Members of the Evolutionarily Conserved PMT Family of ProteinO-Mannosyltransferases Form Distinct Protein Complexes among Themselves*

Protein O-mannosyltransferases (PMTs) initiate the assembly of O-mannosyl glycans, an essential protein modification. Since PMTs are evolutionarily conserved in fungi but are absent in green plants, the PMT family is a putative target for new antifungal drugs, particularly in fighting the threat of phytopathogenic fungi. The PMT family is phylogenetically classified into PMT1, PMT2, and PMT4 subfamilies, which differ in protein substrate specificity. In the model organism Saccharomyces cerevisiae as well as in many other fungi the PMT family is highly redundant, and only the simultaneous deletion of PMT1/PMT2 and PMT4 subfamily members is lethal. In this study we analyzed the molecular organization of PMT family members in S. cerevisiae. We show that members of the PMT1 subfamily (Pmt1p and Pmt5p) interact in pairs with members of the PMT2 subfamily (Pmt2p and Pmt3p) and that Pmt1p-Pmt2p and Pmt5p-Pmt3p complexes represent the predominant forms. Under certain physiological conditions, however, Pmt1p interacts also with Pmt3p, and Pmt5p with Pmt2p, suggesting a compensatory cooperation that guarantees the maintenance ofO-mannosylation. Unlike the PMT1/PMT2 subfamily members, the single member of the PMT4 subfamily (Pmt4p) acts as a homomeric complex. Using mutational analyses we demonstrate that the same conserved protein domains underlie both heteromeric and homomeric interactions, and we identify an invariant arginine residue of transmembrane domain two as essential for the formation and/or stability of PMT complexes in general. Our data suggest that protein-protein interactions between the PMT family members offer a point of attack to shut down overall proteinO-mannosylation in fungi.

grade transport of misfolded proteins across the membrane of the endoplasmic reticulum (ER) 1 (11). O-mannosylation is not only important in yeast, but also in mammals. It was recently shown that in humans, O-mannosyl glycosylation represents a new pathomechanism for muscular dystrophy and neuronal migration disorders (12,13).
In yeast and fungi, O-mannosylation is initiated in the lumen of the ER by an essential family of protein O-mannosyltransferases (PMTs). These enzymes catalyze the transfer of mannose from dolichyl phosphate-activated mannose (Dol-P-Man) to serine or threonine residues of secretory proteins (2). In Saccharomyces cerevisiae, a total of seven PMT family members (Pmt1-7p) have been identified, which share almost identical hydropathy profiles that predict the PMTs to be integral membrane proteins with multiple transmembrane domains. Pmt1-6 proteins feature an overall protein sequence identity of 57.5%. Pmt7p is less conserved. Protein O-mannosyltransferase activity has been demonstrated for Pmt1-4p and Pmt6p (14). Aside from S. cerevisiae PMTs, orthologues are known from many other yeast and fungi, for example, Candida albicans (CaPMT1-2 and CaPMT4 - 6) and Schizosaccharomyces pombe (SpPMT1, SpPMT3, and SpPMT4) (Refs. 3 and 4). 2 Moreover, PMT homologues have been also identified in many multicellular eukaryotes such as Drosophila melanogaster, mouse and, humans (15)(16)(17). Despite their evolutionarily conservation in fungi and throughout the animal kingdom (with the exception of Caenorhabditis elegans), PMTs are not present in green plants. 2 This makes the PMT family in fungi especially attractive as a target for the development of new antifungal drugs in order to combat phytopathogenic fungi.
The protein O-mannosyltransferases can be divided into three subfamilies: PMT1, PMT2, and PMT4, which include transferases closely related to S. cerevisiae Pmt1p, Pmt2p, and Pmt4p, respectively (17,18). Members of the PMT1 and PMT2 subfamilies show marked similarities and distinctions from PMT4 subfamily members. First, all PMT family members share three conserved sequence motifs but, these show significant variations between PMT1/PMT2 and PMT4 subfamily members (18). Second, the PMT1/PMT2 and PMT4 subfamilies use distinct acceptor protein substrates in vivo (9,14). Third, in fungi the PMT1/PMT2 subfamily is highly redundant, whereas the PMT4 subfamily has only one representative per species (17,18).
Among the PMT family members, Pmtp1 from S. cerevisiae has been most extensively characterized. Pmt1p is an integral ER membrane glycoprotein with seven transmembrane-spanning domains (19). Its N terminus faces the cytoplasm whereas the C terminus faces the lumen of the ER. Two major hydrophilic domains that are located between transmembrane spans one and two (loop 1) and transmembrane spans five and six (loop 5), respectively, are oriented toward the ER lumen and are essential for Pmt1p activity (18,19). The replacement of invariant amino acid residues in these regions suggested that these segments are involved in the recognition and/or binding of protein substrates and/or catalysis (18). Comparison of PMTs from different organisms defined highly conserved peptide motifs present in loop 5 (18), which are also found in IP 3 and ryanodine receptors. Their common function is unknown (20). Pmt1p forms a heteromeric complex with Pmt2p in vivo, and this complex formation is essential for maximal mannosyltransferase activity (18,21). N-and C-terminal regions of Pmt1p are involved in Pmt1p-Pmt2p interactions (18). Other than Pmt1p, very little is known about the molecular organization of the rest of the O-mannosylation machinery.
In the present study we analyzed the molecular assembly of the PMT family in yeast. We demonstrate that complex formation is of general validity for all members of the PMT family in S. cerevisiae. Strikingly, members of the PMT1 subfamily form specific heteromeric complexes with members of the PMT2 subfamily in vivo, while Pmt4p acts as a homomeric complex. Despite the differences between PMT1/PMT2 and PMT4 complexes, we show that the same rules and residues govern Pmtp protein-protein interactions.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-The S. cerevisiae strains are listed in Table I. Yeast strains were grown under standard conditions and transformed following the method of Gietz et al. (27) with the yeast shuttle vectors pRS423 (28), YEp352 (29), pSB53 (19), pSB56 (18), PMT2-YEp352 (23), pVG13 (18), pSB114 (18), and the plasmids listed below. Standard procedures were used for all DNA manipulations (30). All cloning and transformations were carried out in Escherichia coli host SURE ® 2 (Stratagene). PCR fragments were routinely checked by sequence analysis. Oligonucleotide sequences are available upon request.
Plasmid pVG80 (PMT2 HA )-A SalI site was introduced downstream of the PMT2 coding region by cloning a 2.96-kb PstI/HindIII fragment from PMT2-YEp352 (23) into pBluescript SK ϩ (Stratagene) digested with the same enzymes. From the resulting plasmid pVG70 a 2.97-kb PstI/SalI fragment was isolated and cloned into YEp352 (cut with PstI and SalI), resulting in plasmid pVG76. A total of six copies of the hemagglutinin (HA) epitope were fused to the C terminus of PMT2 by recombinant PCR (31). Two separate PCR products that overlap in sequence were produced. One was amplified by PCR on pVG76 with the oligonucleotides vg65 and vg66, the other on plasmid pHA-kanMX (gift of U. Schermer) with the oligonucleotides vg67 and vg68. The overlapping, primary PCR products were combined into one longer product using oligonucleotides vg65 and vg68. The resulting 730-bp fragment was cloned into pGEM T-easy (Promega). A 695-bp BglII/SalI fragment of the resulting plasmid pVG78 was subcloned into pVG76 (cut with BglII and SalI). DNA sequence analysis of the resulting plasmid pVG80 (PMT2 HA ) was performed. In the course of this analysis, we realized a discrepancy between the PMT2 sequence we obtained and the yeast data base entry (GenBank TM accession no. AAC04934). To verify the PMT2 sequence, we amplified a 664-bp genomic DNA fragment of PMT2 from the yeast strains S288c, BY4742, W303-1A, and SEY6210 using the oligonucleotides vg63 and vg69. PCR products were cloned into pGEM T-easy, and several independent clones were sequenced. These analyses showed that in contrast to the data base entry, the PMT2 open reading frame contains three additional base pairs (bp ϩ400 to ϩ429 is tgggacttccCttctggGGaaatttaccca; additional bases in capital letters). The insertions result in the predicted amino acid sequence of Pro-Ser-Gly instead of Leu-Leu at position 137 of Pmt2p.
Plasmid pJK4-B1 (PMT4 FLAG )-A copy of the FLAG sequence (32) was obtained by annealing oligonucleotides oligo211 with oligo212. The annealed oligo pair features BamHI and NotI overhang sequences. The FLAG sequence was joined by a three-piece ligation with a 0.7-kb SacI/NotI fragment (isolated from plasmid SAP/EN, Ref. 33) that contains the yeast plasma membrane ATPase terminator (34) and the yeast shuttle vector pRS423 (digested with SacI and NotI) resulting in plasmid pRS423/TER/FLAG. The PMT4 promoter and coding region (bp Ϫ675 to ϩ2286) was amplified from S. cerevisiae genomic DNA using oligonucleotides oligo213 and oligo214. The PCR fragment was digested with BamHI and SalI, and cloned into pRS423/TER/FLAG. Additional FLAG sequences were cloned into the BamHI site of the resulting plasmid pJK4, using the annealed oligonucleotide pair oligo233/ oligo234. In the resulting plasmid (pJK4-B1) four copies of the FLAG epitope were fused to the C terminus of PMT4.
Plasmid pVG45 (PMT4 R142E FLAG )-An arginine residue at position 142 of Pmt4p FLAG was exchanged for a glutamate by site-directed mutagenesis (GeneEditor™, Promega) using the oligonucleotide vg30 and plasmid pVG43 (pUC18, containing bp ϩ24 to ϩ1165 of PMT4). Mutations were confirmed by DNA sequence analysis of the resulting plasmid pVG44. A 476-bp MunI fragment of pVG44 was cloned into pVG42 (pUC18, containing bp ϩ307 to ϩ1165 of PMT4) digested with the same enzyme. From the resulting plasmid pVG69 a 1.87-kb SphI/ HpaI fragment was isolated and cloned into pJK4-B1 resulting in pVG45.
Production of Polyclonal Anti-Pmt3-6p Antibodies in Rabbits-Rabbits were immunized with recombinant fusion proteins consisting of glutathione S-transferase, and the aa Met-1 to Arg-78 of S. cerevisiae Pmt3p, aa Met-1 to Ala-65 of Pmt4p, aa Asp-10 to Thr-121 of Pmt5p, and aa Met-1 to Gln-85 of Pmt6p, respectively. The corresponding DNA fragments were amplified by PCR using genomic S. cerevisiae DNA as template and adapter oligonucleotides oligo151/oligo152 for PMT3, oligo106/oligo107 for PMT4, oligo143b/oligo144 for PMT5, oligo147/ oligo148 for PMT6. The respective DNA fragments were combined with the glutathione S-transferase sequence by EcoRI/BamHI subcloning into a pGEX-2TK expression vector (Amersham Biosciences). The fusion proteins were expressed in E. coli host BL21. The recombinant proteins were excised from SDS-polyacrylamide gels and injected into rabbits. Pineda Antikoerper-Service, Berlin, Germany, performed immunizations. Antibodies were affinity-purified by binding to nitrocellulose derivatized with the glutathione S-transferase fusion protein (35). Preparation of Crude Membranes-Crude membranes were isolated as described (18).
Pmt4p FLAG and FLAG-tagged Pmt4p mutant proteins were solubilized from crude membranes, prepared as described in Ref. 18 using 500 l of solubilization buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.3 mM MgCl 2 , 0.5% Triton X-100, plus protease inhibitors) by vortexing for 30 min at 4°C. The suspension was clarified by centrifugation for 30 min at 20,000 rpm (Sorvall SS34 rotor) to obtain the Triton extract. FLAG-tagged proteins were immunoprecipitated for 1-2 h at 4°C from 300 l of Triton extract, using 30 l of anti-FLAG M2 affinity Gel (Sigma). Precipitates were washed four times with 1 ml of cold solubilization buffer, once with 1 ml of Tris-buffered saline and resuspended in 20 l of 3ϫ SDS-sample buffer.
Chemical Cross-linking of Yeast Membrane Proteins-2.0 ϫ 10 9 yeast cells from a logarithmically growing culture were harvested at 3,000 ϫ g, 4°C and washed once with 20 ml of OPA buffer (50 mM boric acid/sodium tetraborate, pH 8.0). The cell pellet was resuspended in 200 l of OPA buffer plus protease inhibitors. Crude membranes were prepared as described (18). The membrane pellet was resuspended in 20 ml of OPA buffer without protease inhibitors and centrifuged for 30 min at 20,000 rpm (Sorvall, SS34 rotor). Subsequently, membranes were resuspended in 500 l of OPA buffer. For cross-linking, o-phtaldialdehyde (OPA; Sigma) was added to final concentrations of 25-200 M to 100 l of membrane suspension and incubated in the dark for 30 min at 25°C. The reaction was quenched with 100 mM Tris-HCl, pH 6.8.

RESULTS
To uncover common principles that underlie the functionality of Pmtps we investigated whether complex formation is a general feature of the yeast PMT family members. We generated polyclonal antibodies that specifically recognize S. cerevisiae Pmt3p (predicted molecular mass, 86.2 kDa), Pmt4p (predicted mass, 87.8 kDa), Pmt5p (predicted mass, 84.8 kDa), and Pmt6p (predicted mass, 87.9 kDa) in wild-type yeast as shown by Western blot analyses ( Fig. 1, lanes 1, 3, 5, and 7). Extracts from the corresponding pmt deletion strains contained no cross-reactive material, proving that the antibodies are highly specific (Fig. 1, lanes 2, 4, 6, and 8). We also created epitopetagged versions of Pmt2p (Pmt2p HA ) and Pmt4p (Pmt4p FLAG ) (described under "Experimental Procedures"). These tools enabled us to detect PMT complexes isolated by coimmunoprecipitation, BN-PAGE, and chemical cross-linking.
Members of the PMT1 Subfamily Interact in Pairs with Members of the PMT2 Subfamily-To characterize PMT complexes, we performed coimmunoprecipitation experiments using a HA epitope-tagged version of S. cerevisiae Pmt1p (Pmt1p HA , Ref. 18). Pmt1p HA was expressed in a pmt1 deletion strain and solubilized from crude membranes using Triton X-100 and sodium deoxycholate. Pmt1p HA was immunoprecipitated from sodium deoxycholate extracts with monoclonal anti-HA antibodies (see "Experimental Procedures"). The immunoprecipitate and an aliquot of the sodium deoxycholate extract were resolved on 8% SDS-polyacrylamide gels and analyzed by Western blotting and sequentially probing the blots with polyclonal antibodies to Pmt1p to Pmt6p ( Fig. 2A, lanes 1 and 2). To ensure the specificity of the immunoprecipitation reaction the same experiment was performed using strain pmt1⌬ expressing Pmt1p without the HA tag ( Fig. 2A, lanes 3 and 4). As shown previously (21) we confirmed that Pmt2p is the major interacting partner of Pmt1p ( Fig. 2A, lane 2). Moreover, a weak signal for Pmt3p could be specifically detected in the Pmt1p HA immunoprecipitate ( Fig. 2A, compare lanes 2 and 3). The amount of coimmunoprecipitated Pmt3p was small, but this result was highly reproducible. In contrast, Pmt4p, Pmt5p, or Pmt6p could not be detected ( Fig. 2A, lane 2). Coimmunoprecipitation of Pmt1p HA from pmt3 (Fig. 2B, lane 4) and pmt2 (Fig. 2B, lane 6) deletion strains showed that Pmt1p HA interacts with Pmt2p independently of Pmt3p and vice versa.
Pmt4p, or Pmt6p (lane 2) corroborating that Pmt1p interacts with Pmt3p independently of Pmt2p. In addition, in the Pmt2p HA immunoprecipitate a weak signal for Pmt5p could be detected (Fig. 2C, lane 2). The amount of coimmunoprecipitated Pmt5p was small; however, the result was specific (data not shown) and highly reproducible. Furthermore, immunoprecipitation of Pmt2p HA from a pmt1 deletion strain showed that Pmt2p HA interacts with Pmt5p independently of Pmt1p (data not shown).
So far our data indicated that in wild-type yeast Pmt1p and Pmt2p form a dominant protein complex. In addition, we found that Pmt1p also interacts with Pmt3p, and that Pmt2p interacts with Pmt5p. Next we addressed the question of which is the major interacting partner of Pmt3p. Pmt3p was immunoprecipitated from sodium deoxycholate extracts of wild-type and pmt3 mutant strains using polyclonal anti-Pmt3p antibodies (see "Experimental Procedures"). In the Pmt3p immunoprecipitate a small amount of Pmt1p, but not Pmt2p, Pmt4p, or Pmt6p was present (Fig. 2D, lane 2). Only Pmt5p was highly enriched when compared with the input material (Fig. 2D,  compare lanes 1 and 2), demonstrating that Pmt3p predominantly interacts with Pmt5p. Again, the association between Pmt3p and Pmt5p was independent of other Pmt proteins (Fig.  2, A, C, and D and data not shown).
The Pmt1p-Pmt3p and Pmt5p-Pmt2p complexes which we detected in our coimmunoprecipitation experiments are less abundant when compared with Pmt1p-Pmt2p or Pmt5p-Pmt3p complexes. To substantiate that the minor complexes are not formed artificially during coimmunoprecipitation, and to confirm that Pmt1p-Pmt2p and Pmt5p-Pmt3p represent the predominant PMT complexes in wild-type yeast, we performed BN-PAGE (38) that separates native protein complexes. Nonidet P-40 extracts derived from wild-type and pmt1-6 mutant strains were resolved on 4 -13% polyacrylamide gels to separate native PMT complexes (see "Experimental Procedures"). Pmtp-containing complexes were detected by Western blots probed with polyclonal anti-Pmtp antibodies.
Analyses of Pmt1p-containing complexes showed that in Nonidet P-40 extracts from wild-type yeast two protein bands with an apparent mass of ϳ140 kDa and ϳ310 kDa, respectively, were specifically detected by anti-Pmt1p antibodies (Fig.  3A, compare lanes 1 and 7). The ϳ140-kDa band highly likely represents the monomeric Pmt1p. The discrepancy in mass of ϳ140 kDa in BN-PAGE versus 92 kDa in SDS-PAGE (19) is probably due to an abnormal migration behavior of Pmt1p caused by the hydrophobic nature of the protein, an unusual Coomassie Blue to protein ratio, and/or the charge to mass ratio, which is variable in BN-PAGE (41). The formation of Pmt1p homodimers was excluded by coimmunoprecipitation experiments using Pmt1p HA and untagged Pmt1p (data not shown). In addition to monomeric Pmt1p, specific Pmt1p-containing protein complexes with an apparent molecular mass of ϳ310 kDa could be detected (Fig. 3A, lane 1). The ratio of monomeric Pmt1p to Pmt1p-containing high molecular weight complexes varied to some extend in independent experiments (Fig. 3, A and B, lane 1), which is highly likely due to disaggregation of PMT complexes during solubilization. Protein complexes with molecular masses of ϳ300 -320 kDa were also detected by polyclonal anti-Pmt2p, anti-Pmt3p and anti-Pmt5p antibodies (data not shown).
Assuming that in wild-type yeast in vivo Pmt1p-Pmt3p complexes are only present in minor amounts when compared with Pmt1p-Pmt2p, one would expect that in a pmt2 but not in a pmt3 deletion mutant the amount of the ϳ310-kDa complexes recognized by anti-Pmt1p antibodies should decrease dramatically. As shown in Fig. 3A (lanes 2 and 3), this is exactly what we observed. Nevertheless, in the absence of Pmt2p a very small amount of protein complexes with an apparent mass of ϳ310 kDa could be detected after raising the limit of detection of the Western analysis (Fig. 3B, lane 2). These complexes were not only recognized by anti-Pmt1p but also by anti-Pmt3p antibodies (data not shown). In pmt4 -6 mutants Pmt1p-containing complexes were not affected (Fig. 3A, lanes 4 -6). These data corroborate our finding that Pmt1p interacts individually with Pmt2p and Pmt3p; however, Pmt2p represents the major interacting partner. BN-PAGE also showed that Pmt3p-containing protein complexes vanished almost completely only in the absence of Pmt5p (data not shown), consistent with Pmt3p forming an abundant heteromeric complex with Pmt5p.
The amount of Pmt1p-Pmt3p and Pmt5p-Pmt2p complexes in wild-type yeast appeared very minor. Therefore, the question comes up whether these complexes are physiologically relevant. One possibility is that these Pmtp complexes are mainly formed in the absence of their major interacting partners to compensate partially for a lack of protein O-mannosyltransferase activity. To investigate this hypothesis, we analyzed pmt1pmt3 and pmt2pmt5 deletion mutant strains where the preferred interacting partners of Pmt2p/Pmt5p, and Pmt1p/Pmt3p, respectively, are missing. Western analysis of crude membranes isolated from a pmt1pmt3 mutant showed that the amount of both Pmt2p and Pmt5p is increased when compared with wild type (Fig. 4A, compare lanes 1 and 2). Accordingly, in the pmt2pmt5 mutant Pmt3p is more abundant (Fig. 4A, compare lanes 1 and 3). The amount of Pmt1p is not obviously changed, suggesting that Pmt1p is not limiting (Fig.  4A, lane 3). These observations suggested that in the absence of the favored partners Pmt1p-Pmt3p and Pmt5p-Pmt2p complexes accumulate.
To substantiate these data we analyzed Pmt5p-Pmt2p complex formation by chemical cross-linking using OPA. Under the conditions we applied (50 M OPA) in wild-type yeast only upon overexpression of Pmt2p a protein complex with an apparent mass of ϳ152 kDa could be detected which is specifically recognized by polyclonal anti-Pmt5p (Fig. 4B, compare lanes 1 and  3) as well as anti-Pmt2p antibodies (data not shown). In contrast, Pmt5p-Pmt2p complexes could be easily detected in the pmt1pmt3 mutant even without overexpression of Pmt2p (Fig.  4B, lane 4).
To test whether the formation of Pmt5p-Pmt2p and Pmt1p-Pmt3p complexes results in increased O-mannosyltransfer, we determined in vitro O-mannosyltransferase activity in pmt1pmt3 and pmt2pmt5 mutant strains. The in vitro assay system we used preferentially detects O-mannosyltransferase activity of Pmt1p-and Pmt2p-containing complexes (Refs. 14 and 24, Table II). As shown in Table II in vitro O-mannosyltransferase activity is dramatically decreased in pmt1 and pmt1pmt2 mutants when compared with wild-type yeast. Western analysis of crude membranes revealed an equal abundance of Pmt3-6p in pmt1, pmt1pmt2 and wild-type strains (data not shown). In pmt1pmt3 and pmt2pmt5 mutants, in which Pmt5p-Pmt2p and Pmt1p-Pmt3p complexes are formed (see Fig. 4), in vitro O-mannosyltransferase activity is increased by 68.2% and 42.6%, respectively, when compared with pmt1pmt2 mutants (Table II).
In summary, our data show that members of the PMT1 subfamily (Pmt1p and Pmt5p) form distinct complexes with members of the PMT2 subfamily (Pmt2p and Pmt3p), and that Pmt1p-Pmt2p and Pmt5p-Pmt3p pairs are the predominant forms. Under specific conditions Pmt1p-Pmt3p and Pmt5p-Pmt2p complexes can be formed to perpetuate O-mannosyltransfer.
Pmt4p Forms Homomeric Complexes in Vivo-We could not detect interactions between members of the PMT1 or PMT2 subfamily and Pmt4p, the only member of the PMT4 subfamily. In agreement with these results, no other Pmt proteins could be copurified when Pmt4p was precipitated from sodium deoxycholate extracts from a yeast wild-type strain using polyclonal anti-Pmt4p antibodies (data not shown). However, BN-PAGE revealed the presence of larger Pmt4p complexes, which were not influenced by the absence of any other PMT family member (data not shown). We wished to determine whether these complexes were due to Pmt4p-Pmt4p homotypic interactions or whether Pmt4p is associated with other proteins. Therefore, an epitope-tagged version of Pmt4p was constructed  by fusing four copies of the FLAG epitope to the C terminus of Pmt4p (see "Experimental Procedures"). Complementation of the temperature-sensitive phenotype of a pmt1pmt4 mutant by Pmt4p FLAG proved that this construct is fully functional in vivo (Fig. 7). To test for homotypic interactions, PMT4 FLAG was expressed in a S. cerevisiae wild-type strain and crude membranes were prepared. Proteins were solubilized with Triton X-100 (Triton extract), and immunoprecipitation of Pmt4p FLAG was performed using monoclonal anti-FLAG antibodies covalently linked to protein A-Sepharose. Immunoprecipitates were resolved on 8% SDS-polyacrylamide gels. Wild-type Pmt4p and Pmt4p FLAG , which differ in molecular mass by 4.8 kDa, were detected on a Western blot probed with polyclonal anti-Pmt4p antibodies. As shown in Fig. 5A Pmt4p FLAG specifically coimmunoprecipitates wild-type Pmt4p (compare lane 4 with lanes 5 and 6), indicating that Pmt4p is present in homomeric complexes in vivo. Chemical cross-linking experiments corroborated these results. Pmt4p FLAG was expressed in a pmt4 deletion strain, crude membranes were prepared, and crosslinking was performed using OPA at final concentrations of between 25 and 100 M. Proteins were resolved on 8% SDSpolyacrylamide gels and analyzed by Western blotting using anti-Pmt4p antibodies as probe. Fig. 5B (lane 2) shows that in the absence of OPA, Pmt4p FLAG migrates with an apparent mass of ϳ90 kDa, which is in agreement with a deduced mass of 92.6 kDa. Upon addition of OPA, larger complexes with an apparent molecular mass of ϳ165 kDa could be detected (Fig. 5B, lanes  3-5), consistent with the formation of homodimeric Pmt4p complexes. Summarizing, our data suggest that Pmt4p forms homomeric complexes; however, the association with other smaller molecular weight molecules cannot be ruled out completely.

Conserved Protein Domains Underlie Heteromeric Pmt1p-Pmt2p and Homomeric Pmt4p-Pmt4p
Interactions-Because in contrast to the other PMT family members Pmt4p forms homomeric complexes, we asked whether common principles underlie homomeric and heteromeric PMT complex formation and/or stability. We previously showed that a large hydrophilic endoplasmic reticulum-oriented segment of Pmt1p (loop 5, aa 294 -586) is crucial for mannosyltransferase activity but not for Pmt1p-Pmt2p complex formation (Ref. 18, see also Fig. 7). To test whether the same may be the case for Pmt4p we deleted the large predicted luminal loop 5 region of Pmt4p FLAG (aa 394 -521, Fig. 6A) and expressed the internal deletion construct (⌬loop5) in pmt4 mutant and wild-type yeast strains. Pmt4p complex formation was assayed by chemical cross-linking. Fig. 6B shows that in the presence of OPA, larger complexes with an apparent mass of ϳ140 kDa (compare lanes 1 and 2) can be detected in the pmt4 mutant strain in addition to monomeric ⌬loop5. Furthermore, when ⌬loop5 and wild-type Pmt4p are expressed simultaneously, additional complexes varying in size from ϳ156 to ϳ170 kDa appeared (Fig. 6B, lane  4). From these data we conclude that Pmt4p FLAG ⌬loop5 is able to interact with itself as well as with wild-type Pmt4p. When Pmt4p FLAG ⌬loop5 is expressed in a temperature-sensitive pmt1pmt4 mutant strain it does not restore the growth defect were expressed from plasmid pJK4-B1 and pVG45, respectively. Immunoprecipitation of FLAG-tagged proteins was performed as described in Fig. 5A. Blots were probed with polyclonal anti-Pmt4p antibodies.
at 35°C (Fig. 7), indicating that this large hydrophilic segment is essential for Pmt4p activity even though it does not obviously affect Pmt4p dimerization. Our data show that loop 5 domain of Pmt4p appears to behave in the same way as loop 5 of Pmt1p.
For S. cerevisiae Pmt1p, amino acid residue Arg-138, located in transmembrane domain two at the water-membrane interface, is essential for the formation of heteromeric Pmt1p-Pmt2p complexes (Ref. 18, see also Fig. 6C, lane 3). In addition, exchange of Pmt1p Arg-138 for alanine results in a complete loss of mannosyltransferase activity (Ref. 18, see also Fig. 7). When Arg-138 is exchanged for lysine, Pmt1p-Pmt2p complexes (Fig.  6C, compare lanes 1 and 4) as well as O-mannosyltransferase activity (Fig. 7) are partially restored, indicating that a positive charged amino acid at that position is important for the establishment of functional Pmt1-Pmt2p complexes. Because this arginine residue is highly conserved between all PMT family members, we asked whether this residue also affects Pmt4p's homomeric interactions. We therefore replaced Arg-142 of Pmt4p FLAG (the equivalent of Arg-138 in Pmt1p) with glutamate using site-directed mutagenesis. The Pmt4p mutant protein R142E FLAG was expressed and characterized in a yeast wild type, a pmt4 and a pmt1pmt4 mutant strain. SDS-PAGE and Western blotting of Triton extracts with polyclonal anti-Pmt4p antibodies revealed that in wild-type yeast Pmt4p FLAG and the mutant protein R142E FLAG show an identical mass and are expressed at similar levels (data not shown). However, Pmt4p-R142E FLAG failed to complement the temperature-sensitivity of the pmt1pmt4 mutant, indicating that Arg-142 is essential for Pmt4p activity in vivo (Fig. 7). In addition, coimmunoprecipitation experiments were performed on Triton extracts of a wild-type strain coexpressing wild type Pmt4p and, alternatively, Pmt4p FLAG , or Pmt4p-R142E FLAG using monoclonal anti-FLAG antibodies. In contrast to Pmt4p FLAG , which efficiently coimmunoprecipitates wild-type Pmt4p, Pmt4p-R142E FLAG almost completely fails to precipitate wild-type Pmt4p (Fig. 6D, compare lanes 1 and 2), consistent with a critical role for Arg-142 in Pmt4p-Pmt4p complex formation. Taken together, our data show that similar principles underlie the formation of heteromeric complexes between members of the PMT1 and PMT2 subfamilies and homomeric Pmt4p complexes.

DISCUSSION
In yeast the PMT1 and PMT2 subfamilies but not the PMT4 subfamily are highly redundant. In this study we demonstrate that the formation of specific protein complexes is a common feature of PMT family members in yeast. We found that, in general, members of the PMT1 subfamily (Pmt1p and Pmt5p) interact in pairs with members of the PMT2 subfamily (Pmt2p and Pmt3p). As schematically shown in Fig. 8A, Pmt1p-Pmt2p and Pmt5p-Pmt3p are the predominant complexes formed between PMT1 and PMT2 subfamily members in wild-type S. cerevisiae cells. Under certain conditions, however, Pmt1p can interact with Pmt3p, and Pmt5p with Pmt2p, respectively. This can occur, for example, when one of the principle partners is absent, as is the case in pmt mutant strains. In contrast, the unique representative of the PMT4 subfamily forms homomeric complexes (Fig. 8B). Interestingly, we further uncovered that the same conserved protein domains influence both heteromeric and homomeric Pmt-protein interactions.
Heteromeric Protein Complexes between PMT1 and PMT2 Subfamily Members Might Have Evolved by Gene Duplication and Fulfill Similar Tasks in S. cerevisiae-Within the PMT family in S. cerevisiae, Pmt1p is most closely related to Pmt5p (53% identity and 72% homology), and Pmt2p to Pmt3p (65% identity and 81% homology), respectively. In view of this high degree of conservation, the proteins Pmt1p and Pmt5p, as well as Pmt2p and Pmt3p might have evolved by gene duplication. This is supported by the fact that Pmt1p (YDL095w) and Pmt5p (YDL093w) are located directly next to each other on chromosome IV. Further, the report of Wolfe and Shields (42) suggests that Pmt2p (YAL023c; chr. I) and Pmt3p (YOR321w; chr. XV) constitute a protein pair that derives from an ancient duplication of the entire yeast genome. As a consequence of these gene duplication events, the ability to form specific protein complexes between individual members of the PMT1 and the PMT2 subfamily ses of in vitro mannosyltransfer from Dol-P-Man to specific synthetic acceptor peptides have shown that the simultaneous deletion of PMT1 and PMT2 results in the loss of Ͼ80% of in vitro O-mannosyltransferase activity, when compared with wild-type yeast (1,24). In addition, in vivo O-mannosylation of the same protein substrates is dramatically decreased in pmt1 and pmt2 mutant strains, such as the cell wall proteins Kre9p, chitinase (Cts1p), Bar1p, Ccw4p, and Ccw5p (14,25). In contrast, in vitro and in vivo O-mannosylation is not obviously affected in pmt3 or pmt5 mutants (1,14,25). However, deletion of PMT3 in a pmt1pmt2 mutant further decreases in vitro mannosyltransfer from Dol-P-Man to a synthetic Pmt1p/ Pmt2p-acceptor peptide by ϳ25%, when compared with the in vitro activity measured in pmt1pmt2 strains (1). In analyses of the O-mannosylation state of chitinase in pmt1pmt2 and pmt1pmt2pmt3 mutants, Gentzsch and Tanner (1) showed that in vivo chitinase is O-mannosylated by Pmt1p and Pmt2p as well as Pmt3p. Pmt3p, therefore, mannosylates the same proteins as Pmt1p and Pmt2p, but comes into operation mainly in the absence of Pmt1p and Pmt2p. The notion that Pmt1p-Pmt2p and Pmt5p-Pmt3p complexes fulfill similar tasks in vivo is further supported by the fact that transcription of PMT1, PMT2, PMT3, and PMT5, but not PMT4 and PMT6 is enhanced in response to cell stress conditions that cause the accumulation of misfolded proteins in the ER (43). In summary, Pmt1p-Pmt2p and Pmt5p-Pmt3p are likely to O-mannosylate the same set of substrate proteins; yet, Pmt5p-Pmt3p complexes might, for example, exhibit lower substrate affinities and, therefore, play only a minor role in wild-type yeast cells. In the absence of Pmt1p and Pmt2p, however, Pmt5p-Pmt3p might compensate for O-mannosylation deficiency in pmt mutant strains. Similar functions might be assigned to the Pmt1p-Pmt3p and Pmt5p-Pmt2p complexes, which we could detect only in small amounts in wild-type strains (Figs. 2 and 3) or under specific physiological conditions such as in pmt mutants (Fig. 4). The observed increase of in vitro O-mannosyltransferase activity in pmt1pmt3 and pmt2pmt5 mutants (Table II) substantiates that Pmt1p-Pmt3p and Pmt5p-Pmt2p act as mannosyltransferases and feature substrate specificities similar to Pmt1p-Pmt2p. Our data suggest a compensatory cooperation between PMT1/PMT2 subfamily members, which might explain why in S. cerevisiae only the simultaneous deletion of several PMT subfamily members results in a substantial decrease in O-linked oligomannose chains and finally causes cell death (1). This possibility is supported by the fact that overexpression of PMT2 rescues the growth defect of a temperaturesensitive pmt1pmt4 mutant (data not shown). Because overexpression of PMT2 results in the formation of Pmt5p-Pmt2p complexes (Fig. 4), Pmt5p-Pmt2p might at least partially compensate the lack of Pmt1p-Pmt2p. Concordantly, Gentzsch et al. (21) showed that overexpression of PMT2 causes a slight increase of in vitro O-mannosyltransferase activity in wild-type yeast and in pmt1 deletion mutants. A compensatory cooperation between the redundant members of the PMT1/PMT2 subfamily in S. cerevisiae is also supported by the observation that in the fission yeast S. pombe where only one member of each PMT subfamily is present, the deletion of the single PMT2 subfamily member is lethal. 2 The Third Member of the S. cerevisiae PMT2 Subfamily, Pmt6p, Interacts with None of the Other PMT Family Members-In S. cerevisiae Pmt6p shares an overall sequence identity of 46% with Pmt2p and of 45% with Pmt3p. In the course of our analyses, BN-PAGE, coimmunoprecipitation, and chemical cross-linking experiments showed that Pmt6p interacts neither with Pmt1-5p nor with itself (data not shown), and therefore behaves differently from all other Pmtps in yeast.
Nevertheless, BN-PAGE suggested that Pmt6p interacts with other proteins, although not with Pmtps (data not shown). To identify potential interacting partners, we analyzed whether selected components of the N-glycosylation machinery or the translocon are stably associated with Pmt6p. Again, no evidence was obtained that Wbp1p (40), Ost1p (44), Stt3p (45), or Sec61p (46) interact with Pmt6p (data not shown). Further studies are needed to elucidate the nature of these Pmt6p containing complexes.
Common Principles Underlie Heteromeric Pmt1p-Pmt2p and Homomeric Pmt4p-Pmt4p Interactions-Considering Pmtps as antifungal targets it is noteworthy that in S. cerevisiae and in C. albicans only deletion of Pmt4p in combination with PMT1/ PMT2 subfamily members causes lethality (1,47). Therefore, to eliminate protein O-mannosylation both PMT1/PMT2 and PMT4 subfamily members must be inhibited. Pmt4p differs in several respects from PMT1/PMT2 subfamily members, such as substrate specificity and conserved signature sequence motifs (9,14,18). In addition, Pmt4p forms homomeric protein complexes as demonstrated in this study (Fig. 5). In view of these variances, it is of particular importance that common principles form the basis of the formation, structure and/or stability of PMT complexes in both PMT1/PMT2 and PMT4 subfamilies. Our mutational analyses showed that Pmt4p Arg-142, which is highly conserved between all PMT family members, is crucial for Pmt4p-Pmt4p complexes and enzyme activity (Figs. 6 and 7). Analogical, the corresponding mutation in Pmt1p affects Pmt1p-Pmt2p complexes and results in loss of mannosyltransferase activity (18). Thus, PMT complexes offer a point of attack to abolish protein O-mannosylation in fungi.
PMT Complexes Might Ensure Efficient O-Mannosylation-There are a number of reasons why protein O-mannosyltransferases form homo-or heteromeric protein complexes. One is that these complexes ensure an efficient O-glycosylation of a wide range of proteins. A common feature of O-glycosylated proteins is that O-linked carbohydrate chains are clustered in distinct serine/threonine rich regions (48). Such areas are thought to adopt rod-like structures important for protein function. With a few exceptions O-mannosylation occurs while proteins are translocated in the lumen of the ER (49,11). 3 Thus, the clustering of O-linked sugars requires high efficiency sugar transfer, which might be provided by mannosyltransferase complexes. That oligomerization enhances enzyme function has been proven for other glycosyltransferases such as UDP-GlcNAc:dolichol-P GlcNAc-1-P transferase (50) or the mannosyltransferase complexes M-Pol I and M-Pol II (51)(52)(53). Furthermore, even though members of the PMT1 and PMT2 subfamily act on the same protein substrate (14,25), they might actually O-mannosylate different serine and threonine residues within one and the same protein. Since to date no specific consensus sequences are known that are required for O-mannosylation this assumption remains to be verified. However, this hypothesis is further supported by the fact that mutant ␣-factor precursor is O-mannosylated by Pmt2p but no other PMT family member (11). To understand the functioning of PMT complexes it will be important to learn more about their different substrate specificities and what features of PMTs determine specificity.