Membrane association is a determinant for substrate recognition by PMT4 protein O-mannosyltransferases

Abstract

Protein O-mannosylation represents an evolutionarily conserved, essential posttranslational modification with immense impact on a variety of cellular processes. In humans, O-mannosylation defects result in Walker–Warburg syndrome, a severe recessive congenital muscular dystrophy associated with defects in neuronal migration that produce complex brain and eye abnormalities. In mouse and yeasts, loss of O-mannosylation causes lethality. Protein O-mannosyltransferases (PMTs) initiate the assembly of O-mannosyl glycans. The evolutionarily conserved PMT family is classified into PMT1, PMT2, and PMT4 subfamilies, which mannosylate distinct target proteins. In contrast to other types of glycosylation, signal sequences for O-mannosylation have not been identified to date. In the present study, we identified signals that determine PMT4-dependent O-mannosylation. Using specific model proteins, we demonstrate that in yeast Pmt4p mediates O-mannosylation of Ser/Thr-rich membrane-attached proteins. The nature of the membrane-anchoring sequence is nonrelevant, as long as it is flanked by a Ser/Thr-rich domain facing the endoplasmic reticulum lumen. Our work shows that, in contrast to several other types of glycosylation, PMT4 O-mannosylation signals are not just linear protein's primary structure sequences but rather are highly complex. Based on these findings, we performed in silico analyses of the Saccharomyces cerevisiae proteome and identified previously undescribed Pmt4p substrates. This tool for proteome-wide identification of O-mannosylated proteins is of general interest because several of these proteins are major players of a wide variety of cellular processes.

Glycosylation is an essential and abundant protein modification (1). More than half of all proteins in biological systems are glycosylated, and the glycan chains influence a large number of biological processes (2).

Pro- and eukaryotes modify proteins with a variety of carbohydrate residues. Regarding the two major types of protein glycosylation, N- and O-glycosylation, carbohydrate moieties are attached either to the amide group of asparagine (Asn) residues of the sequon AsnXSer/Thr or to hydroxy amino acids, mostly serine (Ser) and threonine (Thr) residues, respectively. In the case of O-glycosylation, a variety of monosaccharides such as N-acetylgalactosamine, N-acetylglucosamine, fucose, glucose, xylose, or mannose are found in O-glycosidic bonds in different organisms (2, 3).

Protein O-mannosylation represents an evolutionarily conserved modification among eukaryotes and mycobacteria (1, 4). In yeasts and filamentous fungi, O-mannosylation serves a variety of different functions. It is required for stability, sorting, and localization of proteins, thereby affecting protein function and being indispensable for cell wall integrity, cell polarity, and morphogenesis (1). In Drosophila melanogaster, reduced O-mannosylation results in altered muscle structures and alignment of adult cuticle (5). In mouse, lack of O-mannosylation results in embryonic lethality (6) and in humans in congenital muscular dystrophies with neuronal migration defects, such as Walker–Warburg syndrome and muscle–eye–brain disease (7).

O-mannosyl glycans are short linear oligosaccharides linked via an α-glycosidically-bound mannose to Ser and Thr residues (7). Biosynthesis is initiated at the endoplasmic reticulum (ER) by the transfer of mannose from dolichyl phosphate-activated mannose to Ser or Thr residues of proteins that are entering the secretory pathway (1). Further chain elongation takes place in the Golgi apparatus using nucleotide activated sugars as donors. The initial mannosyltransfer reaction is catalyzed by an essential family of dolichyl phosphate-d-mannose:protein O-mannosyltransferases (PMTs) that is evolutionarily conserved from yeast to humans (5, 8–11). PMTs have been identified and extensively characterized in yeast. In Saccharomyces cerevisiae, seven PMT family members (Pmt1p to -7p) are present (8, 9) that are integral ER membrane proteins with seven transmembrane-spanning domains (12). Phylogenetic analyses indicate that the PMT family is subdivided into the PMT1, PMT2, and PMT4 subfamilies, whose members include transferases closely related to S. cerevisiae Pmt1p, Pmt2p, and Pmt4p, respectively (1, 13).

In yeast members of the PMT family show a high degree of conservation. Despite the high sequence homology, several features suggest that PMT1/PMT2 and PMT4 members form distinct functional subclusters. First, all PMT family members share three highly conserved sequence motifs that, nonetheless, show significant variations between PMT1/PMT2 and PMT4 subfamily members (13). Second, members of the PMT1 subfamily physically interact with members of the PMT2 subfamily, whereas the unique representative of the PMT4 subfamily forms homomeric complexes (14). Third, Pmt1p/Pmt2p and Pmt4p complexes mannosylate different acceptor proteins in vivo (15, 16). In S. cerevisiae, a subset of O-mannosylated proteins such as Kre9p, Cts1p, Bar1p, Pir2p, and Aga2p is exclusively mannosylated by Pmt1p/Pmt2p complexes (15). In contrast, Kex2p, Gas1p, Axl2p, and Fus1p are O-mannosylated by Pmt4p (15, 17). A third group of proteins including the WSC-family members, Mid2p and Ccw5p/Pir4p, is glycosylated by both complexes; however, Pmt1p/Pmt2p and Pmt4p mannosylate different domains of the protein (16, 18). In contrast to other types of glycosylation, signals causing O-mannosylation of Ser and Thr residues by PMT family members and determinants of the different substrate specificities among the PMT complexes are unknown.

Here, we present signals for Pmt4p-dependent O-mannosylation. Using specific model proteins, we demonstrate that S. cerevisiae Pmt4p mediates O-mannosylation of Ser/Thr-rich membrane attached proteins, whereas Pmt1p/Pmt2p complexes act on both, soluble and membrane bound secretory proteins. The nature of the membrane-anchoring sequence is nonrelevant, as long as it is flanked by a Ser/Thr-rich domain facing the ER lumen. Based on these results, an in silico analysis was performed, which identified previously uncharacterized Pmt4p substrates in S. cerevisiae. Our work shows that, in contrast to many other types of glycosylation, Pmt4p O-mannosylation signals are not just linear sequences of proteins but instead are highly complex.

Results

In quest of determinants that bring about O-mannosylation and/or define specificity of PMTs toward different protein substrates, we performed in silico analyses of known PMT substrates. No obvious signals could be detected at the level of the primary and secondary structure of the proteins (data not shown). However, we realized that all Pmt4p substrates characterized to date are membrane-associated. Thus, we hypothesized that membrane association is a prerequisite of Pmt4p-catalyzed O-mannosylation.

Disruption of Membrane Attachment of FUSwTM^ZZ Changes Its Specificity for PMT Complexes.

To follow up our hypothesis, we designed model substrates derived from the Pmt4p substrate Fus1p (Fig. 1 A–G). Fus1p is a type I integral membrane protein involved in cell fusion during mating (19, 20). It was shown that Pmt4p mannosylation of the Ser/Thr-rich extracellular domain is crucial for cell surface delivery (17).

Fig. 1.

Schematic representation of the fusion proteins used in this study. The boxed sequence represents the entire Fus1p extracellular domain; signal peptide is underlined. Constructs are fused to a C-terminal protein A-tag (ZZ). TM, transmembrane domain; AA, amino acids. Stippled, Fus1p; hatching, Axl2p; black, Kre9p; cross-hatching, Gas1p.

We constructed a tandem protein A-tagged (ZZ) version of the Fus1 protein lacking the cytosolic C-terminal domain (Fus1p with transmembrane domain, FUSwTM^ZZ; Fig. 1B). To determine whether this construct was O-mannosylated we expressed it in WT and various pmtΔ mutants (pmt1Δ, pmt2Δ, pmt1pmt2Δ, pmt3Δ, pmt4Δ, and pmt6Δ). In Western blot analyses of FUSwTM^ZZ expressed in WT cells (Fig. 2A), we detected four specific bands similar to the pattern described for the full-length protein which are due to the processing of Fus1p (17). The same pattern was observed when FUSwTM^ZZ was expressed in pmt1Δ, pmt2Δ, pmt1pmt2Δ, pmt3Δ, and pmt6Δ mutants, indicating that in these mutants the protein was processed normally (Fig. 2A). In contrast, in pmt4Δ cells FUSwTM^ZZ migrated at the level of the predicted molecular mass of the unglycosylated protein (27.6 kDa). This band was not labeled by the lectin Con A, which binds to O-mannosyl glycans [supporting information (SI) Fig. 8]. To further confirm that the 27.6-kDa protein is nonglycosylated FUSwTM^ZZ, we expressed the protein in the thermosensitive sec53 mutant. SEC53 encodes a phospho-mannomutase and when sec53 mutant cells are incubated at 37°C (restrictive temperature), both N- and O-glycosylation are blocked (21, 22). In sec53 cells, FUSwTM^ZZ accumulates at 37°C as a single band with the same mobility as the band detected in pmt4Δ cells (Fig. 2B). Our data show that the vast majority of FUSwTM^ZZ is specifically mannosylated by Pmt4p. In pmt4Δ cells, a very minor fraction is still processed to the mature form detected in WT cells (Fig. 2A). This residual activity is probably due to a compensatory action of one or more of the remaining Pmt proteins, which mannosylate a fraction of FUSwTM^ZZ that accumulates in the absence of Pmt4p.

Fig. 2.

Membrane attachment is a prerequisite for Pmt4p-mediated O-mannosylation. Western blot analyses of FUSwTM^ZZ (A and B) and intracellular FUSw/oTM^ZZ (C and D) expressed in WT and different mutant strains as indicated. Crude membranes isolated from ≈4 × 10⁶ cells were analyzed unless indicated differently. (A) FUSwTM^ZZ is hypoglycosylated in the pmt4Δ mutant. (Left) Ten micrograms of membrane protein was analyzed. (B) The form of FUSwTM^ZZ that is produced in the pmt4Δ mutant migrates at the level of the unglycosylated protein produced in sec53 mutant cells at 37°C. (C and D) Intracellular FUSw/oTM^ZZ is not O-mannosylated (C) and migrates at the level of the unglycosylated protein produced in sec53 mutant cells at 37°C (D).

To further analyze FUSwTM, a green fluorescent protein fusion (FUSwTM^GFP) was expressed in WT and pmt4Δ cells. Whereas in WT cells, FUSwTM^GFP localized to the plasma membrane (Fig. 3), transport of FUSwTM^GFP to the cell surface was reduced in pmt4Δ cells, and the protein accumulated intracellularly (data not shown).

Fig. 3.

Cellular localization of FUSwTM^GFP and intracellular FUSw/oTM^GFP in WT cells. (Upper) FUSwTM^GFP localizes mainly to the plasma membrane. (Lower) Deletion of the TM results in intracellular accumulation of FUSw/oTM^GFP.

In summary, our data show that the deletion of the cytosolic domain of Fus1p has no influence on O-mannosylation by Pmt4p. FUSwTM^ZZ shows very similar features as native Fus1p and therefore represents an ideal basic model protein to analyze signals for Pmt4p-mediated O-mannosylation.

To test the role of the transmembrane domain (TM) for Pmt4p-dependent mannosylation, we deleted the TM from our model substrate, resulting in FUSw/oTM^ZZ (Fus1p without transmembrane domain; Fig. 1C). The protein was expressed in WT and in the pmtΔ mutants and analyzed by Western blotting (Fig. 2C). Cell fractionation showed that in WT cells FUSw/oTM^ZZ is present in the intracellular soluble and the membrane fraction (Fig. 2C and data not shown). The intracellular protein shows an apparent molecular mass of 23.3 kDa. Accumulation of this protein in mutant sec53 at 37°C (Fig. 2D) and lacking ConA staining (SI Fig. 8) confirmed that it is the unglycosylated FUSw/oTM^ZZ. Expression of FUSw/oTM^GFP in WT cells (Fig. 3) showed that the protein localizes intracellularly as previously described for native Fus1p in pmt4Δ cells (17). These data demonstrate that deletion of the TM abolishes Pmt4p-mediated O-mannosylation.

Interestingly, we found that WT cells secrete a minor fraction of FUSw/oTM^ZZ into the culture medium. Extracellular FUSw/oTM^ZZ showed an apparent molecular mass of ≈55 kDa (Fig. 4A). A partially glycosylated precursor of the secreted protein accumulated in temperature sensitive sec18 cells (23, 24), when ER exit is blocked at restrictive temperature (Fig. 4B). In the mutants pmt3Δ, pmt4Δ, and pmt6Δ, FUSw/oTM^ZZ with the same molecular mass as observed in WT cells is secreted (Fig. 4A). In contrast, a shift to a lower molecular mass is observed in the mutants pmt1Δ, pmt2Δ, and pmt1pmt2Δ (Fig. 4A). These data suggest that FUSw/oTM^ZZ is O-mannosylated by PMT1/PMT2-family members. Because only a minor fraction is glycosylated, FUSw/oTM^ZZ does not contain potent signals for PMT1/PMT2-family mediated O-mannosylation.

Fig. 4.

Disruption of membrane attachment changes specificity for PMT complexes. (A) Western blot analyses of culture medium from WT and pmtΔ mutants expressing FUSw/oTM^ZZ. The extracellular protein is hypoglycosylated in pmt1Δ, pmt2Δ, and pmt1pmt2Δ. (B) Extracellular FUSw/oTM^ZZ accumulates in the ER of sec18 mutant cells at 37°C.

In summary, disruption of membrane integration of our model substrate leads to a complete loss of Pmt4p-mediated O-mannosylation and permits modification by Pmt1p/Pmt2p complexes. We conclude that membrane localization of target proteins plays an important role for recognition by Pmt4p.

The Distance of O-Mannosylation Sites from the Membrane Does Not Influence Pmt4p-Dependent Glycosylation.

To address the question whether Pmt4p complexes preferentially mannosylate Ser/Thr-rich protein domains that are in immediate proximity of the ER membrane, we introduced spacer sequences between the Ser/Thr-rich domain and the TM of FUSwTM^ZZ. We generated two constructs, FUSwTM4AA^ZZ (Fig. 1D) and FUSwTM10AA^ZZ (Fig. 1E), which carry 4- and 10-aa spacers of alternating glycine and alanine residues, respectively. As expected, in WT and pmt4Δ cells, the proteins localized to the membrane fraction (data not shown). For both constructs, a pattern similar to that of Fus1p and FUSwTM^ZZ was observed (compare Figs. 5A and 2A) showing accumulation of hypoglycosylated protein in the pmt4Δ mutant, which indicates that FUSwTM^ZZ, FUSwTM4AA^ZZ, and FUSwTM10AA^ZZ are processed in a similar, Pmt4p-dependent way. We conclude that the distance of the target sequence to the ER membrane has no influence on O-mannosylation by Pmt4p, at least in a range of up to 10 aa.

Fig. 5.

The nature of the membrane-anchoring sequence does not affect Pmt4p-mediated mannosylation. Shown are Western blot analyses of model proteins expressed in WT and pmtΔ mutants. (A and C) Crude membranes (10 μg of protein) were analyzed. (B and D) Crude membranes isolated from ≈4 × 10⁶ cells were analyzed. (A) Introduction of 4 or 10 spacing amino acids did not alter the Pmt4p-dependent processing of FUSwTM^ZZ. (B and C) Exchange of the Fus1p TM for TMs derived from Axl2p or Can1p did not disrupt O-mannosylation by Pmt4p. Both, FUS-AXL2TM^ZZ (B) and FUS-CAN1TM^ZZ (C) are hypoglycosylated in pmt4Δ. (D) Attachment of the TM from Can1p to the soluble Pmt1p/Pmt2p substrate Kre9p triggers O-mannosylation by Pmt4p.

Membrane Localization but Not Intrinsic Features of the TM Mediate O-Mannosylation by Pmt4p.

Our experiments showed that the TM of FUSwTM^ZZ is crucial for O-mannosylation by Pmt4p. Thus, we asked whether intrinsic signals for Pmt4p-dependent O-mannosylation exist in the TMs of Pmt4p substrates. For this reason, we exchanged the TM of FUSwTM^ZZ for TMs of other Pmt4p substrates (Axl2p) and non-substrates (Can1p).

The construct FUS-AXL2TM^ZZ (Fig. 1F) resulted from the fusion of the extracellular N-terminal domain of Fus1p with the TM and the cytosolic part of the type I membrane protein Axl2p. Axl2p is a Pmt4p substrate and O-mannosylation affects stability and localization of the protein thus resulting in abnormal budding pattern (25). Cell fractionation experiments showed that FUS-AXL2TM^ZZ is exclusively membrane-associated in WT and pmt4Δ mutants (data not shown). Western blot analyses revealed underglycosylation of the protein in the pmt4Δ mutant (Fig. 5B), demonstrating that FUS-AXL2TM^ZZ is mannosylated by Pmt4p. Thus, the exchange of the Fus1p TM for the TM of another Pmt4p target does not affect the interaction with Pmt4p.

The next step in the analysis was to substitute the TM of FUSwTM^ZZ with a TM derived from a protein that is not O-mannosylated. We selected TM XII of the plasma membrane arginine permease Can1p which is oriented in the same way as the Fus1p TM (26, 27). The Can1p TM was fused to the extracellular N-terminal domain of Fus1p resulting in FUS-CAN1TM^ZZ (Fig. 1G). In WT and pmt4Δ cells, FUS-CAN1TM^ZZ localized specifically to the membrane fraction (data not shown) and showed a similar band pattern as observed for FUSwTM^ZZ (compare Figs. 5C and 2A). In the pmt4Δ mutant, a major fraction of the protein was hypoglycosylated (SI Fig. 8) and showed the molecular mass of the unglycosylated protein (26.2 kDa), suggesting a similar processing of FUSwTM^ZZ, FUS-CAN1TM^ZZ, and native Fus1p. These results show that FUS-CAN1TM^ZZ is a substrate for Pmt4p, although the TM is derived from a non-substrate. Our experiments exclude the possibility of primary sequence based signals for Pmt4p-mediated O-mannosylation encoded in TMs of specific Pmt4p substrates.

To further confirm this result, we designed a gain-of-function approach by converting a non-substrate protein into a Pmt4p substrate. Therefore, we fused TM XII of Can1p (see above) to a part of the soluble secreted protein Kre9p. Kre9p is involved in β1,6-glucan assembly and is O-mannosylated exclusively by Pmt1/Pmt2p complexes (15). The resulting construct KRE9-CAN1TM^ZZ (Fig. 1H) covered the Kre9p-derived N-terminal signal sequence, the Ser/Thr-rich target sequence for Pmt1p/Pmt2p-mediated O-mannosylation and TM XII from Can1p. In WT, pmt4Δ, and pmt1pmt2Δ mutant strains, the fusion protein KRE9-CAN1TM^ZZ localized to the membrane fraction (data not shown). In WT cells, we detected two bands with markedly decreased mobility compared with the predicted molecular mass of 46.3 kDa of the nonglycosylated protein, indicating that KRE9-CAN1TM^ZZ is O-mannosylated (Fig. 5D). In mutant pmt4Δ or pmt1pmt2Δ, respectively, KRE9-CAN1TM^ZZ is hypoglycosylated and in addition to the dominant bands around ≈50 kDa, further hypoglycosylated and/or degraded low molecular mass forms appear (Fig. 5D). These results indicate that Pmt1p/Pmt2p complexes as well as Pmt4p complexes participate in the processing of KRE9-CAN1TM^ZZ.

In summary, the addition of a single TM (derived from a non-substrate of O-mannosylation) to a soluble Ser/Thr-rich protein domain (derived from a Pmt1p/Pmt2p substrate) triggered the recognition and modification by Pmt4p.

Deletion of the Glycosylphosphatidylinositol (GPI) Anchor of Gas1p Diminishes Pmt4p-Dependent O-Mannosylation.

Next we addressed the question whether the type of membrane association is relevant for the modification by Pmt4p and examined the GPI-anchored plasma membrane protein Gas1p (Fig. 1I), a β1,3-glucanosyltransferase required for cell wall assembly (28). The protein is O-mannosylated by Pmt4p (15) and additionally N-glycosylated. We performed Western blot analyses of the endogenous Gas1p from WT, pmt4Δ, and pmt1pmt2Δ mutants to confirm these findings (Fig. 6Upper). Gas1p is hypoglycosylated only in the pmt4Δ mutant, and Pmtl/Pmt2p complexes obviously do not recognize the protein as substrate for mannosyl transfer. To ensure that the change in molecular mass of Gas1p in the pmt4Δ mutant is due to a reduced amount of O-linked glycans, N-linked carbohydrates were removed by treatment with endo-N-acetylglucosaminidase H (EndoH). Although EndoH treatment reduced the apparent molecular mass by ≈30 kDa, the mass difference between Gaslp from WT and pmt4Δ strains remained (Fig. 6 Upper).

Fig. 6.

The type of membrane association is nonrelevant for Pmt4p-mediated modification. Shown are Western blot analyses of endogenous Gas1p and GAS1ΔGPI^ZZ from WT and pmtΔ mutants. N-glycans were removed by EndoH as indicated. (Upper) Gas1p is hypoglycosylated only in pmt4Δ. (Lower) GAS1ΔGPI^ZZ is no longer mannosylated and accumulates intracellularly.

To test the impact of the GPI anchor on the O-mannosylation state of Gas1p, we constructed a tagged truncated version of the protein lacking the C-terminal hydrophobic GPI anchor signal thus giving rise to the soluble secretory protein GAS1ΔGPI^ZZ (Fig. 1J). In WT, a major band with an apparent molecular mass of ≈120 kDa could be detected (Fig. 6 Lower). After EndoH treatment, the apparent molecular mass was decreased by ≈30 kDa indicating similar N-glycosylation as compared with the native Gas1p protein. In pmt4Δ and pmt1pmt2Δ mutant cells, the same results were obtained, indicating that deletion of the GPI anchor attachment site also eliminated Pmt4p O-mannosylation. Interestingly, hypoglycosylated GAS1ΔGPI^ZZ was not secreted into the medium but rather accumulated intracellularly (Fig. 6 Lower and data not shown).

From these results, we conclude that disruption of membrane attachment of Gas1p strongly affects its ability to interact with Pmt4p complexes and that, in accordance with previously published results (17), O-mannosylation, especially that mediated by Pmt4p, plays an important role in the secretion of proteins.

In Silico Identification of Pmt4p Substrates.

Our data demonstrate that Pmt4p specifically acts on secretory proteins with an ER-luminally oriented Ser/Thr-rich region flanked by a membrane anchor. Based on these results, we performed an in silico search for putative Pmt4p substrates in S. cerevisiae. We screened the proteome of S. cerevisiae for proteins with at least one TM (found with a sliding window of 18 aa) and containing a region of >20 aa where the percentage in Ser/Thr is 40% or higher. Of 5,888 sequences scanned, 51 confirmed to these criteria (SI Table 1). Twenty of these proteins are putative GPI-anchored proteins and 31 are integral membrane proteins. Out of the last group, we cloned tagged versions of Opy2p [an integral membrane protein that functions in the signaling branch of the high-osmolarity glycerol pathway (29)], Prm5p [a pheromone-regulated protein that is predicted to have one TM and is induced during cell integrity signaling (30, 31)], Rax2p [which is involved in the maintenance of bud site selection during bipolar budding (32)], and YNL176c and analyzed them in WT and pmt4Δ mutant strains. For the tagged proteins Opy2p, Prm5p, Rax2p, and YNL176c, we could find a shift to lower molecular weights or an accumulation of incompletely processed forms of the proteins in the pmt4Δ mutant, indicating Pmt4p-dependent processing (Fig. 7). The in silico identification of previously uncharacterized Pmt4p substrates confirms that membrane association is a prerequisite of Pmt4p-mediated protein O-mannosylation.

Fig. 7.

In silico screening identified previously uncharacterized Pmt4p substrates. Shown are Western blot analyses of the protein A-tagged proteins Opy2p, Prm5p, Rax2p, and YNL176c from WT and mutant pmt4Δ. Incompletely processed forms of the proteins accumulate in mutant pmt4Δ.

Discussion

Using specific model proteins, we identified signals that determine Pmt4p-dependent O-mannosylation in yeast. We demonstrate that, in contrast to many other types of glycosylation, Pmt4p O-mannosylation signals are not linear sequences of proteins. Pmt4p mediates O-mannosylation of proteins, which are membrane attached and bear a Ser/Thr-rich domain facing the ER lumen.

The only Pmt4p substrate described so far, which at the first glance did not comply with our predictions, was S. cerevisiae Ccw5/Pir4p (16). Ccw5p is a cell wall mannoprotein that is attached to β-1,3-glucan and does not contain an obvious TM or GPI attachment site. However, cell fractionation studies revealed that Ccw5p is membrane-associated during secretion to the cell surface (J.H. and S.S., unpublished data), thus meeting the prerequisites of a Pmt4p substrate. Further, our conclusions are in agreement with the previously published observation that chimeric membrane proteins derived from Fus1p, Mid2p and invertase (Suc2p) were O-mannosylated in a Pmt4p-dependent manner when they carried a Ser/Thr-rich target sequence (17). Moreover, in mammals, the only well characterized substrate of the Pmt4p homologue POMT1 (α-dystroglycan) is membrane-associated (33). Dystroglycan is translated as a type I ER membrane propeptide that is then further processed into two subunits (α and β) during secretion (33). The O-mannosylated Ser/Thr-rich domain of dystroglycan propeptide is separated from the TM by ≈266 aa, suggesting that in mammals further determinants might have evolved.

Our data prove that Pmt4p mannosylates exclusively membrane-bound proteins whereas Pmt1p/Pmt2p complexes act on both soluble and membrane proteins. What could be the molecular base of that specificity? There are several options. (i) The first option could be differences in the kinetics of O-mannosyl transfer catalyzed by Pmt4p versus Pmt1p/Pmt2p complexes and thus residence time of protein substrates at the ER membrane. Similar to N-glycosylation, O-mannosyl transfer occurs while proteins are translocated into the ER lumen (1). Membrane anchoring increases the residence time of nascent proteins at the ER membrane and thus might aid and abet Pmt4p-mediated O-mannosylation. (ii) The second option could be association of PMTs with different translocon complexes in the ER. It was recently reported that two isoforms of the oligosaccharyl transferase (OT) complex associate specifically with two different translocon complexes. Ost3p containing OT complexes associate specifically with the Sec61 translocon, whereas Ost6p containing OT complexes with the Ssh1 translocon (34). Because O-mannosylation and N-glycosylation are competitive processes (16), it is conceivable that distinct PMT complexes associate specifically with translocon–OT supercomplexes providing the molecular base of substrate specificity. (iii) A third option could be positioning of Pmt4p complexes in specific microdomains of the ER membrane. In yeast, the existence of distinct membrane microdomains has been reported that are already formed in the ER (35, 36). Interestingly, the Pmt4p substrates Fus1p, Gas1p, Wsc1p, and Mid2p are associated with such detergent resistant membrane fractions (36, 37). Thus, Pmt4p complexes and the corresponding substrates might colocalize in specific ER membrane microdomains. Because we cannot eliminate any of these possibilities at present, one of the main future tasks will be to elucidate the molecular mode of operation of the O-mannosylation machinery in the ER.

Based on our results, we screened the S. cerevisiae protein database for novel Pmt4p substrates and identified 51 candidates (20 putative GPI-anchored and 31 integral membrane proteins). Among them are the so far known Pmt4p substrates Fus1p (17), Axl2p (25), Gas1p, Kex2p (15), Mid2p, Wsc1p, and Wsc2p (18). Analyses of the glycosylation status of selected putative substrates (Opy2p, Prm5p, Rax2p, and YNL176c) indeed confirmed their O-mannosylation by Pmt4p. Pmt4p-mediated O-mannosylation is crucial for the stability, sorting and/or performance of proteins which are functionally of high relevance for cell growth and development as demonstrated for Axl2p, Mid2p and Fus1p (19, 25, 38). Thus, our findings provide a tool to identify proteins that are potential major players in various cellular processes in yeast and other fungi. The identification of such proteins is highly significant because fungal pathogens represent the major eukaryotic disease causing agents in humans as well as in agricultural important crop plants. Thus, it is important to identify new targets for the development of novel antifungal drugs. In addition, we are now able to screen genomes of higher eukaryotes for putative PMT substrates which might help to elucidate the relevance of O-mannosyl glycans for early stages of development and for vital physiological functions of proteins in mammals and humans (1).

Materials and Methods

Yeast Strains and Plasmids.

Used for this study were the S. cerevisiae WT SEY6210 (MATα, his3-Δ200, leu2-3, -112, lys2-801, trp1-Δ901, ura3-52, suc2-Δ9) (39); sec mutants SEY5188 (MATα, sec18-1, suc2-09, leu2-3,112 ura3-52) (24) and SFNY28-6C (MATa, sec53, ura3-52) (17); and pmt deletion strains pmt1Δ (isogenic to SEY6210, pmt1::HIS3) (40), pmt2Δ (isogenic to SEY6210, pmt2::LEU2) (40), pmt3Δ (isogenic to SEY6210, pmt3::HIS3) (41), pmt6Δ (isogenic to SEY6210, pmt6::LEU2) (41), and pmt1pmt2Δ (isogenic to SEY6210, pmt1::HIS3, pmt2::LEU2) (40). Strain STY100 (isogenic to SEY6210, pmt4::KanMX) was derived by disruption of the PMT4 gene by using a KanMX disruption cassette released from pSB119 by NotI digestion.

Yeast strains were grown under standard conditions and transformed following the method of Gietz et al. (42). Standard procedures were used for all DNA manipulations. S. cerevisiae genes indicated below were amplified by PCR on genomic DNA.

For plasmid pSB119, megaprimers were amplified by PCR on S. cerevisiae genomic DNA. The megaprimers were used for PCR on pFA6a-GFP-kanMX6 (43) to amplify the KanMX cassette. The PCR product was cloned into pGEM-Teasy (Promega, Mannheim, Germany). For plasmid pMS3, Schizosaccharomyces pombe adh1 promoter sequence derived from pREP3-adh (44) was cloned as a HindIII/XhoI fragment into pYEplac195ZZ (gift from J. Stolz). For plasmids pMS7.1 (FUSw/oTM^ZZ) and pMS7.2 (FUSwTM^ZZ), fragments of the FUS1-coding sequence (bp 1–290 or 1–342) were amplified by PCR and cloned into pMS3 by using XhoI/BamHI. For plasmids pMS9.1 (FUSw/oTM^GFP) and pMS9.2 (FUSwTM^GFP), S. pombe adh1 promoter and FUS1-coding sequences (bp 1–290 or 1–342) derived from pMS7.1 and pMS7.2, respectively, were cloned as BamHI/HindIII fragments into YEplac195-GFP (gift from J. Stolz). For plasmid pJH1(FUS-AXL2TM^ZZ), a fragment of the AXL2-coding sequence (bp 1405–2469) was amplified by PCR and cloned into pMS7.1 by using BamHI/NheI. For plasmids pJH6.1 and pJH6.2 (FUSwTM4AA^ZZ and FUSwTM10AA^ZZ), bp 220–342 of the FUS1-coding sequence were amplified by PCR and cloned into pMS7.1 by using BamHI/NheI, resulting in plasmid pJH5. Oligonucleotide 573 (5′-GATCCGGCGCCGGCGCCGGCGCCGGCGCCGGCGCCG-3′) was annealed and ligated into BamHI-digested pJH5. For plasmid pJH8 (FUS-CAN1TM^ZZ), part of the CAN1 coding sequence (bp 1561–1653) was amplified by PCR and cloned into pMS7.1 by using BamHI/NheI. For plasmid pJH12 (KRE9-CAN1TM^ZZ), part of the KRE9-coding sequence (bp 1–768) was amplified by PCR and cloned into pJH8 by using XhoI/BamHI. For plasmids pJH15 (OPY2^ZZ), pJH16 (PRM5^ZZ), pJH17 (RAX2^ZZ), and pJH20 (YNL176c^ZZ), coding sequences lacking the termination codon were amplified by PCR and products were cloned into pMS3 digested with XhoI/BamHI. For plasmid pJH23 (GAS1ΔGPI^ZZ), part of the GAS1-coding sequence (bp 1–1584) was amplified by PCR and cloned into pMS3 by using XhoI/BamHI.

Preparation of Cell Extracts.

Yeast cells (5 × 10⁸) from an exponentially growing culture were harvested and washed with 20 ml 10 mM NaN₃. Cell fractionation was performed as described (14). To enrich secreted proteins, culture media were concentrated by using Vivaspin 500 centrifugal filter units (Sartorius, Goettingen, Germany) with a molecular weight cut-off of 10,000.

Protein Analyses in Temperature-Sensitive sec Mutants.

Yeast cultures were grown to logarithmic phase at 25°C. At time 0, 10⁸ cells were harvested, washed with NaN₃ (10 mg/ml), and frozen in liquid nitrogen. Cultures were then shifted to restrictive temperature (37°C). Samples were taken at different times, and cell extracts were prepared as described above.

Deglycosylation with Endo-β-N-Acetylglucosaminidase H.

Ten to thirty micrograms of protein were treated with endo-β-N-acetylglucosaminidase H (EndoH; Calbiochem, Darmstadt, Germany) according to the manufacturer's instructions. Mock incubations were carried out without EndoH.

Western Blot Analysis.

Proteins were fractionated by SDS/PAGE and transferred to nitrocellulose. Polyclonal anti-Gas1p antibodies were used at a dilution of 1:2,500. Peroxidase-coupled anti-mouse-IgG antibody from rabbit (Sigma, Munich, Germany) and peroxidase-coupled anti-rabbit-IgG antibody from goat (Sigma) were used at a dilution of 1:10,000. Protein–antibody complexes were visualized by using the SuperSignal West Pico Chemiluminescent System (Pierce, Bonn, Germany).

Light Microscopy.

Cells were immobilized by 0.8% agarose before microscopic observation. Specimens were viewed by using an LSM510-Meta confocal microscope (Carl Zeiss, Jena, Germany) with ×100 PlanApochromat objective (numerical aperture 1.4). Fluorescence signal of GFP (excitation 488 nm, Ar laser) was detected by using a bandpass emission filter 505–530 nm.

In Silico Identification of Pmt4p Substrates.

We wrote a computer program that, in a first step, identified the proteins in the data set that have a transmembrane segment, using the Kyte and Doolittle algorithm (45). The selected proteins were then subjected to a search for a region of at least 20 aa rich in serine or threonine residues (at least 40%) adjacent to the transmembrane segment.

Abbreviations

TM

transmembrane domain

PMT

protein O-mannosyltransferase

ER

endoplasmic reticulum

GPI

glycosylphosphatidylinositol

OT

oligosaccharyl transferase.