Roles of O-Mannosylation of Aberrant Proteins in Reduction of the Load for Endoplasmic Reticulum Chaperones in Yeast*

The protein quality control system in the endoplasmic reticulum (ER) ensures that only properly folded proteins are deployed throughout the cells. When nonnative proteins accumulate in the ER, the unfolded protein response is triggered to limit further accumulation of nonnative proteins and the ER is cleared of accumulated nonnative proteins by the ER-associated degradation (ERAD). In the yeast ER, aberrant nonnative proteins are mainly directed for the ERAD, but a distinct fraction of them instead receive O-mannosylation. In order to test whether O-mannosylation might also be a mechanism to process aberrant proteins in the ER, here we analyzed the effect of O-mannosylation on two kinds of model aberrant proteins, a series of N-glycosylation site mutants of prepro-α-factor and a pro-region-deleted derivative of Rhizopus niveus aspartic proteinase-I (Δpro) both in vitro and in vivo. O-Mannosylation increases solubilities of the aberrant proteins and renders them less dependent on the ER chaperone, BiP, for being soluble. The release from ER chaperones allows the aberrant proteins to exit out of the ER for the normal secretory pathway transport. When the gene for Pmt2p, responsible for the O-mannosylation of these aberrant proteins, and that for the ERAD were simultaneously deleted, the cell exhibited enhanced unfolded protein response. O-Mannosylation may therefore function as a fail-safe mechanism for the ERAD by solubilizing the aberrant proteins that overflowed from the ERAD pathway and reducing the load for ER chaperones.

The endoplasmic reticulum (ER) 1 is the entrance for the protein secretory pathway in eukaryotic cells. Nascent secre-tory and membrane proteins translocating into the ER undergo folding, oligomerization, and maturation, including glycosylation and disulfide bridge formation, with the aid of molecular chaperones and folding enzymes. Quality control mechanisms in the ER monitor these processes to ensure that defective proteins that failed to acquire correct functional structures are not deployed throughout the cells (1). Aberrant proteins are continuously exposed to chaperones and folding enzymes to achieve their native conformations. If this protein repair is not successful, aberrant proteins are cleared from the ER by a mechanism called ER-associated degradation (ERAD) (2,3).
Nonnative proteins have such features as the exposure of hydrophobic regions, exposed free cysteines, and the tendency to aggregate. These features are readily recognized by the ER chaperones and folding enzymes including molecular chaperones of the Hsp70 family, calnexin/calreticulin, and thiol oxidoreducatases of the protein-disulfide isomerase (PDI) family (4). In the ER of yeast Saccharomyces cerevisiae, BiP, an ER member of the Hsp70 family, facilitates folding and assembly of nonnative proteins by preventing their aggregate formation, in cooperation with its partner proteins Jem1p and Scj1p (5,6). Interactions of nonnative proteins with these ER chaperones and folding enzymes during the "chaperone-mediated folding process" may well contribute to the block of their deployment out of the ER (1).
Both correctly folded proteins and nonnative proteins with N-linked glycans receive trimming of glucose by glucosidases in the ER lumen. If glycoproteins cannot acquire native conformations within an appropriate time window, they are targeted to the ERAD pathway for disposal (7,8). Entry for the ERAD pathway requires trimming of a single mannose by the ER ␣1,2-mannosidase I and subsequent recognition of the Man8 moiety by a putative lectin Mnlp1/Htm1p, a yeast counterpart of mammalian EDEM (9 -11). Trimming of terminal mannose by mannosidase I is a slow process, so that it regulates the protein flux for the ERAD pathway. PDI was shown to be responsible for directing nonnative proteins to the ERAD pathway in yeast (12). Nonnative proteins that enter the ERAD pathway are not degraded in the ER lumen but instead translocated back to the cytosol for degradation (13). BiP, a yeast ER member of the Hsp70 family, in cooperation with its partner proteins Jem1p and Scj1p facilitates the ERAD process by preventing aggregation of misfolded proteins prior to the retrotranslocation (5). The retrotranslocation of proteins across the ER membrane to the cytosol is mediated by the Sec61 translocon channel, the same translocon used for import (13), and/or Derlin-1 (Der1p in yeast) (14, 15) with the aid of such cytosolic components as the AAA ATPase, Cdc48, and its part-ner proteins (16). Most of the ERAD substrates are ubiquitinated in the cytosol and are finally degraded by the 26 S proteasome (13). In parallel with the ERAD process, accumulation of nonnative proteins in the ER leads to activation of the unfolded protein response (UPR), which induces transcription of genes encoding a set of proteins that can process aberrant proteins, and translational attenuation to reduce the amount of proteins imported newly into the ER (17).
In addition to N-linked glycochains, O-linked carbohydrate chains are attached to proteins in the ER (18). In yeast, Olinked glycosylation starts by the addition of the initial mannose from dolichol-phosphate mannose to Ser/Thr by one of the seven O-mannosyl transferases (Pmt1-7p), although no consensus amino acid sequence was identified as an O-mannosylation site. Roles of O-mannosylation in the protein secretion (18,19), cell wall integrity (20), budding process (21), and stabilization of cell surface proteins (22) have been suggested. Recently, several aberrant proteins were found to receive Omannosylation in the yeast ER; when N-glycosylation site mutants of prepro-␣-factor (pp␣F) were translocated into isolated yeast microsomes, subsequent incubation with cytosol led to O-mannosylation of imported and cleaved pro-␣-factor (p␣F) (23). Mutated bovine pancreatic trypsin inhibitor received Omannosylation if artificially retained in the ER (24). KHN (Kar2-hemagglutinin neuraminidase) also received O-mannosylation during its degradation by the ERAD pathway (25). However, the biological significance of O-mannosylation of aberrant proteins remains unclear.
Although bulky hydrophilic N-linked glycans could increase solubilities of glycoproteins and suppress their irreversible aggregation (26 -28), few studies have been carried out to test whether O-linked glycans have similar properties. In the present study, we analyzed the effect of O-mannosylation on distinct model aberrant proteins: a series of N-glycosylation site mutants of prepro-␣-factor and a pro-region-deleted derivative of Rhizopus niveus aspartic proteinase I (⌬pro). We found that O-mannosylated p␣F mutants and ⌬pro became soluble and partly secreted out of the cell through the normal secretory pathway. O-Mannosylation rendered p␣F mutants less dependent on the ER chaperone, BiP, for being soluble. The impairment of O-mannosylation, especially with that of the ERAD, led to the enhanced UPR. Thus, we propose a new model in which O-mannosylation reduces the load for ER chaperones to contribute to the maintenance of ER homeostasis.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Culturing Conditions-Yeast strains used in this study are listed in Table I. For construction of the pmt2::LEU2 allele, the PMT2 gene was cloned by PCR using primers 5Ј-GCG GAA TTC GTT CTT GAT GTT CAC AGC CAT-3Ј and 5Ј-GCG AAG CTT TCG TGT AAG TCT TGA GGC GTT-3Ј, and the amplified DNA fragment was introduced into the EcoRI and HindIII sites of pUC118. pJJ283 (29) was cut with XbaI and BamHI, and a resulting 2-kb fragment containing the LEU2 gene was introduced into the XbaI and BglII sites of the PMT2 gene. The pmt2::kanMX4 allele was generated by PCR using the kanMX4 gene (30) as a template and using primers 5Ј-ATG TCC TCG  TCT TCG TCT ACC GGG TAC AGC AAA AAC AAT GCC GCC CAC  ATT AAG CAA GAG TTA AGG CGC GCC AGA TCT GT-3Ј and 5Ј-TCA  TGC TTC TTG CTT GTC GGC AAT GTC CCA AGT GGA AAA CCA  GTT TAA GTA GCG GAA GTT ATC ATC GAT GAA TTC GAG CTC-3Ј. The der1::LEU2 allele and cue1::LEU2 allele were constructed as described previously (31,32). The erg6::LEU2 and doa4::LEU2 alleles were constructed as described (5). Disruption of the genes was confirmed by PCR. Multiple mutants were generated by the second round of transformation.
In vitro translation and translocation into isolated yeast microsomes of ⌬pro were performed as above, except that translation and translocation were coupled. Briefly, microsomes were added to the translation reaction mixture 10 min after initiation of the translation at 20°C and further incubated for 60 min at 20°C. ⌬pro-loaded microsomes were collected by centrifugation and subjected to ERAD assays. Proteins bound to ConA-Sepharose were eluted with 1 M ␣-methylmannoside in 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA for 1 h at 30°C. Proteins were analyzed by SDS-PAGE and radioimaging with a Storm 860 image analyzer (Amersham Biosciences).
Sucrose Density Gradient Centrifugation-Sucrose density gradient centrifugation for proteins in microsomes and yeast cell lysate was performed as in Ref. 5.
Halo Assay-A halo assay was carried out according to Nishikawa and Nakano (36). Briefly, ϳ5 ϫ 10 5 tester cells (BC180), which are supersensitive to ␣-factor, were spread onto SCD plates that contained appropriate supplements and were buffered at pH 3.5. Cells to be tested for ␣-factor production were spotted on the plates with sterile toothpicks. The secreted ␣-factor was quantified according to the following equation: halo size (mm) ϭ 1.8 ln(amount of ␣-factor (ng)) ϩ 5.5. 2 Pulse-Chase Experiments-Metabolic labeling of yeast cells with Tran 35 S-label (ICN) and preparation of cell extracts were performed essentially as described (37). ⌬pro was expressed from the chromosome at the URA3 locus under the GAL1 promoter. Before labeling, ⌬pro was induced in the minimal medium containing 2% galactose. Immunoprecipitation of ⌬pro with anti-RNAP-I antibodies and the second round of immunoprecipitation with anti-␣1,3-mannose antibodies were performed as described previously (9,38). The immunoprecipitated proteins were analyzed by SDS-PAGE with or without 6 M urea and radioimaging.
Assay of Cellular ␤-Galactosidase Activity-UPR induction was measured by using a single copy (CEN) plasmid that harbored the CYC1 promoter containing the UPR element fused to the lacZ gene (pSCZ-Y, a gift from K. Mori). ␤-Galactosidase activity of cells was determined as described (39).

N-Glycosylation Mutants of Pro-␣-factor Receive O-Mannosylation-Prepro-␣-factor
is the precursor form of a yeast mating pheromone and consists of an ER-targeting signal sequence that is cleaved off in the ER, a pro-region, and four tandem repeats of ␣-factor peptides, all of which are cleaved in the Golgi complex. Since the pro-region has three potential Nglycosylation sites (Asn-23, Asn-57, and Asn-67) (Fig. 1A), we designate it as p(NNN)p␣F. p(QQQ)p␣F, a mutant form of p(NNN)p␣F that has three glutamine residues substituted for the N-glycosyl acceptor sites (Fig. 1A) and therefore receives no N-glycosylation in the ER, has been used as a model aberrant protein for in vitro ERAD assays with yeast microsomes (34,43,44). Besides, it was reported that a fraction of (QQQ)p␣F imported into yeast microsomes in vitro receives O-mannosylation in the ER lumen instead of being retrotraslocated to the cytosol for degradation under the ERAD conditions, although its biological relevance remained unclear (23).
When radiolabeled wild-type p(NNN)p␣F and mutant p(QQQ)p␣F, synthesized in yeast lysates (pp in Fig. 1B, NNN and QQQ) were incubated with isolated yeast microsomes, they were efficiently translocated into microsomes. p(NNN)p␣F received signal sequence cleavage and N-glycosylation to form 3G(NNN)p␣F (3Gp in Fig. 1B, NNN) in the ER lumen, whereas p(QQQ)p␣F received signal sequence cleavage, but no N-glycosylation and formed ⌬G(QQQ)p␣F (⌬Gp in Fig. 1B, QQQ), since it lacked all of the N-glycosyl acceptor sites. The microsomes were then washed and incubated with the yeast cytosol fraction and "ATP mix" (the ERAD conditions; see "Experimental Procedures"). The amount of 3G(NNN)p␣F remained the same even after 60 min incubation with the cytosol. On the other hand, the amount of ⌬G(QQQ)p␣F decreased, but a smeared band (asterisk in Fig. 1B, QQQ) concomitantly appeared above the band for ⌬G(QQQ)p␣F and shifted upward with increasing time of incubation, which appeared to reflect O-mannosylation (23). Indeed, the ⌬G(QQQ)p␣F species responsible for the smeared band was precipitated with ConA-Sepharose beads (m⌬Gp in Fig. 1C, lane 2) and even after treatment with endoglycosidase H, which removes N-linked glycans (m⌬Gp in Fig. 1C, lane 4). When nonlabeled p(QQQ)p␣F, instead of radiolabeled p(QQQ)p␣F, was translocated into the microsomes and subsequently incubated with the cytosol and ATP mix in the presence of [ 14 C]GDPMan, a [ 14 C]mannose-labeled form of ⌬G(QQQ)p␣F was detected by radioimaging after immunoprecipitation with anti-pp␣F antibodies (m⌬Gp in Fig. 1D). This confirms that mannose was transferred to ⌬G(QQQ)p␣F from GDPMan as a donor.
Since ⌬G(QQQ)p␣F received O-mannosylation whereas 3G(NNN)p␣F did not, we tried to assess the relationship between the number of N-glycosylated moieties and the extent of O-mannosylation on p␣F derivatives. We thus made a series of p(NNN)p␣F mutants in which one or two asparagine residue(s) in the N-glycosyl acceptor sites was converted to glutamine residue(s) (Fig. 1A). p(NNN)p␣F mutants with an Asn 3 Gln substitution at position 67, 57, or 23 are designated as p(NNQ)p␣F, p(NQN)p␣F, and p(QNN)p␣F, respectively, and those with double Asn 3 Gln substitutions at positions 57 and 67, 23 and 67, or 23 and 57 are designated as p(NQQ)p␣F, p(QNQ)p␣F, and p(QQN)p␣F, respectively. The relative mobilities of signal cleaved and N-glycosylated forms of single glycosylation site mutants (2G(NNQ)p␣F, 2G(NQN)p␣F, and 2G(QNN)p␣F) and double glycosylation site mutants (1G(NQQ)p␣F, 1G(QNQ)p␣F, and 1G(QQN)p␣F) to those of 3G(NNN)p␣F and ⌬G(QQQ)p␣F are consistent with one another (pp in Fig. 1B). The bands for 1G(QNQ)p␣F and 1G(QQN)p␣F migrated faster than that for 1G(NQQ)p␣F (45) and were overlapped by those for p(QNQ)p␣F and p(QQN)p␣F (1Gp and pp in Fig. 1B, 0 min of chase of QNQ and QQN). During incubation with the cytosol and ATP mix, 1G(NQQ)p␣F, 1G(QNQ)p␣F, and 1G(QQN)p␣F were partly converted to higher molecular weight forms (asterisk in Fig. 1B, NQQ, QNQ, and QQN), whereas 2G(NNQ)p␣F, 2G(NQN)p␣F, and 2G(QNN)p␣F were not (Fig. 1B). These higher molecular weight forms arose from O-mannosylation, since they were not observed with microsomes prepared from the pmt2⌬ strain, which lacks the PMT2 gene, one of the seven PMT genes for mannosyl transferases in yeast (Fig. 1E). O-Mannosylation of double glycosylation site mutants was also confirmed by precipitation with ConA-Sepharose beads after treatment with endoglycosidase H (Fig. 1C, lanes 1 and 3). Therefore, although 3G(NNN)p␣F and the single glycosylation site mutants (2G(NNQ)p␣F, 2G(NQN)p␣F, and 2G(QNN)p␣F) did not receive O-mannosylation, double glycosylation site mutants (1G(NQQ)p␣F, 1G(QNQ)p␣F, and 1G(QQN)p␣F) and ⌬G(QQQ)p␣F were significantly O-mannosylated. The Omannosylated forms were resistant to externally added protease but were degraded by protease upon disruption of the membranes with detergent (data not shown), suggesting that they remained in the ER lumen.
O-Mannosylation Renders Aggregation-prone p␣F Mutants Soluble-In contrast to 3G(NNN)p␣F, which is soluble in the ER lumen, ⌬G(QQQ)p␣F is prone to aggregation and remains soluble only in the presence of the ER luminal Hsp70, BiP (5). We therefore reasoned that removal of single or double Nglycosylation sites may affect solubilities of the p␣F mutants differently, which would in turn lead to different levels of O-mannosylation of the p␣F mutants.
To test this hypothesis, we first asked if O-mannosylation of ⌬G(QQQ)p␣F required functional BiP, since aggregation-prone proteins tend to interact with BiP to remain soluble (5) (Fig. 2). Radiolabeled p(QQQ)p␣F was translocated into microsomes prepared from wild-type cells or kar2-159 yeast cells at 23°C ("translocation"). The ⌬G(QQQ)p␣F-loaded microsomes were FIG. 1. N-glycosylation mutants of p␣F receive O-mannosylation. A, pp␣F is schematically outlined. The triangles indicate asparagine residues that receive N-glycosylation. N-Glycosylation mutants of p␣F used in this study are listed. Wild-type pp␣F was designated as p(NNN)p␣F. For designation of the N-glycosylation mutants, see "N-Glycosylation Mutants of Pro-␣-factor Receive O-Mannosylation" (in "Results"). pre, cleavable signal sequence; pro, pro-sequence. B, radiolabeled pp␣F and its N-glycosylation mutants were translocated into wild-type (SEY6210) microsomes at 30°C for 60 min. After reisolation, the membranes were subjected to incubation with ATP mix and the yeast cytosol fraction at 30°C for the indicated times (chase). Proteins were analyzed by SDS-PAGE. (NNN)p␣F, a signal cleaved form of p(NNN)p␣F, receives three N-linked glycans to form 3G(NNN)p␣F (0 min of NNN). The bands for 1G(QNQ)p␣F and 1G(QQN)p␣F migrate faster than that for 1G(NQQ)p␣F (45) and are overlapped by those for p(QNQ)p␣F and p(QQN)p␣F (0 min of QNQ and QQN). The asterisks indicate O-mannosylated forms. ⌬Gp, 1Gp, 2Gp, and 3Gp, p␣F mutants with 0, 1, 2, or 3 N-linked glycan(s), respectively. m⌬Gp and m1Gp indicate O-mannosylated p␣F mutants with 0 or 1 N-linked glycan, respectively. pp, pp␣F; p, p␣F. C, radiolabeled p(NQQ)p␣F and p(QQQ)p␣F were translocated into wild-type (SEY6210) microsomes at 30°C for 60 min. After reisolation, the membranes were incubated with ATP mix and the yeast cytosol fraction at 30°C for 60 min and subsequently solubilized with 1% SDS. The proteins were precipitated with ConA-Sepharose (lanes 1 and 2) or first treated with endoglycosidase H and then precipitated with ConA-Sepharose (lanes 3 and 4).  ). B, radiolabeled p(QQQ)p␣F was translocated into wild-type (RSY607) microsomes at 23°C for 60 min. The reaction mixtures were split in half, and the ⌬G(QQQ)p␣F-loaded membranes were isolated. One sample was subjected to incubation with ATP mix at 37°C for 30 min. The remaining sample was incubated with ATP mix and the yeast cytosol fraction at 23°C for 50 min to generate the O-mannosylated form of ⌬G(QQQ)p␣F (m⌬G(QQQ)p␣F), and subsequently, the membranes were isolated and incubated with ATP mix at 37°C for 30 min. Samples were solubilized with 1% Triton X-100 and subjected to sucrose density gradient centrifugation. Fractions were collected, and proteins were analyzed by SDS-PAGE and radioimaging. Relative amounts of ⌬G(QQQ)p␣F and m⌬G(QQQ)p␣F were quantified. C and D, radiolabeled p(NQQ)p␣F, p(QNQ)p␣F, and p(QQQ)p␣F were translocated into kar2-159 (RSY579) microsomes at 23°C for 60 min. The reaction mixtures were subsequently divided into halves, and the 1G(NQQ)p␣F-, 1G(QNQ)p␣F-, and ⌬G(QQQ)p␣F-loaded membranes were isolated. One aliquot was subjected to incubation with ATP mix at 37°C for 30 min (C). The other aliquot was incubated with ATP mix and yeast cytosol at 23°C for 50 min to generate O-mannosylated forms of 1G(NQQ)p␣F, 1G(QNQ)p␣F, and ⌬G(QQQ)p␣F (m1G(NQQ)p␣F, m1G(QNQ)p␣F, and m⌬G(QQQ)p␣F). The membranes were subsequently isolated and incubated with ATP mix at 37°C for 30 min (D). Samples were subjected to incubated at 23 or 37°C for 30 min in the presence of ATP mix ("first incubation") but in the absence of the cytosol. During this incubation, BiP in kar2-159 microsomes is fully functional at permissive temperature (23°C) but becomes defective at restrictive temperature (37°C). The microsomes were then incubated with both the cytosol and ATP mix at the same temperature as the first incubation for 50 min ("O-mannosylation"). Whereas ⌬G(QQQ)p␣F was efficiently O-mannosylated (m⌬Gp) in wild-type microsomes in the presence of the cytosol and ATP mix both at the permissive temperature of 23 and 37°C (asterisk in Fig. 2, lanes 2 and 4) and in the kar2-159 microsomes at 23°C (asterisk in Fig. 2, lane 6), it was hardly O-mannosylated in the kar2-159 microsomes at the restrictive temperature of 37°C (Fig. 2, lane 8). This result indicates that functional BiP is required for O-mannosylation of ⌬G(QQQ)p␣F, probably because substrates for O-mannosylation should be soluble in the ER lumen.
Aggregation states of N-glycosylation site mutants of p␣F and its O-mannosylated forms in the microsomes incubated at restrictive temperature (37°C) were analyzed by sucrose density gradient centrifugation after solubilization of the microsomes with Triton X-100 (schematized in Fig. 3A). After translocation of N-glycosylation site mutants of p␣F into wild-type and kar2-159 microsomes at 23°C ("translocation"), incubation of the microsomes at 23°C with ATP mix and cytosol is required to generate O-mannosylated forms ("O-mannosylation"). These microsomes were then incubated at 37°C (second incubation). ⌬G(QQQ)p␣F solubilized from the wild-type microsomes were fractionated with a molecular mass corresponding to 70 kDa, but its mannosylated form m⌬G(QQQ)p␣F was recovered exclusively in the fractions corresponding to Ͻ30 kDa (Fig. 3B). Therefore, m⌬G(QQQ)p␣F and ⌬G(QQQ)p␣F are soluble as monomers and probably as a small oligomer, respectively. In kar2-159 microsomes, p␣F mutants that lack two or three N-glycosylation sites (1G(NQQ)p␣F, 1G(QNQ)p␣F, and ⌬G(QQQ)p␣F) were aggregated and recovered only at the bottom of the tube after incubation at restrictive temperature (37°C; BiP is not functional) (Fig. 3C). On the other hand, their O-mannosylated forms (m1G(NQQ)p␣F, m1G(QNQ)p␣F, and m⌬G(QQQ)p␣F) were recovered in the fractions corresponding to Ͻ30 kDa after incubation at restrictive temperature (37°C) (Fig. 3D). This result suggests that the O-mannosylated forms are soluble as a monomer without the assistance of BiP. The high solubilities of O-mannosylated p␣F mutants may be compared with those of 3G(NNN)p␣F, 2G(NQN)p␣F, and 2G(QNN)p␣F. In the kar2-159 microsomes, 3G(NNN)p␣F, 2G(NQN)p␣F, and 2G(QNN)p␣F were recovered in the fractions for 30 -70 kDa after incubation at restrictive temperature (37°C) (Fig. 3E), suggesting that they also remained soluble, but probably in oligomeric forms.
These results indicate that the p␣F mutants that lack two or three N-glycosylation sites are prone to aggregate if BiP is inactivated. In other words, BiP is required for these p␣F mutants to remain soluble in the ER lumen. However, a significant fraction of these N-glycosylation site mutants of p␣F become O-mannosylated forms in the ER lumen, which are, in contrast to the nonglycosylated forms, highly soluble independent of the BiP function.
Pmt2p Facilitates Secretion of p␣F Mutants-Although the previous in vitro study indicated that O-mannosylated ⌬G(QQQ)p␣F failed to be secreted from the ER (23), the in vitro ERAD assay system with yeast microsomes did not reconstitute downstream membrane trafficking along the secretory pathway. To analyze these downstream events, we followed the fate of O-mannosylated p␣F mutants or their secretion by using an in vivo halo assay.
The halo assay relies on the ability of secreted ␣-factor to block growth of ␣-factor-sensitive yeast mutant cells and a zone of growth inhibition that manifests as a "halo" (46). The amount of the secretion of ␣-factor can be estimated semiquantitatively. Since N-and O-glycosylation sites of p␣F are present in the pro-region, but not in the mature region, changes in glycosylation will not affect the ␣-factor activity of the mature peptide pheromone itself. p(NNN)p␣F, p(NQQ)p␣F, and p(QQQ)p␣F were expressed from a single copy (CEN) plasmid in PMT2 and pmt2⌬ cells and subjected to the halo assay (Fig.  4). In wild-type cells, the amount of ␣-factor secreted by p(NQQ)p␣F and p(QQQ)p␣F were ϳ36 and ϳ8% as compared with p(NNN)p␣F, respectively, indicating that elimination of N-glycosylation sites led to reduction in ␣-factor secretion.  ⌬Gp, 1Gp, 2Gp, and 3Gp, p␣F mutants with 0, 1, 2, or 3 N-linked glycan(s), respectively. m⌬Gp and m1Gp, O-mannosylated p␣F mutants  with 0 or 1 N-linked glycan, respectively. pp, pp␣F; p, p␣F.
Larger effects were observed for p(QQQ)p␣F with all three glycosylation sites eliminated than for p(NQQ)p␣F with two sites eliminated (Fig. 4), a result that is consistent with previous data (45). In pmt2⌬ cells, the amount of ␣-factor secreted by p(NQQ)p␣F was ϳ15% as compared with p(NNN)p␣F, and no halo was observed for p(QQQ)p␣F, indicating that elimination of the N-linked glycosylation sites on p␣F mutant secretion became more prominent in pmt2⌬ cells (Fig. 4), in which Omannosylation of 1G(NQQ)p␣F and ⌬G(QQQ)p␣F was suppressed. This result strongly suggested that the O-mannosylated forms of p␣F mutants, which were highly soluble in the ER lumen, became competent for transport from the ER to the Golgi and were finally secreted out of the cell.
⌬pro Is O-Mannosylated When ERAD Is Blocked in Vivo-Is solubilization of aberrant proteins by O-mannosylation in the ER specific to p␣F mutants? Harty et al. (23) observed that, in addition to authentic substrates for O-mannosylation, many proteins isolated from tunicamycin-treated yeast cells can be O-mannosylated. As a test for the generalization of our model for the role of O-mannosylation, we analyzed another substrate, ⌬pro, a pro-region-deleted derivative of RNAP-I, for possible O-mannosylation. RNAP-I does not have potential Nglycosylation sites and the N-terminal pro-region is essential for its folding and secretion (47). Upon expression in yeast, ⌬pro does not exit from the ER but is instead degraded irrespective of the block of the ER-to-Golgi transport (38). When ⌬pro was expressed in wild-type cells under the control of the GAL1 promoter and analyzed by pulse-chase experiments, the amount of ⌬pro decreased rapidly (Fig. 5A, wild-type) (38). We found that a weak smeared band, which escaped detection in the previous study (38), appeared above the band for ⌬pro and shifted upward on SDS-polyacrylamide gels with increasing time of incubation (Fig. 5A, wild-type). Intensities of the smeared bands were markedly enhanced when analyzed with der1⌬ cells and cue1⌬ cells, in which the gene for Der1p, an integral ER membrane protein involved in the ERAD (31), or Cue1p, which recruits the ubiquitin-conjugating enzyme to the ER membrane for the ERAD (32), was deleted, respectively (Fig. 5A, der1⌬ and cue1⌬). The smeared bands disappeared in pmt2⌬, pmt2⌬der1⌬, and pmt2⌬cue1⌬ cells, in which the PMT2 gene or both the PMT2 and DER1 or CUE1 genes were deleted, respectively (Fig. 5B, pmt2⌬, pmt2⌬der1⌬, and  pmt2⌬cue1⌬); hence, they reflect O-mannosylation of ⌬pro.
The sum of the amounts of ⌬pro and its O-mannosylated form (m⌬pro) decreased with half-lives of 23, 60, and 65 min in wild-type, der1⌬, and cue1⌬ cells, respectively (Fig. 5A, der1⌬  and cue1⌬). The retardation of degradation of the ⌬pro species (⌬pro and m⌬pro) in der1⌬ and cue1⌬ cells suggests that they were at least partly degraded via the ERAD pathway. This is in contrast to the suggestion in the previous report (38). We thus reevaluated the degradation of the total ⌬pro species by compromising the activity of the ubiquitin-proteasome pathway with MG132, an inhibitor of the 26 S proteasome (48), or depletion of Doa4p, a deubiquitination enzyme (49). In both cases, ⌬pro species (⌬pro and m⌬pro) were partially stabilized; ⌬pro and m⌬pro were degraded now with half-lives of 50 and 60 min, respectively (Fig. 5C, ϩMG132 and doa4⌬). O-Mannosylation of ⌬pro was hardly observed, probably due to the weak inhibition of degradation. It should be noted that, in pmt2⌬ cells, ⌬pro was degraded efficiently with a half-life of 25 min (Fig. 5B, pmt2⌬), suggesting that O-mannosylation takes place for a fraction of ⌬pro that already escaped degradation, as in the case of the p␣F mutants.
⌬pro Is O-Mannosylated in Vitro-We next tested the possibility that ⌬pro was also O-mannosylated in vitro. Radiolabeled p⌬pro synthesized in vitro could be co-translationally but not posttranlationally translcoated into yeast microsomes (Fig. 6). p⌬pro received signal sequence cleavage and formed ⌬pro, which is resistant against digestion with externally added trypsin (Fig. 6, upper panel, trypsin (ϩ), Triton X-100 (Ϫ)) but, after solubilization of the membrane with Triton X-100, became susceptible to trypsin digestion (Fig. 6, upper panel, trypsin (ϩ), Triton X-100 (ϩ)). Upon incubation of ⌬pro-loaded microsomes with ATP mix and yeast cytosol (the ERAD conditions; see Refs. 5 and 34), the amount of ⌬pro decreased with increasing time of incubation, but a smeared band (Fig. 6, upper panel,  m⌬pro) concomitantly appeared above the band for ⌬pro and increased its intensity. m⌬pro responsible for the smeared band was precipitated with ConA-Sepharose beads (Fig. 6,  upper panel, ConA). When we used microsomes prepared from the pmt2⌬ strain, that lacks the PMT2 gene, one of the seven PMT genes for mannosyl transferases in yeast, we did not observe the smeared band above the band for translocated ⌬pro (Fig. 6, lower panel). This confirms that ⌬pro is O-mannosylated by Pmt2p in wild-type microsomes in vitro. We did not observe degradation of ⌬pro during incubation of the microsomes under the present ERAD conditions, and this absence of degradation may have facilitated O-mannosylation of ⌬pro in microsomes.
O-Mannosylation Partly Restores the Normal Secretory Pathway Transport of ⌬pro-Since O-mannosylation increased the solubility of p␣F mutants, we analyzed the solubility of m⌬pro formed in vivo. Proteins extracted from wild-type, der1⌬, and cue1⌬ cells that express ⌬pro were analyzed by sucrose density gradient centrifugation (5) followed by immunoblotting with anti-RNAP-I antibodies. After centrifugation, ⌬pro was mainly recovered in the pellet (Fig. 7A) (42), whereas m⌬pro in der1⌬ and cue1⌬ cells were recovered in the fractions corresponding to 30 -60 kDa (Fig. 7, B and C). This result indicates that O-mannosylation renders ⌬pro soluble.
We next followed the fate of m⌬pro in various mutant strains that have defects in Der1p or Cue1p. When cell extracts were analyzed for ⌬pro in der1⌬ and cue1⌬ cells by immunoblotting with anti-RNAP-I antibodies, smeared bands arising from Omannosylation of ⌬pro were observed (Fig. 8A, lanes 2 and 3). In pmt2⌬der1⌬ and pmt2⌬cue1⌬ cells, only ⌬pro without Omannosylation was detected (Fig. 8A, lanes 5 and 6). Since Pmt2p is not required for degradation of ⌬pro (Fig. 5B), ⌬pro did not significantly accumulate in pmt2⌬ cells (Fig. 8A, lane 4) but did accumulate in pmt2⌬der1⌬ and pmt2⌬cue1⌬ cells (Fig.   8A, lanes 5 and 6). These results are consistent with our interpretation that ⌬pro is at least partly degraded via the ubiquitin-proteasome pathway and is O-mannosylated when this pathway is blocked.
When extracellular medium was analyzed for ⌬pro with anti-RNAP-I antibodies, m⌬pro, but not unmodified ⌬pro, was detected in the medium for der1⌬ and cue1⌬ cells (Fig. 8A, lanes  8 and 9). m⌬pro in the medium bound to ConA-Sepharose beads, confirming that m⌬pro indeed contained O-linked mannose (Fig. 8B, lanes 5 and 6). As a control, when ⌬pro was expressed in pmt2⌬der1⌬ and pmt2⌬cue1⌬ cells, m⌬pro was not detected in the cell extract or in the medium (Fig. 8A, lanes  11 and 12). An ER chaperone BiP and a cytosolic chaperone Ssb1p were not detected in the medium for der1⌬ and cue1⌬ cells (Fig. 8C), ruling out the trivial possibility that m⌬pro detected in the medium was due to cell lysis.
In parallel, we confirmed the ER-to-Golgi transport of m⌬pro by analyzing its Golgi-specific glycosylation by using anti-␣1,3- mannose antibodies. der1⌬ and cue1⌬ cells expressing ⌬pro under the control of the GAL1 promoter were pulse-labeled, chased, and subjected to immunoprecipitation with anti-RNAP-I antibodies to precipitate ⌬pro and m⌬pro and subsequently with the anti-␣1,3-mannose antibodies to precipitate proteins with the Golgi-specific addition of ␣1,3-mannose. m⌬pro was immunoprecipitated with anti-␣1,3-mannose antibodies (Fig. 8D, lanes 4 and 8), suggesting that m⌬pro was indeed modified by the late Golgi enzymes. Therefore, m⌬pro was secreted to the extracellular medium via the normal secretory pathway through the Golgi complex.
Mutations Affecting ERAD and O-Mannosylation Cause Synthetic Phenotypes-Partial recovery of the normal transport of aberrant proteins by O-mannosylation reflects their disengagement from interactions with BiP or other ER chaperones, which usually prevents further secretion through the organelles of the secretory pathway. However, although they are released from ER chaperones, a large fraction of the O-mannosylated forms appear to remain in the ER (data not shown). Nevertheless, since the O-mannosylated proteins remain soluble irrespective of functional BiP, O-mannosylation may reduce the load for ER chaperones significantly. We tested this possibility by monitoring the effects of O-mannosylation on the UPR. Under normal growth conditions, defects in the ERAD lead to an increase in the amounts of nonnative proteins in the ER, which results in constitutive and moderate activation of the UPR (50,51). If O-mannosylation contributes to the ER quality control by solubilization of aberrant proteins in the ER when ERAD is blocked, deletion of genes involved both in O-mannosylation and in ERAD should further activate the UPR. We thus monitored the UPR using a ␤-galactosidase reporter under the control of the promoter containing the UPR-response element (Fig. 9A). Whereas deletion of the DER1 or PMT2 gene alone led to moderate induction of the ␤-galactosidase activity, pmt2⌬der1⌬ and pmt2⌬cue1⌬ mutants displayed significant increase in the ␤-galactosidase activity. This result suggested that simultaneous deletion of the DER1/CUE1 and PMT2 genes caused synergistic activation of the UPR.
Umebayashi et al. (42) reported previously that ⌬pro, which was overexpressed from a multicopy plasmid under the GAL1 promoter, formed aggregates containing BiP but not PDI. In addition, immunoelectron microscopic analyses revealed that the aggregates of ⌬pro were visible as electron-dense regions in the ER and nuclear envelope. In this study, although ⌬pro is expressed from the chromosome under the GAL1 promoter, we reasoned that ⌬pro, which was neither degraded nor O-mannosylated in pmt2⌬der1⌬ and pmt2⌬cue1⌬ cells, became insoluble and accumulated in the ER lumen. When wild-type cells expressing ⌬pro were stained with anti-RNAP-I, anti-BiP, and anti-PDI antibodies, indirect immunofluorescence microscopy showed perinuclear staining with several extensions in the cytoplasm (Fig. 9B, a, e, and i). This staining is typical for yeast ER proteins and suggests that ⌬pro is distributed uniformly in the ER lumen at this expression level of ⌬pro. However, 35% of der1⌬ cells (b), 35% of pmt2⌬ cells (c), and 76% of pmt2⌬der1⌬ cells (d) showed a dotlike distribution along the perinuclear structure, when stained with anti-RNAP-I antibodies (arrowhead in d). When stained with anti-BiP antibodies, 25% of der1⌬ cells (f), 33% of pmt2⌬ cells (g), and 90% of pmt2⌬der1⌬ cells (h) showed similar distribution patterns (arrowhead in h). Since anti-PDI antibodies exhibited normal staining of ER proteins (j-l), the dotlike subcellular distribution is specific for ⌬pro and the ER chaperone BiP, but not for PDI. These structures probably represent aggregates formed by ⌬pro with BiP, which was neither degraded nor O-mannosylated but accumulated in the ER lumen, as observed in the previous study (42). DISCUSSION In the present study, we analyzed the consequence of the ER O-mannosylation of model nonnative proteins in the ERAD. We found, by using N-glycosylation site mutants of p␣F as aberrant proteins, that removal of N-glycosylation sites decreased the solubility of p␣F, but in turn, subsequent O-mannosylation regained the solubility. Aberrant substrate proteins have to be soluble, hence require BiP or ER chaperones, for receiving O-mannosylation. However, once they have become O-mannosylated, they do not require BiP to remain soluble. We also demonstrated that ⌬pro, an aberrant secretory protein derivative that lacks its pro-sequence, was O-mannosylated in the ER both in vivo and in vitro and became soluble.
Since "early aggregate" of misfolded proteins are often toxic rather than simply nonfunctional (52), the cell possesses a protein quality control system to remove aggregation-prone proteins (1,8,53). Avoidance of early aggregate is achieved by association with molecular chaperones that shield the exposed hydrophobic surfaces from one another or by degradation of the proteins into small peptides. Attachment of hydrophilic glycans to incompletely folded or misfolded proteins would also contribute to the avoidance of aggregate formation by increasing their solubilities, as shown in the present study. In the case of p␣F mutants, serine and threonine residues, potential O-mannosylation acceptor sites, are clustered in its pro-region, where N-glycosylation sites are also localized in wild-type p␣F (54), so that O-mannosylation may well compensate the lowered solubility due to failure of N-glycosylation. Reduced solubility of ⌬pro due to deletion of the pro-region of RNAP-I could be also cured by simply attaching highly hydrophilic mannosyl chains. In this connection, it is interesting to note that O-mannosylation by Pmt proteins has a preference for unfolded substrates but not for fully folded proteins (18,23). This raises an attractive hypothesis that Pmt proteins may be, like glucosyl transferase in the ER lumen (55), directly involved in recognition of unfolded and therefore aggregation-prone substrate proteins. On the other hand, whereas N-linked glycosylation (56) and presumably O-mannosylation of authentic substrate proteins are compatible with correct folding process of substrate proteins, O-mannosylation of aberrant proteins may interfere with their achievement of native conformations and/or assembly in the ER. Therefore, the role of O-mannosylation of aberrant proteins would be to increase solubility of the proteins, irrespective of their abilities to fold into compact native conformations.
Although the O-mannosylated form of N-glycosylation site mutants of p␣F and ⌬pro was partly secreted to the extracellular medium, the secreted fraction was not so large. It is thus unclear whether clearance of aberrant proteins from the ER along the secretory pathway is the major biological consequence of O-mannosylation. Rather, a more likely scenario could be that O-mannosylation functions as a mechanism to avoid accumulation and subsequent aggregate formation of aggregation-prone proteins in the ER, which could be otherwise detrimental to cell survival. The aberrant proteins that failed to proceed to degradation and were retained in the ER are instead O-mannosylated and remain soluble without the assistance of ER molecular chaperones, which are up-regulated by UPR in response to accumulation of nonnative proteins. Although molecular chaperones are major components that prevent aggregation of aberrant proteins and the ERAD is the major pathway to eliminate them, O-mannosylation of aggregation-prone proteins could contribute to reduction of the load of the ER stress for the ER protein quality control systems. Indeed, deletions of the ERAD-related genes, DER1 and CUE1, and the PMT2 gene led to aggregate formation of ⌬pro involv- FIG. 9. Impairment of O-mannosylation and ERAD causes synthetic phenotypes. A, wild-type (R27-7C-1C), der1⌬ (KNY48), cue1⌬ (KNY49), pmt2⌬ (KNY50), pmt2⌬der1⌬ (KNY51), and pmt2⌬cue1⌬ (KNY52) cells were transformed with pSCZ-Y (UPR-reporter construct) and grown in SD medium to the early logarithmic phase and ␤-galactosidase activity was measured. The values are means Ϯ S.D. B, subcellular localization of BiP and ⌬pro by indirect immunofluorescence. Cells of wild-type (KUY90; a, e, and i), der1⌬ (KNY13; b, f, and j), pmt2⌬ (KNY40; c, g, and k), and pmt2⌬der1⌬ (KNY41; d, h, and l) expressing ⌬pro from the chromosome under the GAL1 promoter were fixed and subjected to immunofluorescence microscopy using anti-RNAP-I (a-d), anti-BiP (e-h), and anti-PDI (i-l) antibodies. The arrowheads show a dotlike distribution along the perinuclear structure. These structures probably represent aggregates formed by ⌬pro with BiP, which was neither degraded nor O-mannosylated but accumulated in the ER lumen. Bar, 2 m.
ing BiP when ⌬pro was overexpressed and showed synergic activation of the UPR (Fig. 9B). In addition, PMT1-4 genes are transcriptionally up-regulated by the UPR (51). These observations suggest that O-mannosylation as well as the ERAD eliminates nonnative misfolded proteins whose accumulation increases the load for ER chaperones and triggers the UPR. In other words, O-mannosylation may function as a fail-safe mechanism for the ERAD by solubilizing the aberrant proteins that overflowed from the ERAD pathway, and reducing the load for ER chaperones.
Is our observation of solubilization of aberrant proteins by O-mannosylation biologically relevant to normal cell functions? Although we analyzed O-mannosylation of only two proteins, p␣F and ⌬pro, under conditions where the ERAD was more or less compromised, many endogenous proteins were reported to be O-mannosylated when N-glycosylation was inhibited by tunicamycin treatment (23). Besides, we found that impairment of O-mannosylation by deletion of the PMT2 gene alone led to activation of the UPR even under normal growth conditions (Fig. 9A). This result suggests that stochastically generated misfolded proteins in the ER of wild-type cells may be handled with partly by O-mannosylation. Alternatively, the inactivation of Pmt2p could prevent O-mannosylation of authentic substrate proteins for Pmt2p, which may lead to an increase in the load for the ER quality control system. Nevertheless, the latter interpretation raises the possibility that the role of O-mannosylation of authentic substrate proteins is also to enhance their activity and efficient secretion by preventing their aggregation. This hypothesis can be tested in future studies.