Functional characterization of the YUR1, KTR1, and KTR2 genes as members of the yeast KRE2/MNT1 mannosyltransferase gene family.

Eukaryotic glycan structures are progressively elaborated in the secretory pathway. Following the addition of a core N-linked carbohydrate in the endoplasmic reticulum, glycoproteins move to the Golgi complex where the elongation of O-linked sugar chains and processing of complex N-linked oligosaccharide structures take place. In order to better define how such post-translational modifications occur, we have been studying a yeast gene family in which at least one member, KRE2/MNT1, is involved in protein glycosylation. The family currently contains five other members: YUR1, KTR1, KTR2, KTR3 and KTR4 (Mallet, L., Bussereau, F., and Jacquet, M. (1994) Yeast 10, 819-831). All encode putative type II membrane proteins with a short cytoplasmic N terminus, a membrane-spanning region, and a highly conserved catalytic lumenal domain. Kre2p/Mnt1p is a alpha 1,2-mannosyltransferase involved in O- and N-linked glycosylation (Häusler, A., Ballou, L., Ballou, C.E., and Robbins, P.W. (1992) Proc. Natl. Acad. Sci. U.S.A. 89, 6846-6850); however, the role of the other proteins has not yet been established. We have carried out a functional analysis of Ktr1p, Ktr2p, and Yur1p. By in vitro assays, Ktr1p, Ktr2p, and Yur1p have been shown to be mannosyltransferase but, in vivo, do not appear to be involved in O-glycosylation. Examination of the electrophoretic mobility of the N-linked modified protein invertase in null mutant strains indicates that Ktr1p, Ktr2p, and Yur1p are involved in N-linked glycosylation, possibly as redundant enzymes. As found with Kre2p (Hill, K., Boone, C., Goebl, M., Puccia, R., Sdicu, A.-M., and Bussey, H. (1992) Genetics 130, 273-283), Ktr1p, Ktr2p, and Yur1p also seem to be implicated in the glycosylation of cell wall mannoproteins, since yeast cells containing different gene disruptions become K1 killer toxin-resistant. Immunofluorescence microscopy reveals that like Kre2p; Ktr1p, Ktr2p and Yur1p are localized in the Golgi complex.

The covalent addition of glycans to secretory and membrane proteins constitutes one of the major post-translational modifications known to occur in eukaryotes. The biosynthetic path-way leading to N-glycosylation has been studied in considerable detail and involves the ordered assembly of a core oligosaccharide on the lipid carrier dolichol phosphate, which is embedded in the ER 1 membrane. Once this oligosaccharide has been completed, it is transferred onto specific asparagine residues of proteins and subsequently altered by specific glycosidases and glycosyltransferases. The elaboration and initial processing of N-linked oligosaccharides in the ER are similar in all eukaryotes, but subsequent phases of glycosylation are different in a broad range of organisms (Tanner and Lehle, 1987;Herscovics and Orlean, 1993;Knauer and Lehle, 1994;Lehle and Tanner, 1995).
In the yeast Saccharomyces cerevisiae, the N-linked core oligosaccharide is mainly constituted of Man 8 GlcNAc 2 and may undergo Golgi maturation resulting in Man 9 -13 GlcNAc 2 . In other cases, glycoproteins traversing the Golgi have their core oligosaccharide extended by outer chains containing up to 200 mannose residues (Ballou, 1990;Herscovics and Orlean, 1993;Lehle and Tanner, 1995). Protein N-glycosylation appears essential for cell function since mutants of S. cerevisiae lacking protein subunits of the core oligosaccharyltransferase or mutants defective in the synthesis of the dolichol pyrophosphateoligosaccharyl precursor are not viable (Huffaker and Robbins, 1982;te Heesen et al., 1992te Heesen et al., , 1993Stagljar et al., 1994;Kelleher and Gilmore, 1994), although the biochemical basis of this lethality remains unclear (Tanner and Lehle, 1987;Lehle and Tanner, 1995).
The structure and biosynthesis of O-linked carbohydrate chains attached to serine and threonine show considerable evolutionary diversity. The primary reaction in the modification of mammalian O-linked proteins involves the attachment of a GalNAc that has been transferred from UDP-GalNAc within the Golgi (Roth, 1984). The carbohydrate chains of mammalian O-linked modified proteins are variable in length and composition and include galactose, sialic acid, fucose, Gal-NAc, and GlcNAc (Elhammer and Kornfeld, 1984;Roussel et al., 1988;Jentoft, 1990;Krijnse Locker et al., 1992). In contrast, it has been demonstrated that in S. cerevisiae O-modified proteins possess a linear carbohydrate chain consisting of up to 5 mannose residues (Tanner and Lehle, 1987;Herscovics and Orlean, 1993;Lehle and Tanner, 1995).
Some of the structural genes coding for yeast mannosyltransferases have been isolated. OCH1 encodes the first ␣1,6-mannosyltransferase involved in initiating outer chain elaboration (Nakayama et al., 1992;Nakanishi-Shindo et al., 1993). KRE2/ MNT1 is the only known ␣1,2-mannosyltransferase gene isolated to date  and is implicated in N-linked outer chain oligosaccharide synthesis (Hill et al., 1992) and is also responsible for the addition of the third mannose residue of O-linked carbohydrate chains . Outer chain and core modified oligosaccharides are brought to completion by the action of a terminal ␣1,3-mannosyltransferase encoded by the MNN1 gene and similarly to Kre2p/Mnt1p, Mnn1p also mannosylates O-linked glycans (Ballou, 1990;Yip et al., 1994).
To examine how post-translational modifications occur in Saccharomyces cerevisiae and to further define the responsible enzymes, we have functionally characterized three members of the KRE2/MNT1 putative mannosyltransferase gene family. This growing gene family was known to contain KRE2/MNT1, YUR1, KTR1, KTR2 (Hä usler and Robbins, 1992;Lussier et al., 1993), and recently two other homologues, KTR3 and KTR4, have been found by the yeast genome project (Mallet et al., 1994). These genes are predicted to encode type II membrane proteins with a short cytoplasmic N terminus, a membranespanning region, and a highly conserved catalytic lumenal domain. While the precise role of Kre2p/Mnt1p as a mannosyltransferase in O-glycosylation has been established , the role of the other genes remains to be determined. We have carried out a functional analysis of Yur1p, Ktr1p, and Ktr2p and demonstrate that they are mannosyltransferases involved in N-linked glycosylation.

EXPERIMENTAL PROCEDURES
Yeast Strains, Culture Conditions, and Methods-All yeast constructions used strain SEY6210 (MATa,. Yeast cells were grown under standard conditions (yeast extract peptone dextrose, Yeast Nitrogen Base buffered with Halvorson medium, when required) as described previously (Boone et al., 1990). Strains were transformed using the lithium acetate procedure using sheared, denatured carrier DNA (Gietz et al., 1995). Transformants were selected on synthetic minimal medium with auxotrophic supplements. Levels of sensitivity to K1 killer toxin was evaluated in SEY6210 by a seeded plate assay using a modified medium consisting of 0.67% YNB, 0.0025% required amino acids, 1.0% Bacto Agar (DIFCO, Detroit, MI), 0.001% methylene blue, 2% glucose, and buffered to pH 4.7 with Halvorson minimal medium (Lussier et al., 1993;Brown et al., 1994).
Gene Disruptions-Deletional disruptions of the KRE2/MNT1, KTR1, KTR2 and YUR1 genes were made by a single-step gene replacement procedure (Rothstein, 1991). The KRE2/MNT1 locus was disrupted with the TRP1 gene using the pAHI1 plasmid (kindly provided by Drs. A. Hä usler and P. W. Robbins, Massachusetts Institutes of Technology, Cambridge, MA). Briefly, the 1329-bp KRE2 gene was interrupted by replacement of the region from bases 78 to 1315 with a 1.2-kb fragment encoding TRP1. A linearized DNA fragment obtained by digestion with Asp718, comprising the kre2::TRP1 allele, was used for integration. The KTR2 gene disruption is described elsewhere (Lussier et al., 1993). Briefly, the gene was digested by XhoI and ClaI, which are situated, respectively, 146 bp downstream from the ATG and 388 bp upstream from the stop codon. This digestion removed a 744-bp fragment encompassing 248 amino acids of the KTR2 sequence. After treatment to fill in both restriction sites, a 1170-bp blunted DNA fragment containing the complete URA3 gene was ligated into the blunted XhoI and ClaI sites of KTR2. A linear 2.7-kb SpeI fragment containing the complete URA3 gene, as well as the coding plus flanking sequences from the KTR2 gene, was excised and used to disrupt the KTR2 locus into the isogenic diploid SEY6210 cells. The YUR1 gene was disrupted by deleting a 716-bp AccI-NsiI fragment in the coding region and replacing it by the complete 1.8-kb HIS3 gene fragment. Both deleted YUR1 gene fragment and HIS3 gene fragment were treated with the Klenow fragment of DNA polymerase I to accommodate the ligation. A 2.5-kb SpeI-EcoRV linearized fragment containing the complete HIS3 gene and the remains of YUR1 coding and flanking regions at each ends were used for integration in yeast. The KTR1 gene disruption was obtained by digestion of the gene with EcoRV, which deleted a 705-base pair fragment including 87 base pairs upstream from the ATG and replacing it by a 4.8-kb fragment containing the LYS2 gene. A linear 6-kb EcoRI-HindIII fragment containing the complete LYS2 gene, as well as the coding and flanking sequences from the KTR1 gene, was used to achieve integration at the KTR1 locus in yeast. All gene disruptions were confirmed by Southern analysis (data not shown).
Immunoblotting-The extracellular yeast proteins Kre9p and invertase were overexpressed from yeast cells harboring 2-based multicopy vectors and concentrated from cultures exponentially growing in Yeast Nitrogen Base selective media containing 5% (v/v) glycerol and 2% glucose using Amicon Centriprep concentrators (W. R. Grace & Co., Danvers, MA). Proteins were then separated by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes (Schleicher & Schuell). Blots were treated in TBST buffer (10 mM Tris, pH 8.0, 150 mM NaCl, 0.05% Tween 20, 5% nonfat dried milk powder) and subsequently incubated in TBST buffer with affinity purified anti-Kre9p or anti-invertase antibodies. After antibody binding, membranes were washed in TBST buffer and a second antibody directed against rabbit immunoglobulins and conjugated with horseradish peroxydase, was then added. The blots were again washed and proteins detected using an enhanced chemiluminescence procedure (Amersham Canada, Oakville, Ontario).
Mannan Acceptor Preparation-The mannoprotein acceptors were prepared from yeast cells according to a modification of the method of Ballou (1990). Pelleted cells from a 200-ml culture of quadruple mutant yeast (containing disruption in the KRE2, YUR1, KTR1, and KTR2 genes) grown in Yeast Nitrogen Base media were washed once with 0.9% NaCl, then with water, and the paste was resuspended in 50 ml of 0.02 M sodium citrate buffer, pH 6.8, and autoclaved for 90 min. After cooling and low speed centrifugation, the supernatant was kept, and the gelatinous solid pellet was re-extracted with citrate buffer. The supernatants were combined and poured into three volumes of methanol to precipitate the crude mannoprotein. After an overnight precipitation at 4°C, the mixture was centrifuged and the pelleted mannoprotein dissolved in water and dialyzed overnight against two changes of water. The crude mannoprotein was lyophilized, then dissolved in 30 ml of 7.5% CTAB, and left to stand at room temperature for 24 h. The solution was centrifuged, the supernatant was kept and an equal volume of 2% boric acid in water was added. The pH of the solution was then adjusted to 8.8 with KOH, and precipitation of the mannoprotein-CTAB complex was left to proceed overnight at room temperature. The solution was then centrifuged, the pellet was washed with 0.5% sodium borate buffer at pH 8.8 and dissolved in 12 ml of 2% acetic acid to dissociate mannoproteins from CTAB. Finally, the mannoprotein was precipitated in three volumes of ethanol, washed once with 2% acetic acid in ethanol then once in ethanol, and dissolved in water. The solution was dialyzed overnight at room temperature against water, then lyophilized. For use as acceptor in enzyme reactions, the mannoprotein fraction was resuspended at a concentration of 10 mg/ml in 50 mM Hepes, pH 7.2, 10 mM MnCl 2 , and 0.1% Triton X-100.
Preparation of Membranes and Assay of Mannosyltransferase Activity in Vitro-Mannosyltransferase activity assays were performed essentially as described (Lewis and Ballou, 1991;Lussier et al., 1995b). Briefly, reactions were carried out in a solution containing yeast membranes and consisting of 50 mM Hepes, pH 7.2, 10 mM MnCl 2 , 0.1% Triton X-100, 0.2 mM GDP-[ 14 C]mannose, and 10 mM ␣-D-methylmannopyranoside or 5 g of mannan acceptors. Values of specific mannosyltransferase activities (see Fig. 6) are expressed in dpm for a 10-min reaction and for 10 g of membrane proteins. Results represent the average of three independent determinations.
Preparation of Antisera-Invertase antiserum was obtained as described previously (Cooper and Bussey, 1989) by injecting into rabbits a recombinant protein corresponding to the N-terminal half of invertase. Briefly, a DNA fragment was excised with SmaI and XbaI from pSEY304 (kindly provided by Dr. T. H. Stevens, Institute of Molecular Biology, Eugene, OR), cloned into plasmid pEXP2, and expressed in Escherichia coli. Ktr1p antibodies were raised in rabbits against a bovine serum albumin-coupled synthetic peptide corresponding to the last 14 amino acid residues of the protein (NH 2 -NKLPKPAGWQNHIG-COOH; obtained from the Sheldon Biotechnology Centre, McGill University, Montréal, Québec, Canada). Initially, rabbits were injected with 500 g of conjugated peptide in Freund's complete adjuvant, followed by three subsequent injections with equivalent amounts of peptide in Freund's incomplete adjuvant at 3-4-week intervals. The conjugated peptides were coupled to cyanogen bromide-activated Sepharose CL-6B (Pharmacia Biotech Inc., Montréal, Québec, Canada) and used in a column to affinity-purify the antiserum as described by Raymond et al. (1990).
Epitope Tagging of YUR1 and KTR2-The influenza virus hemagglutinin epitope (sequence YPYDVPDYA; Kolodziej and Young (1991)) was inserted directly at the C-terminal domain of both Ktr2p and Yur1p. A NotI site was inserted in frame in both coding sequences by site-directed mutagenesis just before the termination codon. A triple epitope with NotI ends prepared from plasmid pSM191 (kindly provided by Dr. M. Manolson, Hospital for Sick Children, Toronto, Canada) was ligated into the modified coding sequences. Clones possessing the epitope insertion could be identified by restriction mapping a unique BamHI site situated in the triple hemagglutinin epitope. All positive clones were verified by DNA sequencing using the dideoxy chain-termination procedure (Sanger et al., 1977) with the Sequenase enzyme (Amersham), [␣-35 S]dATP, and specific DNA primers.

RESULTS
Functional Analysis of KTR1, KTR2, and YUR1-Ktr1p, Ktr2p, and Yur1p are predicted to possess a similar structure to Kre2p and mammalian glycosyltransferases: a short N-terminal cytoplasmic tail, a hydrophobic transmembrane domain, and a stem region, which links the large lumenal catalytic domain to the membrane-spanning region (Shaper and Shaper, 1992;Kleene and Berger, 1993;Lussier et al., 1995b) (see Fig.  1). The conserved region in the Kre2p family encompasses a large central region, with the N-and C-terminal portions being unique to each protein (for sequence comparison, see Lussier et al. (1993) and Mallet et al. (1994)). Yur1p and Ktr2p are the most similar members of the family, with 62% identity. However, both proteins are the most diverged from Kre2p, with just 37-38% identity at the amino acid level.
In view of their sequence and structural similarities with Kre2p, the possible role of the KTR1, KTR2 and YUR1 gene products as protein mannosyltransferases was analyzed. Onestep gene replacements were carried out using different marker genes (see "Experimental Procedures" and Fig. 2). KRE2 and KTR2 single gene disruptions were previously shown to have no growth phenotypes at 30°C Lussier et al., 1993). Analysis of spore progeny derived from SEY6210 ktr1::LYS2 or yur1::HIS3 heterozygotes showed that neither gene was essential for cell viability nor were they required for normal vegetative cell growth. 2 To assess whether a haploid strain carrying deletions in several of these genes possessed a more severe phenotype, double, triple, and quadruple disruptions were sequentially constructed using standard genetic techniques. Meiotic tetrads segregating combinations of these disrupted genes were dissected and haploid spore progeny grown at 30°C. Haploid strains harboring ktr1:LYS2 ktr2::URA3 yur1::HIS3 triple null mutations or haploids carrying a set of four disruptions were viable and did not grow noticeably slower than wild type cells. mannose-labeled yeast cells by ␤-elimination and resolved by chromatography. The wild type strain showed the normal profile of five oligosaccharide peaks (Fig. 3, Man1-Man5). The pattern obtained from the kre2 null strain gave two peaks (Man1 and Man2), consistent with failure to add the third ␣1,2-linked mannose residue in this mannosyltransferase-defective mutant. A ktr2 null mutation was previously shown to possess a wild type pattern of five oligosaccharide peaks (Lussier et al., 1993). The ktr1 and yur1 single null mutant strains also gave a wild type pattern of five mannose peaks 2 as did the ktr1 ktr2 yur1 triple null mutant (Fig. 3), providing no evidence for their involvement in O-linked chain elaboration.
The extent of O-glycosylation in yeast strains with mutations in these genes was also analyzed by measuring the mobility of a yeast O-glycoprotein, Kre9p. Kre9p is an extracellular matrix protein involved in cell wall assembly that is extensively Omannosylated but lacks N-linked modifications (Brown et al., 1993). When synthesized in a wild type strain, Kre9p migrates at an apparent mass of 55 kDa. As expected, Kre9p isolated from a kre2 null strain migrated more quickly than did the wild type Kre9p, with an apparent molecular mass of approximately 47 kDa (Fig. 4). However, Kre9p produced by ktr1, ktr2, or yur1 single null disruptants or by a triple null mutant strain was indistinguishable from that produced by a wild type strain.
Role of Ktr1p,Ktr2p,and Yur1p in N-Glycosylation-The Kre2p O-mannosyltransferase is also involved in the elaboration of N-linked carbohydrate chains (Hill et al., 1992); consequently, the effect of KTR1, KTR2 and YUR1 gene disruptions on invertase N-glycosyl modifications was analyzed. The prod-FIG. 1. Topologic representation and sequence similarities of Kre2p Ktr1p, Ktr2p, and Yur1p. A, the different members of the family are presumed to be oriented as a type II membrane-anchored protein, a topology characteristic of all isolated glycosyltransferases that consists of a short N-terminal cytoplasmic domain, a hydrophobic transmembrane domain, and a large C-terminal lumenal catalytic domain (Shaper and Shaper, 1992;Kleene and Berger, 1993). The catalytic domain is linked to the transmembrane domain by a "stem" region thought to be devoid of secondary structure. B, the degree of sequence homologies between the different proteins is represented as percentage of identities over the smallest protein and was calculated from sequence alignments with gaps to maximize homology. uct of the SUC2 gene, invertase, is a specifically N-modified protein, which is extensively glycosylated (Orlean, 1991;Ziegler, 1988). Three different classes of invertase can be distinguished: 1) a cytoplasmic form (60 kDa), which lacks a signal sequence and is therefore not glycosylated; 2) a transient ER form, which is heterogeneous in size (80 -90 kDa), as a consequence of the number of core oligosaccharide chains that are attached to the protein; and 3) a secreted form, which constitutes a array of differently modified proteins (Ͼ100 kDa) resulting in elongation from the core oligosaccharide of outer chain glycans in the Golgi complex (see Fig. 5).
As found previously (Hill et al., 1992), invertase synthesized in a kre2 null mutant has a molecular mass (Ϸ 137 kDa) that is smaller than the secreted wild type protein (Ϸ145 kDa). In contrast, the carbohydrate chains of invertase produced in ktr1, ktr2, or yur1 single null mutants appear to be intact, as the molecular mass of the protein made in these strains is wild type (Ϸ145 kDa). Similarly, in ktr1 ktr2, ktr1 yur1, or ktr2 yur1 double null mutants, no obvious reduction in size of invertase was apparent. However, invertase synthesized in a ktr1 ktr2 yur1 triple null mutant possessed a molecular mass of approximately Ϸ127 kDa. Invertase was smallest (Ϸ120 kDa) when produced in a quadruple kre2 ktr1 ktr2 yur1 mutant, indicating a cumulative involvement of all four proteins in N-linked modifications.
Mannosyltransferase Activity of Ktr1p, Ktr2p, and Yur1p-In parallel to the in vivo glycosylation studies, an analysis of the in vitro enzymatic activity of Ktr1p, Ktr2p, and Yur1p in carbohydrate chain elaboration was performed by measuring the transfer of [ 14 C]Man residues from GDP-[ 14 C]Man to a specific acceptor. To reduce the possible background mannosyltransferase activity, the enzymatic source for the in vitro assays consisted of membrane preparations from a kre2 ktr1 ktr2 yur1 quadruple null strain in which each of the genes was individually overexpressed (Fig. 6). In one assay, ␣-methylmannoside was used as an acceptor. A second assay used mannoprotein   type, kre2, yur1, ktr1, ktr2, and yur1 ktr1 ktr2 triple null mutants. Kre9p was overexpressed from plasmid YEp351 (Hill et al., 1986) in different yeast strains and concentrated from exponentially growing cultures (Brown et al., 1993). Yeast extracellular protein extracts were immunoblotted with affinity-purified anti-Kre9p polyclonal antibodies (see "Experimental Procedures"). The molecular mass standards are shown in kilodaltons. The S. cerevisiae O-linked oligosaccharide structures are also shown. Arrows depict ␣1,2, and ␣1,3 linkages between mannoses. The Kre2p/Mnt1p ␣1,2-mannosyltransferase is responsible for the addition of the third mannose in a medial Golgi compartment (Ballou, 1990;Lussier et al., 1995b). prepared from a kre2 ktr1 ktr2 yur1 quadruple null strain as an acceptor. If these mutant acceptor proteins are incompletely mannosylated, they may allow detection of an expanded range of mannosyltransferase activities. Using ␣-methylmannoside, a kre2 disruptant possesses about 38% residual activity when compared to that of wild type, 2 an activity consistent with that found previously . A strain carrying deletions of all four genes showed diminished enzymatic activity corresponding to about 18% of that found in a wild type, 2 indicating these deleted genes contribute to the total mannosyltransferase activity. The enzymatic activity of YUR1, KTR1, and KTR2 was subsequently assessed (Fig. 6). Using ␣-methylmannoside as an acceptor, an extract from the quadruple null strain overexpressing Kre2p showed an elevated (6.2-fold) enzymatic activity compared with an extract from the quadruple null strain. YUR1, when overexpressed, also displayed increased enzymatic activity (5.5-fold), demonstrating that Yur1p is a mannosyltransferase. However, using this acceptor, extracts from strain overexpressing KTR1 and KTR2 displayed activity levels equivalent to that of background. To attempt to detect mannosyltransferase activities using an alternative assay, a mannoprotein preparation from the quadruple null strain was used as an acceptor. Again a net increase in activity was seen with KRE2 (2.8-fold) and YUR1 (2.1-fold) when these genes were overexpressed singly in the quadruple null background. Activity higher than background was also detected with the KTR1 and KTR2 genes, although at lower levels. Strains overexpressing KTR2 or KTR1 had mannosyltrans-ferase activities that were 1.8-fold higher and 1.5-fold higher than background, respectively.
Multiple Disruptions of KTR1, KTR2, and YUR1 Lead to K1 Killer Toxin Resistance-The role of the Ktr1p, Ktr2p, and Yur1p proteins as mannosyltransferases was also assessed in vivo by using a K1 killer toxin sensitivity assay (Fig. 7). K1 killer yeast strains secrete a small pore-forming toxin that requires a cell wall receptor for function (Bussey, 1991). This receptor appears to consist of the glycosyl moieties of cell wall glucomannoproteins. Killer-resistant mutants have been found to be defective in ␤1,6-glucan and in O-mannosylation, suggesting that the in vivo receptor includes these polymers, which are cross-linked in cell wall glucomannoproteins (Montijn et al., 1994;Lu et al., 1995).
Yeast strains harboring single and double mutations, as well as a triple null mutation, of KTR1, KTR2 and YUR1 were assayed for killer toxin sensitivity (Fig. 7). When evaluated by seeded plate assays, the wild type toxin-sensitive SEY6210 strain displayed a large killing zone (15 mm), whereas the kre2 mutant was completely toxin-resistant. A strain bearing a single KTR1 disruption showed no phenotypic resistance to K1 killer toxin. When compared to the wild type strain, yur1 (10 mm) and ktr2 (12.5 mm) single null disruptions were both partially resistant to the killer toxin, yur1 being more resistant. Yeast cells carrying ktr1 ktr2, ktr1 yur1, or ktr2 yur1 double disruptions all displayed pronounced levels of resistance. The ktr1 ktr2 (10 mm) double null strain showed a stronger phenotype than either the ktr1 or ktr2 single null. Both ktr1 yur1 (8 mm; clear) and ktr2 yur1 (7 mm; fuzzy) double nulls are more resistant than a strain carrying a yur1 null mutation indicating that disruption of either KTR1 or KTR2 exacerbates the cell wall defect of a yur1 mutant. Finally, a ktr1 ktr2 yur1 FIG. 5. Immunological detection of invertase synthesized in wild type and different null mutants. Invertase was overexpressed from plasmid YEp351 (Hill et al., 1986) in different yeast strains and concentrated from exponentially growing cultures. Yeast extracellular protein extracts were immunoblotted with anti-invertase polyclonal antibodies (see "Experimental Procedures"). The molecular mass standards are shown in kilodaltons. The S. cerevisiae possible N-linked oligosaccharide structures are also shown (adapted from Ballou (1990)). ␤1,4, ␣1,6, ␣1,2, and ␣1,3 linkages between mannoses of the core and outer chain are depicted. x equals 10 on average.

FIG. 6. Mannosyltransferase activity in vitro.
Total membrane preparations from a kre2 ktr1 ktr2 yur1 quadruple null strain (SEY 6210) overexpressing each gene individually from YEp351 (Hill et al., 1986) were assayed for their ability to transfer of [ 14 C]mannose from GDP-[ 14 C] mannose to a specific acceptor. Assays were carried out as described under "Experimental Procedures." Values are given as specific activities obtained with ␣-methylmannoside or mannoprotein prepared from the quadruple null strain as acceptors.
triple null mutant is almost totally resistant, suggesting a cumulative effect on the reduction of carbohydrate chains leading to killer resistance. These results thus appear to also implicate N-linked chains as part of the killer receptor.
The killer phenotype of some single null mutants allowed a test of possible suppression of the loss of one gene by another homologous counterpart. Ktr1p, Ktr2p, and Yur1p could not suppress the killer resistance of a KRE2 null mutant and thus could not functionally substitute for it. 2 Functional suppression could only be established between the YUR1 and KTR2 genes, with overexpression of KTR2 in a strain carrying a yur1 null mutation completely suppressing the yur1 killer resistance phenotype, indicating that when expressed at very high levels, Ktr2p has the capacity to substitute in vivo for the absence of the Yur1p. 2 These two proteins are 62% identical and constitute the most homologous pair among members of the Kre2p family (Lussier et al., 1993;Mallet et al., 1994).
Ktr1p, Ktr2p, and Yur1p Are Localized in the Yeast Golgi Complex-The Kre2p ␣1,2-mannosyltransferase has been localized to a medial Golgi compartment (Chapman and Munro, 1994;Lussier et al., 1995b). The apparent role of Ktr1p, Ktr2p and Yur1p as glycosyltransferases and their similarity to Kre2p make these proteins candidates for Golgi localization and this was examined by indirect immunofluorescence. To identify and analyze the localization of the three proteins, a specific rabbit antiserum was raised against Ktr1p, and the influenza hemagglutinin virus epitope (Kolodziej and Young, 1991) was inserted directly at the C-terminal domain of Ktr2p and Yur1p (see "Experimental Procedures"). An affinity-purified anti-Ktr1p Ab detected Ktr1p in Western blotting of total cell protein extracts as a 55-kDa protein that was absent from the ktr1::LYS2 strain. 2 Immunoblots using the 12CA5 hemagglutinin epitope specific monoclonal antibody detected only the epitope-tagged version of Ktr2p and Yur1p. 2 When the relevant antibodies, plasmids and strains were used for whole cell indirect immunofluorescence, all three proteins showed a punctate pattern of fluorescent signals (Fig. 8) indicative of Golgi localization (Redding et al., 1991;Cooper and Bussey, 1992;Roberts et al., 1992;Graham et al., 1994;Lussier et al., 1995b). In each case, between 3 and 10 structures/cell can be seen, depending on individual cells and the plane of focus. Fluorescence signals were never seen with the anti-Ktr1p antibody in ktr1::LYS2 cells nor with the 12CA5 monoclonal antibody with cells not expressing tagged versions of the Ktr2 and Yur1 proteins. 2 The signal distribution of Ktr1p, Ktr2p, and Yur1p did not overlap with nuclei or mitochondria as viewed by DNA staining with DAPI (Fig. 8).

DISCUSSION
Kre2p is an ␣1,2-mannosyltransferase , and we present here evidence showing that Ktr1p, Ktr2p, and Yur1p are also involved in protein glycosylation. As Kre2p is a mannosyltransferase adding the third mannose residue on O-linked mannose carbohydrate chains, the possible role of Ktr1p, Ktr2p, and Yur1p in O-glycosylation was analyzed. Experiments indicated that neither the O-glycosylation of total yeast mannoprotein nor the O-glycosylation of Kre9p is affected by these proteins, since no differences from wild type were seen in single or triple null mutants (ktr1 ktr2 yur1).
The influence of KTR1, KTR2, and YUR1 gene disruptions on protein N-glycosylation was analyzed. The N-glycosylated protein invertase was found to be underglycosylated in the ktr1 ktr2 yur1 triple null mutant compared to a wild type strain but not in single or double disruptants, except in the case of kre2 where, as expected, an effect was seen (Hill et al., 1992). Invertase receives even less glycosylation when synthesized in the quadruple, ktr1 ktr2 yur1 kre2, null strain but is still heavily N-modified since its migration pattern (Ϸ120 kDa) remains considerably larger than the molecular mass of the protein predicted from the DNA sequence (59 kDa). These  (Hill et al., 1986), were fixed, spheroplasted, attached to polylysine-coated glass slides, and then incubated with affinity-purified anti-Ktr1p Ab or 12CA5 monoclonal antibody and DAPI. Texas Redcoupled secondary Ab was added to detect antigen-immunoglobulin complexes. Cellular DAPI staining of nuclear and mitochondrial DNA is shown.
results are consistent with these enzymes having redundant functions in N-linked glycosylation.
Possible additional roles for Ktr1p, Ktr2p, Yur1p, and Kre2p were also assessed. S. cerevisiae carries several phosphoinositol)-containing sphingolipids, specifically inositol phosphoceramides, mannosylinositol phosphoceramides (which contain a mannose attached to the inositol), and mannosyl(inositol phospho) 2 ceramides (which is substituted with one mannose and 2 phosphoinositol groups). Ktr1p, Ktr2p, Yur1p, and Kre2p do not appear to be involved in this lipid mannosylation, as none of the strains containing single or multiple deletions of these genes lacked any of the mannosylated inositol phospho-ceramides. 3 Another possibility is that Ktr1p, Ktr2p, Yur1p, and Kre2p elaborate part of the short ␣-linked mannose side chains found on protein-bound GPI anchors. None of the enzymes could be solely responsible for a single biosynthetic step in GPI core synthesis, since this is an essential process in yeast (Leidich et al., 1994, and none have a lethal phenotype when disrupted. Strains carrying single or multiple deletions of these four genes all synthesized normal GPI anchors (Sipos et al., 1995).
Evidence for the function of Yur1p, Ktr1p and Ktr2p as mannosyltransferases was obtained by evaluating their in vitro enzymatic activities. Using ␣-methylmannoside and oligosaccharides found on mannoproteins from a kre2 ktr1 ktr2 yur1 quadruple null strain as acceptors, YUR1 overproducing cells showed a 5.5-and 2.1-fold increase in activity over background respectively. These values are similar to those obtained by overproduction of the known mannosyltransferase encoding gene, KRE2. By comparison, when ␣-methylmannoside was used as a saccharide acceptor, high levels of expression of KTR1 and KTR2 did not result in activity levels higher than those obtained with the parental quadruple null strain. However, when the mannoprotein fraction was used as an acceptor, increased levels of activity similar to those obtained with YUR1 were reproducibly found with both KTR1 and KTR2, indicating that their gene products also are mannosyltransferases. The activity difference seen between these enzymes in the two assays suggests that these mannosyltransferases differ in substrate specificity.
Further evidence that Ktr1p, Ktr2p, and Yur1p are mannosyltransferases comes from an in vivo analysis of their function. Strains carrying non-functional copies of KTR1, KTR2, and/or YUR1 genes, became to varying extents K1 killer toxinresistant, the triple null mutant being most resistant. These results indicate that, as is the case with KRE2 null mutations, singly or in combination disruptions of KTR1, KTR2, and YUR1 lead to a reduced amount of N-linked glycans on cell wall mannoproteins perturbing the cell surface toxin receptor and leading to resistance. The fact that functional replacement by overproduction could only be obtained between the most similar gene pair, YUR1 and KTR2, also suggests that these mannosyltransferases likely perform different functions.
From the above results, it can be concluded that Ktr1p, Ktr2p, and Yur1p are implicated as mannosyltransferases in N-linked glycan elaboration. However, these enzymes do not participate in the synthesis of the basic N-linked core oligosaccharide, as they are situated in the Golgi apparatus and the core oligosaccharide is elaborated and transferred to protein in the ER (Herscovics and Orlean, 1993;Lehle and Tanner, 1995). Similarly, Ktr1p, Ktr2p, and Yur1p do not participate in core Golgi modifications (see Fig. 5), as the size of the core modified oligosaccharide received by the late Golgi protein Kex1p (Coo-per and Bussey, 1992) is the same in the triple ktr1 ktr2 yur1 null mutant and in wild type. 2 Thus, it is likely that the role of Ktr1p, Ktr2p, and Yur1p is to participate in the elaboration of the outer chain glycans of N-linked oligosaccharides. A number of distinct ␣1,2-linked mannosylation reactions are required for the synthesis of the outer chain of yeast N-linked modified proteins (see Fig. 5), and we speculate that Ktr1p, Ktr2p, Yur1p, and also Kre2p (Hill et al., 1992) are partially responsible for establishing some of these ␣1,2-linkages in the Golgi apparatus.
The cumulative effect of multiple gene disruptions seen on the size of the N-linked carbohydrates carried by invertase and on the degree of in vivo killer toxin resistance can be rationalized in distinct ways that are not necessarily mutually exclusive. Ktr1p, Ktr2p, Yur1p, Kre2p, and other similar mannosyltransferases could function redundantly in the sense of having overlapping specificities. Different forms of functional redundancy can be envisaged in the context of a large family of glycosyltransferases elaborating complex glycans. 1) More than one enzyme could be able to establish one specific class of glycosyl linkage. This could happen in normal vegetative growth or could be achieved by differential regulation under specific conditions. The PMT gene family encoding protein Omannosyltransferases constitutes an example of this type of redundancy (Strahl-Bolsinger et al., 1993;Lussier et al., 1995a;Immervoll et al., 1995). 2) Conversely, an individual mannosyltransferase may catalyze the assembly of one type of carbohydrate linkage to more than one type of oligosaccharide, as is the case for the Mnn1p terminal ␣1,3-mannosyltransferase (Ballou, 1990;Cooper and Bussey, 1992;Graham et al., 1994;Yip et al., 1994;Lussier et al., 1995b). 3) Added complexity may arise if members of the Kre2p family are not localized in the same Golgi compartment. Some member proteins could possess the same enzymatic specificity, but their Golgi retention signal would be different, targeting them to distinct intracellular locations. A key element in the targeting of Kre2p to the medial Golgi has been shown to lie in the N-terminal region (Lussier et al., 1995b). This interesting region of the sequence of Kre2p, Ktr1p, Ktr2p, and Yur1p is unique to each protein hinting that they may be localized to different Golgi subcompartments.
Taken together, our results indicate that Ktr1p, Ktr2p, Yur1p, and Kre2p are involved in the elaboration of outer chain N-linked glycans. The specificity of the mannosyltransferase reactions catalyzed by these four enzymes is likely to vary, as each showed a different pattern of activity toward the two acceptors used in our in vitro assays. Multicopy suppression of the phenotype caused by one deleted transferase by another provides an indication of at least some partial overlap of mannosyltransferase specificity. The lack of multicopy suppression, however, provides little information as enzymes with similar specificity may reside in different Golgi compartments or be differentially regulated. A clearer picture of the overall specificity, location, and regulation of the KRE2/MNT1 family awaits the identification and characterization of the entire gene family in S. cerevisiae, a goal likely attainable by the completion of the genome sequence of this organism.