sec61-32 and sec61-41; Pilon et al., 1997). We conclude that, in vitro, conditions that promote misfolded protein export from the ER stimulate O-mannosylation of mutant alpha-factor precursor.

View larger version (22K): [in this window] [in a new window]   Figure 2.   O-mannosylation of gpf in vitro is stimulated by ATP and cytosol and requires export-proficient Sec61p. Left, in vitro translated, [35S]methionine-labeled pgpf was translocated into wild-type yeast microsomes, and the washed membranes were incubated at 24°C for the indicated periods of time (minutes) in the presence of ATP and an ATP-regenerating system, cytosol or both. At the end of the incubation period, samples were TCA precipitated and analyzed by SDS-PAGE and autoradiography. gpf and mgpf were quantified using a phophorimager and expressed as percentages of gpf at 0 min. Note that a fraction of gpf is degraded after 60 min, even in the absence of cytosol; the degree of cytosol-independent degradation varies between microsome preparations and is likely due to proteasomes associated with the cytoplasmic faces of the microsomes. Middle, samples contained SEC61 wild-type or the indicated mutant microsomes and were incubated for 20 min in the presence of ATP and cytosol before TCA precipitation. Right, samples contained wild-type microsomes and were incubated with ATP and either wild-type (PRE) or proteasome-mutant (pre1pre2) cytosol for 0 or 20 min before TCA precipitation.

Pmt2p Is Responsible for Mutant Alpha-Factor Precursor O-Mannosylation but Not Required for Its Degradation

Based on the stimulation of gpf O-mannosylation under ER export conditions and its SEC61 dependence, we assumed that O-mannosylation was intimately linked to gpf export through the Sec61 channel and that the O-mannosylated bands might be export intermediates. We therefore sought to identify the protein O-mannosyl transferase (Pmt) responsible for gpf O-mannosylation. There are seven PMT genes in the Saccharomyces cerevisiae genome. For Pmt1p-Pmt4p and Pmt6p, mannosyl transferase activity has been demonstrated (Strahl-Bolsinger et al., 1999). The individual transferases show distinct specificities toward their protein substrates (Gentzsch and Tanner, 1996, 1997). The active sites of the Pmts are predicted to be on the luminal face of the ER membrane (Girrbach et al., 2000). The transferases Pmt1p and Pmt2p have been shown to act as heterodimers in vitro and in vivo (Girrbach et al., 2000). We prepared microsomes from all viable pmt1-4 double and triple mutants and from strains with individual deletions of PMT5 and PMT6 (Gentzsch and Tanner, 1996); we subsequently translocated pgpf into wild-type and pmt mutant microsomes for 50 min at 24°C, lysed the membranes, and assessed gpf mannosylation by precipitation with ConA-Sepharose. We found that all mutants with a deletion of PMT2 were defective in gpf O-mannosylation (Figure 3A). Deletion of PMT2 on its own also resulted in lack of gpf O-mannosylation, and no stimulation of gpf O-mannosylation was seen in the presence of ATP and cytosol in pmt2 microsomes.

View larger version (27K): [in this window] [in a new window]   Figure 3.   Pmt2p is required for gpf mannosylation but not for its degradation. (A) Microsomes were prepared from PMT wild-type SEY6210 (lane 1) and SEY6211 (lane 2) and the indicated mutant strains. Equal amounts of in vitro translated, [35S]methionine-labeled pgpf were translocated into wild-type yeast microsomes at 24°C in the presence of ATP and an ATP-regenerating system for 50 min, followed by membrane lysis and ConA precipitation. Translocation efficiencies were similar for all microsome preparations tested. Lectin-bound material was analyzed by SDS-PAGE and autoradiography. (B) PMT2 wild-type and pmt2 cells expressing pgpf were pulse-labeled for 5 min with [35S]methionine/cysteine and chased for the indicated periods of time. At each time point, cells were lysed and alpha-factor precursor immunoprecipitated and quantified by SDS-PAGE and phosphorimager analysis. The experiment was repeated twice.

We then investigated whether O-mannosylation was required for gpf degradation. We transformed PMT2 wild-type and pmt2 strains with a plasmid expressing pgpf and performed pulse-chase experiments. As shown in Figure 3B, we found that mannosylation by Pmt2p was dispensable for gpf turnover and that, in fact, the half-life of gpf was slightly reduced in the pmt2 strain, suggesting that O-mannosylation of gpf interfered with its degradation (Figure 3B, half-time of gpf in PMT2 = 12 min, in pmt2 =10 min). This observation was confirmed in vitro where we observed a slightly reduced half-life of gpf in pmt2pmt4 mutant microsomes. The expression levels of pgpf were similar in PMT2 wild-type and pmt2 cells, but by quantitative immunoblotting we found that the increased turnover of gpf in pmt2 cells resulted in a reduction to ~50% of intracellular gpf compared with wild-type cells. We observed no effects of PMT2 deletion on vesicular transport of wild-type or mutant alpha-factor precursor and CPY and no effects on alpha-factor secretion. Deletion of PMT2 also did not affect the ER export and degradation of a mutant form of CPY, CPY*; nor did we observe O-mannosylation of CPY*. We conclude that gpf is specifically O-mannosylated by Pmt2p in the ER and that this modification is not required for gpf degradation.

O-Mannosylation Protects Mutant Alpha-Factor Precursor from Export to the Cytosol

The increased turnover of gpf in pmt2 cells raised the possibility that mannosylation interfered with the degradation of gpf. We therefore investigated the fate of mgpf during a degradation time course in vitro. We incubated wild-type microsomes containing gpf in the presence of ATP, an ATP-regenerating system, and 6 mg/ml wild-type cytosol for up to 60 min at 24°C and at each time point precipitated one aliquot with TCA and ConA-precipitated mgpf from twice the amount of material. We found that over time gpf was increasingly O-mannosylated, which resulted in an increased molecular weight close to the position on pgpf on 18% polyacrylamide 4 M urea gels (Figure 4A, top). The highly O-mannosylated form was not subject to degradation, however, but accumulated during the time course of the reaction (Figure 4A, top).

View larger version (29K): [in this window] [in a new window]   Figure 4.   O-mannosylation protects gpf from export to the cytosol. (A) Wild-type yeast microsomes containing gpf were incubated in the presence of ATP, an ATP-regenerating system, and 6 mg/ml wild-type yeast cytosol at 24°C for the indicated periods (minutes). Top, At each time point, samples were either TCA precipitated or membranes lysed and mgpf ConA precipitated; note that lectin precipitation was done from twice the amount of material as the TCA precipitation. Bottom, at each time point, samples were transferred to ice and either mock incubated (PK) or incubated with 0.1 mg/ml proteinase K for 20 min (+PK) before TCA precipitation and gel electrophoresis. (B) Top, microsomes from PMT1/2-overexpressing cells containing gpf were incubated as in A and mgpf ConA precipitated. Bottom, wild-type, PMT1/2-overexpressing, and pmt2/4 microsomes containing gpf were incubated the presence of ATP, an ATP-regenerating system, and 6 mg/ml wild-type yeast cytosol at 24°C for 10 min; proteins were cross-linked by the addition of DSP, membranes were lysed, and Sec61p and associated proteins were immunoprecipitated. Cross-links were cleaved with dithiothreitol before gel electrophoresis. ConA and TCA precipitates of 10% of the material used for cross-linking are shown in the middle and right. All samples were run on the same gel. Left panel was exposed 10× longer than the right panel. The reason for the faster migrating bands below gpf in the pmt2/4 sample is unknown but specific to this strain.

To clarify whether mannosylation interfered with degradation of gpf by cytosolic proteasomes, or whether the modification prevented export of mgpf from the ER lumen to the cytosol, we performed protease protection experiments. At each time point during a degradation reaction. we transferred samples to ice and digested them with 0.1 mg/ml proteinase K for 20 min. We found that the mannosylated forms of gpf were refractory to proteinase K digestion in contrast to pgpf, which is associated with the cytosolic face of the microsomes and was protease sensitive (Figure 4A, bottom). In the presence of 0.1% Triton X-100, mgpf was fully digested by proteinase K. Our data suggest that mgpf resides in the lumen of the ER and, thus, that O-mannosylation of gpf interferes with its export through the Sec61 channel to the cytosol.

We next asked whether overexpression of PMT2 would lead to retention of a higher proportion of gpf in the ER. We found that to increase Pmt2p activity, we needed to co-overexpress PMT1 and PMT2. Expression of PMT2 from a 2µ plasmid in a strain that had PMT1 integrated under control of the GAL1 promoter resulted in a threefold increase of mannosyl-transferase activity in vitro (S.S., unpublished data). In microsomes derived from this strain, gpf was maximally O-mannosylated with a half-life of <10 min (Figure 4B, top) in contrast to wild-type microsomes, in which the maximally O-mannosylated form appeared with a half-life of 20-30 min (Figure 4A, top, ConA). The proportion of gpf that was O-mannosylated did not change significantly upon PMT1/2 overexpression, however, and consistent with this observation there was no effect of turnover of gpf (Figure 4, A and B, top, and data not shown).

We suspected that O-mannosylation might stabilize a nonproductive interaction of mgpf with the Sec61 channel, which would thus lead to increased mannosylation if higher Pmt2 activity was present in the ER membrane. We investigated whether mgpf could be cross-linked to Sec61p. The cross-linking efficiency of gpf to Sec61p is low (1-2%; Pilon et al., 1997), and the maximal proportion of gpf that was mannosylated in the ER was <20% (Figure 2). To maximize our chances for detecting mgpf interacting with Sec61p, we performed the cross-linking experiment in microsomes derived from PMT1/2-overexpressing cells, in which most mgpf is present as a single, highly mannosylated band that migrates close to the signal sequence containing pgpf in our gel system (Figure 4A, bottom). Wild-type microsomes and microsomes derived from a pmt2/4 deletion strain, in which gpf mannosylation is reduced (Figure 3A), were included as controls. Mutant alpha-factor precursor was translocated into the microsomes; the membranes were washed and incubated with ATP, an ATP-regenerating system, and 6 mg/ml cytosol for 10 min at 24°C before cross-linking was initiated. Membranes were lysed, and Sec61p and associated proteins were immunoprecipitated with affinity-purified Sec61 antibodies. Cross-links were cleaved with dithiothreitol before electrophoresis on 18% 4 M urea SDS gels. The amount of signal-cleaved gpf associated with Sec61p was proportional to the amount of gpf in the ER lumen and identical for wild-type and PMT1/2-overexpressing microsomes (Figure 4B, bottom, compare X-link and TCA). In wild-type and pmt2/4 microsomes, a small amount of cytosolic pgpf was found cross-linked to Sec61p (Figure 4B, bottom). In PMT1/2-overexpressing microsomes, the intensity of this upper band was increased approximately threefold (Figure 4B, bottom, X-link). This increase was dependent on the presence of cross-linking reagent and Sec61 antibodies and corresponds well to the increased amount of maximally mannosylated spf present in these microsomes (ConA, bottom, Figure 4B). Sec61p itself is not glycosylated, but a fraction of the Sec61 complex in the ER membrane is bound to the glycoprotein-containing Sec63 complex. We were therefore unable to determine the association of Sec61p with mgpf by lectin precipitation. Because we observed no change in the amount of pgpf associated with the cytoplasmic face of PMT1/2-overexpressing microsomes, however, we conclude that the upper band found associated with Sec61p in the cross-linking experiment in PMT1/2-overexpressing microsomes consists primarily of highly mannosylated gpf. Interestingly, this interaction was specific for the highly mannosylated form of gpf; the fast migrating mgpf band could not be cross-linked to Sec61p (Figure 4B, bottom, ConA versus X-link).

Competition between N-Glycosylation and O-Mannosylation in the ER

Do N-linked and O-linked glycosylation compete in proteins other than alpha-factor precursor? N-glycosylation acceptor sites are well defined (N-X-S/T), but there is currently no known consensus acceptor site for O-linked mannosylation; thus, the general proximity or the degree of overlap between the two types of acceptor sites could not be determined by studying the available protein data bases (Strahl-Bolsinger et al., 1999). We therefore addressed the question experimentally. We metabolically labeled wild-type yeast with [³⁵S]methionine/cysteine for 10 min at 30°C in the absence or presence of the N-glycosylation inhibitor tunicamycin. Cells were lysed, and a protein with five N-glycosylation acceptor sites, protein disulfide isomerase (PDI), was immunoprecipitated. Immunocomplexes were digested with PNGase F or mock incubated, followed by precipitation with the mannose-specific lectin ConA. In the absence of tunicamycin, PDI is heterogeneously N-glycosylated on four or five of its five acceptor sites and therefore binds to ConA (Figure 5, top, lanes 1 and 2; gPDI). PNGase treatment removes the N-glycosyl side chains and results in complete loss of affinity of PDI for the lectin (Figure 5, top, lanes 3 and 4). PDI isolated from tunicamycin-treated cells migrates at a lower mobility, consistent with lack of N-glycosyl side chains, but a fraction of non-N-glycosylated PDI nevertheless binds to ConA (Figure 5, top, lanes 5 and 6; mPDI). Digestion of PDI isolated from tunicamycin-treated cells with PNGase does not alter its mobility on the gel or its ability to bind to ConA (Figure 5, top, lanes 7 and 8; mPDI). Our data suggest that like alpha-factor precursor PDI can be O-mannosylated in the absence of N-glycosylation.

View larger version (82K): [in this window] [in a new window]   Figure 5.   Competition between N-glycosylation and O-mannosylation in the ER. Wild-type cells were incubated in the presence (+) or absence () of 10 µg/ml tunicamycin (tuni) for 20 min before radiolabeling for 10 min with [35S]methionine/cysteine. Cells were lysed and proteins either immunoprecipitated (IP) with a polyclonal anti-PDI antiserum (top) or ConA precipitated (bottom). Precipitates were digested with PNGase or mock incubated, followed by a second round of ConA precipitation. Proteins were eluted from ConA-Sepharose with -methyl mannoside before electrophoresis on 10% (top) and 12.5% gels (bottom).

How general is the competition of N-glycosylation and O-mannosylation in the ER? To answer this question, we labeled wild-type cells in the presence or absence of tunicamycin, as above, and precipitated glycoproteins with ConA-Sepharose. Lectin precipitates were PNGase treated or mock incubated and subsequently subjected to a second round of ConA precipitation. We found that an increased proportion of glycoproteins isolated from tunicamycin-treated cells bound to ConA-Sepharose in a PNGase-resistant manner (Figure 5, bottom, compare lanes 3, 4 to 7, 8; 35% increase in total signal). Our data confirm that the competition of N- and O-linked glycosylation that we documented for two specific proteins, alpha-factor precursor and PDI, is a general phenomenon in the ER that affects a large number of proteins.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

We have shown that a misfolded secretory protein, gpf, is specifically O-mannosylated in the ER lumen by Pmt2p. This O-mannosylation is stimulated in vitro by the presence of ATP and cytosol and requires functional Sec61p, suggesting that O-mannosylation is linked to the initiation of misfolded protein export from the ER through the Sec61 channel. The O-mannosylated fraction of gpf, however, is not transported to the cytosol but remains protease protected in the ER lumen and is therefore stable. In addition, a fraction of mgpf is associated with Sec61p. Our work demonstrates that there is a limited window of time during which misfolded proteins can be removed from the ER before they acquire inappropriate modifications that can interfere with protein disposal through the Sec61 channel.

The yeast ER contains several O-mannosyl transferases with distinct substrate specificities (Gentzsch and Tanner, 1996, 1997). The basis of substrate recognition has been difficult to characterize; the correlation between data derived from peptide substrates and protein substrates is poor, suggesting that protein conformation contributes significantly to the recognition of O-mannosyl acceptor sites by Pmts (Strahl-Bolsinger et al., 1999). Wild-type alpha-factor precursor is N-glycosylated at three sites in its proregion (Kurjan and Herskowitz, 1982). Conversion of the asparagine residues in the N-glycosyl acceptor sites to glutamine residues results in a protein that is recognized in the ER as misfolded and subsequently transported to the cytosol for disposal by the proteasomes (Mayinger and Meyer, 1993; McCracken and Brodsky, 1996). Most serine and threonine residues in alpha-factor precursor that could serve as O-mannosyl acceptors are clustered around the N-glycosylation sites (Kurjan and Herskowitz, 1982). Our finding that only non-N-glycosylated alpha-factor precursor is O-mannosylated suggests that access to the O-mannosyl acceptor sites in the N-glycosylated precursor is sterically blocked by the N-glycans. Alternatively, N-glycosylation may introduce a conformational change into the gpf proregion that prevents its O-mannosylation by Pmt2p.

N-glycosylation of wild-type alpha-factor precursor is rapid and efficient and is initiated during import into the ER (Figure 1A, K.R. unpublished data). By contrast, O-mannosylation of gpf is inefficient and slow (Figure 1A). The fact that significant O-glycosylation occurs only after prolonged incubation of yeast microsomes containing gpf at physiological temperature suggests that it is a posttranslocational event (Figure 1A). Similarly, a polytopic ER membrane protein fused to a degradation signal, Deg1-Hmg1p, is only glycosylated under conditions that prevent its degradation and therefore prolong its residence in the ER (Wilhovsky et al., 2000); the nature of the glycosylation, however, was not identified in this case. The addition of ATP and cytosol, which promote targeting of gpf to the Sec61 channel for export, stimulates gpf O-mannosylation (Figure 2). This increased O-mannosylation is dependent on export-competent Sec61 channels, suggesting that the substrate may be in contact with the export channel when it is mannosylated (Figure 2). This was confirmed by cross-linking mgpf to Sec61p (Figure 4B). The Pmts in the ER are polytopic transmembrane proteins with large ER-luminal domains containing the active sites (Girrbach et al., 2000). It is therefore conceivable that Pmts modify membrane-associated substrates more efficiently; Pmts may even be located in close proximity to the translocon, which would ensure that wild-type secretory proteins with O-mannosyl acceptor sites are modified efficiently and without interference from protein folding during entry into the ER through the Sec61 channel. This notion is supported by the observation that in vitro only short peptides and partially hydrolyzed proteins are modified by Pmts, but fully folded proteins are not substrates for O-mannosylation ((Strahl-Bolsinger et al., 1999; S.S., unpublished data). In their preference for unfolded substrates, Pmts resemble glucosyl transferase in the ER lumen, which specifically tags unfolded proteins (Parodi, 2000).

Because alpha-factor precursor is not normally a substrate for O-mannosylation, it is possible that its orientation in the translocation channel during import is incompatible with O-mannosylation but that it exposes the appropriate sites during initiation of export (Figures 1 and 2). An alternative explanation may be that import of pgpf into the ER is extremely rapid and efficient, whereas export proceeds with much slower kinetics and may therefore result in prolonged exposure of O-mannosyl acceptor sites in the appropriate vicinity of Pmt2p. A third possibility is that the Sec63 complex associated with the Sec61 channel during posttranslational pgpf import into the ER interferes with access of Pmt2p to proteins in the translocon (Panzner et al., 1995). The Sec63 complex is not required for misfolded protein export from the ER and thus most likely absent from Sec61 channels engaged in export, which may allow mannosylation of export substrates by Pmt2p (Pilon et al., 1997).

We have demonstrated that mgpf remains in the ER lumen, which suggests that the O-mannosyl moieties prevent export through the Sec61 channel (Figure 4). O-mannosylation is largely Sec61 dependent, suggesting that the modification occurs after the export substrate has made contact with the Sec61 channel (Figures 2 and 4B); O-mannosylation therefore aborts export at a posttargeting step (Figure 4B). Protease digestion of microsomes containing mannosyl-gpf did not result in any loss of signal or in the occurrence of partially protease-protected intermediate bands (Figure 4A); because our gel system resolves size differences of <10 amino acids, our data suggest that mgpf is still fully contained inside the ER lumen. It is so far unknown whether proteins insert into the Sec61 channel for export with their N or their C termini first. The lack of protease-protected, O-mannosylated export intermediates combined with the fact that all serine and threonine residues of gpf that can serve as O-mannosyl acceptor sites are contained in the proregion of the proteinthe first is close to the extreme N terminus of the signal-cleaved protein at amino acid position 5may suggest that the N terminus of gpf interacts with the Sec61 channel first and that O-mannosylation of the N terminus aborts its subsequent insertion into the channel (Figure 4; Kurjan and Herskowitz, 1982). Given that misfolded proteins with several N-glycans attached, such as CPY*, can be exported through the Sec61 channel to the cytosol, we consider it unlikely that O-mannosylation sterically interferes with protein transport through the channel (Plemper et al., 1997). O-mannosylation may, however, introduce a conformational change into or increase the rigidity of the alpha-factor precursor proregion and thus abort export of the protein. One intriguing alternative explanation for the effect of O-mannosylation on mutant alpha-factor precursor export is that the O-mannosyl acceptor sites in gpf precursor overlap with an essential recognition motif for misfolded protein transport through the Sec61 channel to the cytosol, which may be recognized by the so far uncharacterized machinery that drives misfolded protein export from the ER.