Structure-function analysis of the dolichyl phosphate-mannose: protein O-mannosyltransferase ScPmt1p.

Protein O-mannosylation is an essential protein modification. It is initiated at the endoplasmic reticulum by a family of dolichyl phosphate-mannose:protein O-mannosyltransferases (Pmts), which is evolutionarily conserved from yeast to humans. Saccharomyces cerevisiae Pmt1p is an integral membrane protein of the endoplasmic reticulum. ScPmt1p forms a complex with ScPmt2p that is required for maximum transferase activity. Recently, we proposed a seven-transmembrane structural model for ScPmt1p. A large, hydrophilic, endoplasmic reticulum-oriented segment is flanked by five amino-terminal and two carboxyl-terminal membrane-spanning domains. Based on this model, a structure-function analysis of ScPmt1p was performed. Deletion mutagenesis identified the N-terminal third of the transferase as being essential for the formation of a functional ScPmt1p-ScPmt2p complex. Deletion of the central hydrophilic loop eliminates mannosyltransferase activity, but not ScPmt1p-ScPmt2p interactions. Alignment of all fully characterized PMT family members revealed that this central loop region contains three highly conserved peptide motifs, which can be considered as signatures of the PMT family. In addition, a number of invariant amino acid residues were identified throughout the entire protein sequence. In order to evaluate the functional significance of these conserved residues site-directed mutagenesis was performed. We show that several amino acid substitutions in the conserved motifs significantly reduce ScPmt1p activity. Further, the invariant residues Arg-64, Glu-78, Arg-138, and Leu-408 are essential for ScPmt1p function. In particular, Arg-138 is crucial for ScPmt1p-ScPmt2p complex formation.

Protein O-mannosylation is of fundamental importance in numerous physiological processes. In the yeast Saccharomyces cerevisiae, O-mannosylation is an essential protein modification that is indispensable for cell morphology and cell wall integrity (1,2). Furthermore, O-linked polymannose chains are required for the stability and/or correct localization of proteins (3,4) and also affect protein function (5,6). In the opportunistic pathogenic fungus Candida albicans, reduced protein O-mannosylation results in defects in morphogenesis, abated adherence to host cells, and a strong attenuation of virulence in mouse model systems (7,9). 1 In yeasts and filamentous fungi, O-linked mannooligosaccharide synthesis is initiated in the endoplasmic reticulum (ER) 2 by the transfer of a mannosyl residue from dolichyl phosphate-activated mannose (Dol-P-Man) to specific serine/ threonine residues of proteins entering the secretory pathway (2). Further modification of the O-linked mannose takes place in the Golgi apparatus (2,10).
Dol-P-Man:protein O-mannosyltransferases (Pmts) comprise a large family of enzymes that initiate protein O-mannosylation at the ER. In Saccharomyces cerevisiae, the seven members of this protein family (ScPmt1p to ScPmt7p) feature an overall sequence similarity of 50 -55% and, more strikingly, a nearly identical hydropathy profile wherein an integral membrane protein with multiple transmembrane domains is predicted (11)(12)(13)(14). Pmtp orthologues have been identified from the yeast Schizosaccharomyces pombe (S. pombe genome sequencing project) and from the human pathogen C. albicans, and both exhibit in vivo as well as in vitro mannosyltransferase activity (7). 3 Moreover, a PMT homologous gene, rotated abdomen (rt) from Drosophila melanogaster has been described (16). The poorly viable homozygous rt mutant flies show strong defects in embryonic muscle development, indicating an essential role of the PMT family members not only in lower but also in higher eucaryotes. Very recently, a PMT homologue from humans (POMT1) has been cloned (17). Interspecies comparison revealed that POMT1-homologous sequences are present in mammals, birds, reptiles, and amphibians, affirming that the PMT gene family is evolutionarily conserved from yeast to humans. However, despite their evolutionary conservation and functional importance, still little is known about the biochemical and biophysical properties of these key enzymes of protein O-mannosylation.
Of the PMT family members, those from S. cerevisiae, especially ScPMT1, have been studied most thoroughly. ScPmt1p is an integral membrane glycoprotein of 817 amino acids located at the endoplasmic reticulum (18 -20). ScPmt1p interacts with ScPmt2p in vivo, and the formation of this complex is required for maximum transferase activity (21). ScPmt1p-catalyzed transfer of mannose from Dol-P-␤-D-mannose (with inversion of its anomeric configuration) to serine/threonine residues of specific protein acceptors can be assayed both in vitro and in vivo (18,(22)(23)(24). Thereby, a strict stereospecificity for Dol-P-␤-D-mannose and the recognition of the ␣-isoprene unit of the dolichyl moiety has been observed (25,26). In vitro studies have shown that ScPmt1p glycosylates a restricted spectrum of peptides (18,27,28). Concordantly, in vivo studies demonstrated that the individual mannosyltransferases ScPmt1p to ScPmt6p recognize specific protein substrates (29). However, the structural basis for the substrate specificities is unknown.
Recently, we have analyzed the transmembrane topology of S. cerevisiae Pmt1p to establish a basis for the identification and characterization of structurally and functionally important protein domains. Both genetic and biochemical evidence supports a seven-transmembrane helical model (30) wherein a large, central, hydrophilic segment that is oriented toward the lumen of the ER is flanked by five amino-terminal and two carboxyl-terminal membrane-spanning domains. The ScPmt1p amino terminus faces the cytoplasm, while the carboxyl terminus faces the lumen of the ER. In the current study, a structure-function analysis of ScPmt1p was performed. Using deletion mutagenesis, we demonstrate that the N-terminal third of the protein is essential for ScPmt1p-ScPmt2p complex formation. Moreover, the central hydrophilic region does not participate in ScPmt1p-ScPmt2p interaction but rather bears part of the catalytic unit. Inside this region, we recognized highly conserved sequence motifs that can be considered as hallmarks of the PMT family. In addition, numerous, perfectly conserved, single amino acid residues were found throughout the entire sequence of ScPmt1p. Using site-directed mutagenesis, we demonstrate the functional importance of these highly conserved residues for ScPmt1p mannosyltransferase activity.
Standard procedures were used for all DNA manipulations (33). All cloning and transformations were carried out in Escherichia coli hosts SURE ® 2 or ABLE TM K (Stratagene). PCR fragments were routinely checked by sequence analysis.
Plasmids pSB63 (⌬732-817) and pSB64 (⌬617-817)-Six copies of the HA epitope and the first 430 bp of the 3Ј-untranslated region of PMT1 were amplified by PCR on pSB56 with the primer pair oligo133 and oligo142a (5Ј-agtagtcgaccgcggccgcatcttttacccatac-3Ј; SalI site in boldface type and NotI site underlined). The PCR fragment was digested with SalI and SphI and subcloned into pC731 or pP616 (30) cut with XhoI and SphI, resulting in pSB63 and pSB64, respectively.
Plasmid pVG13 (⌬304 -531)-The plasmid pSB53 was cut with SphI and NarI, treated with DNA polymerase I (Klenow fragment), and religated to remove a HindIII site from the multiple cloning site. Subsequently, an internal 683-bp HindIII fragment was removed, resulting in the deletion of the PMT1 coding region from bp ϩ916 to bp ϩ1599. The created plasmid (pSB79) was digested with RsrII and XhoI and ligated with a 432-bp RsrII-XhoI fragment isolated from pSB56 fusing six copies of the HA epitope to the C-terminal region of the deletion construct.

Computer Analyses
Multiple sequence alignments of PMT family members were prepared with ClustalW (36).

Preparation of Cell Walls and Crude Membranes
Yeast cells were grown in synthetic complete medium (37). At a concentration of 2.0 ϫ 10 7 cells/ml, 50 ml of cells were harvested; washed with 20 ml of 50 mM Tris-HCl, pH 7.5, 0.3 mM MgCl 2 ; and resuspended in 200 l of the same buffer plus 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 0.25 mM TLCK, 50 g/ml TPCK, 10 g/ml antipain, 1 g/ml leupeptin, and 1 g/ml pepstatin. An equal volume of glass beads was added, and the cells were lysed by vortexing, for 1 min, four times (with 1-min intervals on ice). The bottom of the tube was punctured, and the lysate was collected. Cell debris was removed by centrifugation for 5 min at 3000 rpm at 4°C. The pellet (cell walls) was stored at Ϫ20°C. Crude membranes were collected from the supernatant by centrifugation for 30 min at 20,000 rpm at 4°C (Sorvall SS34 rotor); resuspended in 250 l of 50 mM Tris-HCl, pH 7.5, 7.5 mM MgCl 2 , 15% glycerol; and stored in liquid nitrogen.

Subcellular Fractionation by Velocity Sedimentation on Sucrose Density Gradients
Yeast cells were grown to 2-4 ϫ 10 7 cells/ml in synthetic complete medium. A total of 10 10 cells were harvested; washed with H 2 O; resuspended in 10 ml of 100 mM Tris-HCl, pH 9.4, 10 mM dithiothreitol; and incubated for 15 min at 30°C under moderate shaking. Cells were washed with 20 ml of spheroplasting buffer (20 mM Tris-HCl, pH 7.5, 1.2 M sorbitol), resuspended in 10 ml of the same buffer plus 8 mg of Zymolyase (103,000 units; Seikagaku Corp.), and digested for 1.5 h at 30°C under moderate shaking. Cells were washed in 10 ml of spheroplasting buffer and resuspended in 1 ml of lysis buffer (20 mM Tris-HCl, pH 7.5, 1.2 M sorbitol, 1 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride, 50 g/ml TPCK, 0.25 mM TLCK, 1 mM benzamidine, 20 g/ml antipain, 1 g/ml leupeptin, 1 g/ml pepstatin). Cell lysis was carried out on ice in a 5-ml Dounce homogenizer by homogenizing the spheroplast suspension 8 -10 times. The cell lysate was cleared three times for 5 min at 2,400 rpm. The supernatant was loaded onto an 11-step sucrose gradient (18 -60%). The subcellular fractionation was carried out according to Schröder et al. (38) with the following modification. The gradients were centrifuged for 2.5 h in a TST41.14 rotor (Kontron Instruments) at 35,700 rpm. 30-l aliquots from each gradient fraction were analyzed by Western blot.

Deglycosylation by Endoglycosidase F/N-Glycosidase F Digestion
A total of 2.5 l of crude membranes were suspended in 20 l of PNGase buffer (50 mM potassium phosphate buffer, pH 5.5, 35 mM EDTA, 0.02% SDS, 1% 2-mercaptoethanol, protease inhibitors as above) and digested with 0.5-1 unit/l of endoglycosidase F/N-glycosidase F (Roche Molecular Biochemicals) for 1 h at 30°C. Mock samples were incubated without enzyme. The reaction was stopped by adding 10 l of 5ϫ SDS-sample buffer.

In Vitro Dol-P-Man:Protein O-Mannosyltransferase Assay
15-50 g of membrane protein (see above) were incubated in the in vitro assay for Dol-P-Man:protein O-mannosyltransferase Pmt1p, as described previously (18). The pentapeptide acetyl-YATAV-NH 2 was used at a final concentration of 3.5 mM.

Isolation of Chitinase and Heat Shock Protein Hsp150p
Chitinase (Cts1p) was isolated from cell walls (see above), and Hsp150p was isolated from the medium as described by Gentzsch and Tanner (29).

Immunoprecipitation of Pmt1p HA
Crude membranes were isolated as described above and resuspended in 400 l of lysis buffer (20 mM Tris-HCl, pH 7.5, 140 mM NaCl, 0.3 mM MgCl 2 , 10% glycerol, 0.35% sodium deoxycholate, 0.5% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 0.25 mM TLCK, 50 g/ml TPCK, 10 g/ml antipain, 1 g/ml leupeptin, and 1 g/ml pepstatin). Proteins were solubilized by vortexing on an Eppendorf shaker 5432 for 30 min at 4°C. The suspension was clarified by centrifugation for 30 min at 20,000 rpm (Sorvall SS34 rotor) to yield the DOC extract. Pmt1p HA and Pmt1p HA mutant proteins were immunoprecipitated from 300 l of DOC extract with 10 l of anti-HA monoclonal antibody covalently coupled to protein A-Sepharose (16B12; Babco) for 1-2 h at 4°C. The immunoprecipitates were washed four times with 1 ml of lysis buffer and once with 1 ml of Tris-buffered saline. Subsequently, the precipitates were resuspended in 20 l of 3ϫ SDSsample buffer.

Deletion Mutants of Yeast Pmt1p HA Affect Protein O-Mannosyltransferase Activity-The protein O-mannosyltransferase
Pmt1p from S. cerevisiae is an integral ER membrane protein of 817 amino acids that is modified by three N-linked carbohydrate chains (11,18). In a recent study, we presented a seventransmembrane helical model of the topology of Pmt1p (30). The Pmt1p amino terminus is located in the cytoplasm, the carboxyl terminus in the ER lumen. A large hydrophilic region, facing the lumen of the ER, is separated from the N terminus by five, and from the C terminus by two, membrane-spanning domains (Fig. 1). The functions of the various regions of Pmt1p are unknown. To identify domains important for Pmt1p mannosyltransferase activity, we created a series of deletion mutants (⌬732-817 to ⌬76 -124; Table I; schematically shown in Fig. 2). For immunological detection, six copies of the HA epitope (45) were fused to the C terminus of Pmt1p and the individual mutants. The C-terminal HA epitope does not interfere with mannosyltransferase activity, since wild type Pmt1p and epitope-tagged Pmt1p HA have identical in vitro as well as in vivo mannosyltransferase activities when expressed in the strain pmt1⌬ (12) in which the endogenous PMT1 gene is deleted (data not shown).
As described under "Experimental Procedures," we constructed two C-terminal deletion mutants, ⌬732-817 and ⌬617-817, lacking the C-terminal 86 and 201 amino acids (aa) respectively, including the N-glycosylation sequon N743QT (Table I; Figs. 1 and 2). In addition, the internal deletion mutants ⌬76 -124, ⌬161-211, ⌬203-259, and ⌬304 -531 were created (Table I;  In order to test whether the deleted segments are crucial for mannosyltransferase activity, the various deletion mutants were assessed for their ability to complement defects of in vitro and in vivo mannosyltransferase activity in the yeast pmt1 deletion strain pmt1⌬. In vitro mannosyltransferase activity of membranes was measured by the transfer of [ 14 C]mannose from Dol-P-[ 14 C]Man to the pentapeptide Ac-YATAV-NH 2 (18,24). Under the conditions we used, this assay exclusively records the activity of  1 and 10 with lanes 3 and  9). Even more dramatically, the absence of Pmt1p HA causes a change in the molecular mass of Hsp150p from 150 kDa (Fig.  3B, lane 1) to a series of smaller fragments with a maximum peak of ϳ60 -66 kDa (Fig. 3B, lane 2) due to the lack of O-linked sugars.
Using the in vitro and the in vivo system, we checked the functionality of the Pmt1p HA mutants. Deletion of the C-termi-nal 86 aa (⌬732-817) results in a small but reproducible increase of in vitro activity when compared with Pmt1 HA (Table  II). In vivo, however, no significant differences in the glycosylation of Cts1p and Hsp150 could be detected between strains expressing Pmt1 HA or ⌬732-817 (Fig. 3, A, lane 2, and B, lane  3). In contrast, all other deletions failed to restore Cts1p as well as Hsp150p to its normal glycosylation levels (Fig. 3, A and B,  lanes 4 -8). Concordantly, in vitro mannosyltransferase activity of ⌬617-817 and ⌬304 -531 is diminished by more than 97 and 93%, respectively (Table II). No significant in vitro activity is found for ⌬203-259, ⌬161-211, and ⌬76 -124 (Table II).   . 3. In vivo mannosyltransferase activity of the Pmt1p HA deletion mutants. PMT1 HA and the deletion mutants ⌬732-817 to ⌬76 -124 were individually expressed from the multicopy plasmid YEp352 in the S. cerevisiae strain pmt1⌬. Chitinase (A) and Hsp150p (B) were isolated from cell walls and culture medium, respectively, as described by Gentzsch and Tanner (29). Proteins were resolved on 8% SDS-polyacrylamide gels and analyzed by Western blot using anti-Cts1p or anti-Hsp150p polyclonal antibodies.
The detected differences in mannosyltransferase activity could be due to alterations in expression, stability, localization, or transmembrane topology of the mutant proteins. To rule out changes in expression levels and/or protein stability, all deletion mutants were analyzed by Western blot in a pmt1 mutant background. When compared with Pmt1p HA , very similar amounts for all deletion constructs could be detected with anti-HA monoclonal antibodies (Fig. 4A).
Next, we made sure that the mutant proteins were still localized at the ER. To prove this intracellular localization, cell fractionations were performed. Fig. 4B shows that the individual mutant proteins colocalize with the ER fractions identified by the marker proteins Pmt1p, Pmt2p, and Wbp1, a subunit of the oligosaccharyl transferase complex (41). On the velocity sucrose gradient, the ER (Fig. 4B, lanes 8 -12) clearly separates from the Golgi (Fig. 4B, lanes 3-7) and vacuolar (Fig. 4B, lanes  1 and 2) fractions, identified by the cis-Golgi ␣1,6-mannosyltransferase Och1p (47) To exclude the possibility that the deletions were causing major changes in protein topology, we examined whether the central hydrophilic loop 5 and the C terminus of the Pmt1p HA mutants were still facing the ER lumen by evaluating their glycosylation status. Full-length Pmt1p HA contains N-glycosylation sites at positions Asn-390, Asn-513, and Asn-743. Asn-390 and Asn-513 are located in the central loop 5, and Asn-743 is close to the C terminus (Fig. 1). Treatment of Pmt1p HA with endoglycosidase F/N-glycosidase F (Endo F) reduces the apparent molecular mass of Pmt1p HA from 98 to ϳ93 kDa (Fig. 4C,  lanes 1 and 2), indicating that all three N-glycosylation sites bear N-linked core carbohydrate chains as previously demonstrated for wild type Pmt1p (30). Since N-glycosylation takes place only in the ER lumen, these data confirm the ER-luminal orientation of the loop 5 region and the C terminus (Fig. 1). The Pmt1p HA mutant proteins were then analyzed in like manner. On SDS-polyacrylamide gel electrophoresis, ⌬732-817 migrates as a doublet with apparent masses of ϳ72 kDa (major species) and ϳ76 kDa (minor species) (Fig. 4C, lane 3). After Endo F treatment, a single band of ϳ68 kDa, with anomalous mobility due to the hydrophobic nature of the protein, is detected (Fig. 4C, lane 4). These data show that both N-glycosylation sites (Asn-390 and Asn-513) present in ⌬732-817 bear an N-linked core carbohydrate chain, thus confirming that the central loop 5 is ER-oriented. Very similar results were obtained for ⌬617-817 (Fig. 4C, lanes 5 and 6). ⌬304 -531 migrates as a band with a mobility of 59 kDa (Fig. 4C, lane 7). The reduced mass of ϳ57 kDa after deglycosylation (Fig. 4C, lane 8) corroborates the occupancy of the only N-glycosylation site in ⌬304 -531 (Asn-743) and, therefore, the ER-luminal orientation of C-terminal region. In addition to the mature form of ⌬203-259 with an apparent molecular mass of ϳ96 kDa, two underglycosylated forms are detected (Fig. 4C, lane 9). Endo F decreases the molecular mass to ϳ88.5 kDa (Fig. 4C, lane 10), demonstrating the presence of three core N-linked sugar chains. The same is true for ⌬161-211 and ⌬76 -124 (Fig. 4C,  lanes 11-14), confirming that loop 5 and the C termini of these constructs are facing the ER lumen.
In summary, we found that (i) Pmt1p HA and the various mutant proteins are expressed at similar levels, (ii) the deletions do not affect protein stability, (iii) the mutant proteins localize to the ER, and (iv) while minor local changes cannot be ruled out, the deletions do not cause major changes in Pmt1p topology. From our data, we conclude that the regions spanning the residues Phe-76 to Ile-124, Val-161 to Ala-211, Phe-203 to Leu-259, Ser-304 to Ala-531, and Ile-617 to Cys-731 but not the C-terminal 86 aa are crucial for Pmt1p function.

The Large Luminal Loop 5 Region Is Essential for Mannosyltransferase Activity but Not for Pmt1p-Pmt2p
Complex Formation-With the exception of ⌬732-817, all of the deletion mutants analyzed failed to catalyze the transfer of mannose. It has been previously shown that S. cerevisiae Pmt1p and Pmt2p interact and that the formation of this complex is essential for maximal mannosyltransfer (21). These data suggest the possibility that the Pmt1p HA mutant proteins are defective in forming a functional complex with Pmt2p. To test this assumption, protein-protein interactions were analyzed using coimmunoprecipitation. An anti-HA monoclonal antibody was used to immunoprecipitate Pmt1p HA and the deletion mutant proteins from membranes solubilized with the detergents sodium deoxycholate and Triton X-100 (see "Experimental Procedures"). The immunoprecipitates obtained were analyzed by Western blots sequentially probed with polyclonal anti-Pmt2p and monoclonal anti-HA antibodies. Fig. 5 shows that Pmt1p HA coimmunoprecipitates Pmt2p, substantiating the results reported by Gentzsch et al. (21). Under the conditions we used, the coimmunoprecipitation of Pmt1p HA and Pmt2p is specific, since we could not detect Pmt4p, Pmt6p, or Wbp1p in the immunoprecipitate (data not shown).
In accordance with its functionality, the C-terminal deletion construct ⌬732-817 still forms a complex with Pmt2p (Fig. 5,  lane 3). Interestingly, in several independent experiments, the amount of Pmt2p coprecipitating with ⌬732-817 was slightly increased when compared with wild type Pmt1p HA . Further truncation of the C terminus (⌬617-817) appears to disrupt the Pmt1p-Pmt2p interaction (Fig. 5, lane 4); however, a very weak Pmt2p signal can be detected after coimmunoprecipitation with ⌬617-817. 4 The internal deletion mutants ⌬203-259, ⌬161-211, and ⌬76 -124 (Fig. 5, lanes 6 -8) did not coimmunoprecipitate Pmt2p under all of the conditions we tested. In contrast, the mutant protein ⌬304 -531 lacking the large ER-oriented loop 5 (Fig. 1) shows no detectable in vivo mannosyltransferase activity but still coimmunoprecipitates Pmt2p (Fig. 5, lane 5). Quantification of the Western signals indicated that the degree of association is 50 -60% of the wild type (data not shown). To exclude the unlikely possibility that loss of activity is due to the partial reduced affinity of ⌬304 -351 for Pmt2p, we replaced amino acids 304 -351 with the homologous region of yeast Pmt4p (aa 313-539). Pmt4p has a different substrate specificity compared with Pmt1p or Pmt2p (29) and could not be detected in the Pmt1p-Pmt2p complex (data not shown).

Structure-Function Analysis of ScPmt1p
Whereas substitution of loop 5 completely restored Pmt1p-Pmt2p complex formation, mannosyltransferase activity was not recovered (data not shown). Gentzsch et al. (21) suggested that Pmt1p-Pmt2p complexes might be heterodimers; however, their data did not rule out the formation of heteromultimers. Thus, mutant ⌬304 -531 might affect monotypic Pmt1p-Pmt1p interactions essential for mannosyltransferase activity. We excluded monotypic contacts by expressing epitope-tagged Pmt1p HA in a wild type background where the endogenous copy of Pmt1p is present. Immunoprecipitation of Pmt1p HA did not bring down wild type Pmt1p (data not shown). From our data, we conclude that most of the N-terminal third of Pmt1p (aa Phe-76 to Ile-124, Val-161 to Ala-211, and Phe-203 to Leu-259) is essential for the formation of a functional Pmt1p-Pmt2p complex in vivo. In contrast, deletion of the region between Ser-304 and Ala-531 still allows Pmt1p-Pmt2p interaction, but this complex is enzymatically inactive.
Determination of Critical Amino Acids by Site-directed Mutagenesis-To test the role of the conserved features of the PMT family in enzyme function, we used site-directed mutagenesis to replace conserved residues of Pmt1p HA with alanine. The   FIG. 4. Levels of protein expression, ER localization, and N-glycosylation of the Pmt1p HA deletion mutants. Cell lysates or crude membranes were isolated from the yeast strains pmt1⌬/pSB56 (Pmt1 HA ), pmt1⌬/pSB63 (⌬732-817), pmt1⌬/pSB64 (⌬617-817), pmt1⌬/pVG13 (⌬304 -531), pmt1⌬/pVG11 (⌬203-259), pmt1⌬/pVG9 (⌬161-211), and pmt1⌬/pSB101 (⌬76 -124). A, membrane proteins (25 g) were resolved on an 8% SDS-polyacrylamide gel, and tagged species were detected with the monoclonal anti-HA antibody 16B12. B, subcellular fractionation by velocity sedimentation on sucrose density gradients was performed as described under "Experimental Procedures." 30-l samples from each fraction were resolved on 8% SDS-polyacrylamide gels. Western blots were sequentially probed with polyclonal antibodies directed against the vacuolar marker protein Cpy1p, the late Golgi marker Kex2p, and the ER marker proteins Wbp1p, Pmt1p, and Pmt2p. Monoclonal anti-HA antibody was used to detect the HA-tagged cis-Golgi marker Och1p (gift from L. Lehle) and the Pmt1 HA mutant proteins. C, 25 g of membrane proteins were treated with Endo F or mock-treated, as indicated. Proteins were resolved on 8% SDS-polyacrylamide gels and analyzed by Western blot using a monoclonal anti-HA antibody. Immunoprecipitates were analyzed by Western blots sequentially probed with polyclonal anti-Pmt2p and monoclonal anti-HA antibody. Western blots of the input material prior to immunoprecipitation made sure that equal amounts of Pmt1p HA mutant proteins and/or Pmt2p were present in the individual extracts (data not shown). mutant proteins were expressed and characterized in the yeast strain pmt1⌬. Endo F treatment and Western blotting of wild type Pmt1p HA and all mutant proteins described here revealed bands of comparable intensity, molecular mass, and identical extent of glycosylation (Fig. 8A).
In order to determine the significance of the conserved motifs A-C in region loop 5, we made single or double substitutions of invariant or highly conserved amino acids. The fact that a histidine residue has been implicated in the active site of inverting glycosyltransferases (49) and that the sequence LHSH is partially conserved in motif A and motif B suggests that these residues are particularly relevant. Thus, mutants H346A/H348A (motif A) and L408A/H411A (motif B) were constructed. In addition, we mutated Q359A/Q360A and N370A (motif A), R398A and L399A (motif B), Q493A/E495A (motif C), R469A, and H472A. As summarized in Fig. 7, the majority of the single and multiple substitutions in loop 5 yielded enzymes with activities roughly 30 -40% of normal. This indicates, on the one hand, that the conserved motifs are indeed crucial for enzyme function but, on the other hand, that the enzyme can to some extent tolerate mutations in these domains. Unlike most substitutions, mutant L408A/H411A resulted in almost complete loss of transferase activity. Surprisingly, when individually exchanged Leu-408, but not His-411, proved essential for transferase activity, arguing for a more structural function rather than direct involvement in catalysis.
Since, in loop 5, only mutation of Leu-408 resulted in loss of Pmt1p activity, we wanted to identify additional residues essential for Pmt1p function. Therefore, in the N-terminal third of the protein the 100% conserved polar amino acids Arg-64 (TM I), Arg-138 (TM II), Glu-78, Asp-96 (loop 1), and Lys-234 (loop 4) were changed to alanine. Fig. 7 shows that despite its strict conservation, mutation of Lys-234 only slightly affects transferase activity. Also, mutant D96A still has about one-half of wild type Pmt1p activity. Mutagenesis of Arg-64, Glu-78, and Arg-138, however, caused a severe loss of transferase activity. This is not due to poor expression, stability, or changes in general topology of the mutant proteins (Fig. 8A). Since deletion analysis defined the N-terminal third of Pmt1p to be essential for Pmt1p-Pmt2p interactions, Arg-64, Glu-78, or Arg-138 might affect these protein-protein contacts. Thus, the nonfunctional mutant proteins were tested for Pmt1p-Pmt2p complex formation. All the mutant proteins, except for R138A, coimmunoprecipitate Pmt2p ( Fig. 8B and data not shown). From these data, we conclude that Arg-64 and Glu-78 of region loop 1 are functionally important for Pmt1p activity; in contrast, Arg-138 (TM II) is crucial for Pmt1p-Pmt2p complex formation. DISCUSSION The PMT enzymes constitute an essential family of protein O-mannosyltransferases that are highly conserved from yeast to humans. In this study, we present the first structure-function analysis of a PMT family member, S. cerevisiae Pmt1p. Our data demonstrate that the N-terminal one-third of Pmt1p, in particular residue Arg-138, is crucial for Pmt1p-Pmt2p dimerization. Alignment of all characterized PMT sequences revealed highly conserved sequence motifs and invariant residues throughout the Pmt1p protein. Among them, Arg-64, Glu-78, and Leu-408 are essential for transferase activity, arguing for the involvement of the ER-oriented loop 1 and loop 5 regions in catalysis and/or substrate recognition.
Recently, we presented a seven-transmembrane helical model for yeast Pmt1p ( Fig. 1; Ref. 30). The ER-oriented, hydrophilic region loop 5 is separated from the N terminus by five, and from the C terminus by two, membrane-spanning domains. Deletion analysis now verifies and advances this model. Mutant ⌬161-211 proves the presence of two TM spans between amino acid residues 161 and 211 that were not precisely defined. The deletion of one membrane span would result in inversion of the topology of the subsequent parts of the protein and leave loop 5 and the C terminus in the cytoplasm. However, ⌬161-211 did not affect the ER-luminal orientation of these domains (Fig. 4C). Further, the cytoplasmic loop 4 segment contains a highly hydrophobic region (aa 235-251) that despite its hydrophobicity was shown to not span the membrane (30). Deletion of this region (⌬203-259) did not cause inversion of the ER-luminal orientation of loop 5, corroborating that it does not participate in the specification of the topology of Pmt1p (Fig. 4C).
Using coimmunoprecipitation, we demonstrated that Pmt1p and Pmt2p form a specific heterodimeric complex in vivo, completing the data of Gentzsch et al. (21). The formation of this complex is essential for maximal mannosyltransferase activity, since in both pmt1 and pmt2 deletion strains, in vitro transferase activity is more than 90% decreased when compared with wild type (Ref. 21 and data not shown). Deletion analysis revealed that the N-terminal one-third of Pmt1p is involved in Pmt1p-Pmt2p interactions. In particular, Arg-138, which is located close to or in TM II is essential for Pmt1p-Pmt2p complex formation (Fig. 8B). Arg-138 is 100% conserved in all PMTs, and therefore it is unlikely that this residue directly mediates specific Pmt1p-Pmt2p contacts. Mutant R138A does not cause major changes in Pmt1p topology (Fig. 8A) and still localizes to the ER (data not shown). Thus, the simplest interpretation of our data is that Arg-138 preserves a local arrangement of the N-terminal region essential for mannosyltransferase activity. This specific structure might make Pmt1p-Pmt2p interactions possible, and its destruction would cause the breakup of these complexes.
Deletion of loop 5 that is facing the ER lumen, resulted in complete loss of Pmt1p activity in vivo. However, the in vitro mannosyltransferase activity detected was 7% of wild type levels. It is highly likely that the residual activity is due to Pmt2p. The nonfunctional Pmt1p mutant ⌬304 -531 still interacts with Pmt2p, and this might cause stimulation of Pmt2p activity. This explanation was corroborated by the finding that no significant in vitro activity could be detected when ⌬304 -531 was expressed in a pmt1pmt2 deletion strain (data not shown).
S. cerevisiae Pmt1p differs from ScPmt2-7p in that its Cterminal region is extended by approximately 80 amino acids (2). Therefore, we were tempted to speculate that the C-terminal extension might play a specific role for Pmt1p function. Interestingly, deletion of the 86 C-terminal amino acids resulted in a small but significant increase of in vitro mannosyltransferase activity. In addition, we observed a reproducible increase in the amount of Pmt2p coimmunoprecipitating with mutant ⌬732-817 when compared with Pmt1p HA (Fig. 5). From these data, we conclude that the C-terminal 86 amino acids are not essential for Pmt1p mannosyltransferase activity per se. However, the C-terminal segment might be involved in regulating the extent of Pmt1p-Pmt2p complex formation and thereby allow the cell to rapidly raise the level of protein O-mannosyltransferase activity when necessary.
The formation of multimeric or oligomeric complexes between enzymes with multiple transmembrane domains is not common. One example is the UDP-GlcNAc:Dolichol-P GlcNAc- with Endo F and resolved on 8% SDS-polyacrylamide gels. Epitopetagged species were detected by Western with monoclonal anti-HA antibody 16B12. B, mutant proteins were immunoprecipitated from DOC extracts using anti-HA protein A-Sepharose beads. Immunoprecipitates were analyzed by Western blots probed with polyclonal anti-Pmt2p and monoclonal anti-HA antibody.
1-P transferase. This ER glycosyltransferase, with potentially 10 transmembrane spans, forms oligomeric associations that influence the activity and structure of individual subunits (50). Why would the formation of complexes be essential for protein O-mannosyltransferase activity? One can imagine several scenarios. First, by analogy to some channel proteins Pmt1p and Pmt2p might form a physical pore necessary for activity. Such a structure could be involved in the flipping of the sugar donor Dol-P-Man across the ER membrane, as suggested by Burda and Aebi (51) for the Dol-P-glucose-utilizing enzymes Alg8p and Alg10p. Second, O-linked carbohydrate chains are often clustered in distinct serine/threonine-rich regions of the polypeptide, which are thought to adopt rodlike structures important for protein function (52). Since protein O-mannosylation occurs cotranslationally (53), clustering of O-linked sugars requires high efficiency sugar transfer, which might be provided by mannosyltransferase complexes. Supporting this model, Munro and co-workers (8,15) identified the Golgi glycosyltransferase complexes M-Pol I and M-Pol II that catalyze the elongation of the ␣1,6-linked mannose backbone of N-linked carbohydrate chains. The M-Pol II complex contains two ␣1,6-mannosyltransferases (Mnn10p and Mnn11p), which are thought to facilitate the rapid synthesis of a long ␣1,6linked backbone.
Comparison of PMTs from different organisms defined highly conserved motifs present in the essential loop 5 that is oriented toward the ER lumen. The replacement of perfectly conserved residues in this region (e.g. the simultaneous substitution of His-346 and His-348 or Gln-493 and Glu-495) is tolerated to some extent (Fig. 7). However, mutation L408A results in an almost complete loss of transferase activity. These data suggest that a defined structure of loop 5 is essential for Pmt1p activity. In accordance, secondary structure predictions of loop 5 indicate that the sequence is arranged in a series of alternating ␣-helices and ␤-strands (data not shown). One possible function of this structured domain could be the binding of protein substrates, probably in co-operation with Pmt2p. Since Pmt1p needs to recognize many diverse protein substrates, multiple interactions might facilitate substrate binding and, in addition, define specificity. Therefore, single amino acid substitutions in this region would not completely abolish substrate binding. This hypothesis is supported by the fact that despite their conservation, motifs A-C show striking differences between the PMT subfamilies. Since PMT1 and PMT4 subfamilies use distinct acceptor protein substrates (29), these sequences might provide protein substrate specificity.
Site-directed mutagenesis demonstrated that Arg-64 and Arg-138, both of which are at or are close to the membranewater interface, and Glu-78, which is located in loop region 1 (Fig. 1), are essential for transferase activity. Since a general acid catalytic group and a general base have been discussed in terms of the mechanism of inverting glycosyltransferases (49), including PMTs, these residues might be involved in catalyzes. However, further studies are needed to clarify the role of the essential amino acid residues for Pmt1p activity.
In summary, our data suggest that TM I-V come together to bring the loop 1 and loop 5 domains of Pmt1p into close proximity to form a functional catalytic unit capable of transferring mannose from Dol-P-Man to protein substrates at the luminal side of the ER. Pmt1p-Pmt2p complex formation further stimulates transferase activity. Future work will be necessary to clarify whether the features identified for S. cerevisiae Pmt1p are of general validity for other members of the PMT family.