key: cord-104268-q1jx0n0l authors: nan title: Localization and targeting of the Saccharomyces cerevisiae Kre2p/Mnt1p alpha 1,2-mannosyltransferase to a medial-Golgi compartment date: 1995-11-02 journal: J Cell Biol DOI: nan sha: doc_id: 104268 cord_uid: q1jx0n0l The yeast Kre2p/Mnt1p alpha 1,2-mannosyltransferase is a type II membrane protein with a short cytoplasmic amino terminus, a membrane- spanning region, and a large catalytic luminal domain containing one N- glycosylation site. Anti-Kre2p/Mnt1p antibodies identify a 60-kD integral membrane protein that is progressively N-glycosylated in an MNN1-dependent manner. Kre2p/Mnt1p is localized in a Golgi compartment that overlaps with that containing the medial-Golgi mannosyltransferase Mnn1p, and distinct from that including the late Golgi protein Kex1p. To determine which regions of Kre2p/Mnt1p are required for Golgi localization, Kre2p/Mnt1p mutant proteins were assembled by substitution of Kre2p domains with equivalent sequences from the vacuolar proteins DPAP B and Pho8p. Chimeric proteins were tested for correct topology, in vitro and in vivo activity, and were localized intracellularly by indirect immunofluorescence. The results demonstrate that the NH2-terminal cytoplasmic domain is necessary for correct Kre2p Golgi localization whereas, the membrane-spanning and stem domains are dispensable. However, in a test of targeting sufficiency, the presence of the entire Kre2p cytoplasmic tail, plus the transmembrane domain and a 36-amino acid residue luminal stem region was required to localize a Pho8p reporter protein to the yeast Golgi. cesses have been studied from genetic and biochemical perspectives (Kukuruzinska et al., 1987; Pryer et al., 1992) resulting in a functionally defined Golgi apparatus that has been subdivided into several subcompartments on the basis of asparagine-linked (N-linked) 1 oligosaccharide modifications (Franzusoff and Schekman, 1989; Graham and Emr, 1991; Wilsbach and Payne, 1993; Gaynor et al., 1994) . In yeast, secretory pathway glycoproteins can acquire two types of N-linked oligosaccharides, either a simple core carbohydrate or one extended by an outer-chain glycan structure. The N-linked core oligosaccharide elaborated in the ER is mainly constituted of MansGlcNAc2 and may undergo Golgi maturation resulting in Man9.13_ GlcNAc2. In other cases, glycoproteins traversing the Golgi apparatus have their core oligosaccharide extended by outer chains (see Fig. 2 ; Ballou, 1990; Herscovics and Orlean, 1993) . The initial step in outer-chain synthesis takes place and defines an early Golgi compartment in which a backbone of ed,6-1inked mannose residues is attached to the core oligosaccharide. This Golgi compartment has been likened to the mammalian cis-Golgi network (Wilsbach and Payne, 1993) . Outer chain elaboration is then brought to completion: the al,6-1inked mannose structure is enlarged in a sequential fashion by al,2-and al,3-mannosyltransferases in a subsequent distinct com-partment defining the yeast medial-Golgi compartment. Mature glycosylated proteins are finally transported to a late Golgi compartment where proteolytic processing of secreted protein precursors occurs. Some evidence suggests that this Golgi subcompartment is also involved in vacuolar protein sorting, making this late Golgi compartment comparable to the mammalian TGN (Wilsbach and Payne, 1993) . In S. cerevisiae, the various Golgi compartments have not been morphologically well characterized. The yeast Golgi complex is not structurally similar to the perinuclear stacked cisternal subcompartments characteristic of the Golgi apparatus of mammalian cells. Immunoelectron microscopy revealed that the yeast Golgi is composed of as many as 30 concave, disklike membranous cisternae scattered in the cytoplasm and most often not organized in parallel stacks (Preuss et al., 1992) . In indirect immunofluorescence using antibodies to different yeast Golgi proteins, the isolated cisternae are visualized as dispersed punctate spots (Franzusoff et al., 1991; Redding et al., 1991; Antebi and Fink, 1992; Cooper and Bussey, 1992; Roberts et al., 1992) . The establishment and maintenance of the polarized organization of the Golgi apparatus relies on the existence of complex sorting and transport mechanisms permitting specific Golgi proteins to be delivered to their precise cisternal destinations. Resident proteins of the secretory pathway of eukaryotic cells require particular targeting signals that specify their final location (Pelham and Munro, 1993; Gleeson et al., 1994; Low and Hong, 1994) . Mammalian membrane glycoproteins not possessing positive sorting sequences are transported to the cell surface through a bulk flow of proteins and lipids (Pelham and Munro, 1993) . In yeast, secretory pathway membrane glycoproteins lacking targeting sequences are brought by default to the vacuole (Cooper and Bussey, 1992; Roberts et al., 1992; Wilcox et al., 1992; Nothwehr et al., 1993; Gaynor et al., 1994; Hill and Stevens, 1994) . Proteins associated with glycan modifications constitute a major class of resident Golgi proteins. Enzymes belonging to different mammalian glycosyltransferase families share a similar type II structural organization, but lack amino acid sequence homology or discernible targeting motifs even if situated in the same Golgi compartment (Shaper and Shaper, 1992; Kleene and Berger, 1993) . It has been established that the membrane-spanning domain of animal glycosyltransferases plays a central role in Golgi localization (Munro, 1991; Nilsson et al., 1991; Swift and Machamer, 1991; Aoki et al., 1992; Burke et al., 1992; Colley et al., 1992; Tang et al., 1992; Teasdale et al., 1992; Wong et al., 1992; Gleeson et al., 1994; Low and Hong, i994) . To define more precisely the S. cerevisiae Golgi complex and to better understand how posttranslational modifications occur, we have studied the glycosyltransferase, Kre2p/Mntlp. KRE2/MNT1 was isolated as a gene implicated in cell-wall assembly conferring K1 killer toxin resistance when mutated (Hill et al., 1992) and found to encode an cd,2-mannosyltransferase (H/iusler and . KRE2/MNT1 encodes a 442-amino acid residue predicted type II membrane protein containing a putative transmembrane domain near its NH2 terminus and one potential luminal N-glycosylation site. Kre2p/Mntlp specifically adds a third mannose during the linear elongation of O-linked carbohydrate chains that may contain up to five mannose residues (H~usler et al., 1992) and is also apparently involved in asparagine-linked glycosylation (Hill et al., 1992; Lussier et al., 1995b) . Here we report further characterization of the KRE2/ MNTl-encoded mannosyltransferase. We demonstrate that Kre2p/Mntlp possesses a type II orientation, is progressively N-glycosylated in an MNNl-dependent fashion, and is localized in a Golgi subcompartment that overlaps with the Mnnlp medial-Golgi compartment. In addition, we show that the short NH2-terminal cytoplasmic tail domain of Kre2p/Mntlp is required for correct Golgi localization, whereas the membrane-spanning and stem domains are dispensable. However, localization of a reporter protein to the yeast Golgi requires a region of Kre2p/Mntlp encompassing the cytoplasmic tail, the transmembrane domain (TMD), and a segment of the luminal stem domain. All yeast constructions used strains SEY6210 (MATa, (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 were 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, 0.001% methylene blue, 2% glucose, and buffered to pH 4.7 with Halvorson minimal medium (Lussier et al., 1993 (Lussier et al., , 1995a Brown et al., 1994) . (underlined) . Finally, KPKP was obtained by replacing the Kre2p TMD in KKKP with that of Pho8p using the same oligonucleotide that was used to obtain KPKK All chimeric protein constructions were verified by DNA sequencing using the dideoxy chain termination procedure (Sanger et al., 1977) with the Sequenase enzyme (United States Biochemical Corp., Cleveland, OH), a-asS-dATP and specific DNA primers. Kre2p antibodies were raised in rabbits against a BSA-coupled synthetic peptide corresponding to the last 10 amino acid residues of the protein (NH2-KPKNWKKFRE-COOH; obtained from the Sheldon Biotechnology Centre, McGill University, Montr6al, Qu6bec, 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-wk intervals. The conjugated Kre2p peptide was coupled to cyanogen bromide-activated Sepharose CL-6B (Pharmacia LKB Biotechnology, Inc., Montr6al, Canada) and used in a column to affinity purify the antiserum as described by Raymond et aL (1990) . Yeast total-ceU protein extracts were prepared from cultures growing exponentially in yeast nitrogen base selective medium by cell lysis with glass beads in the presence of protease inhibitors. Membrane fractions were prepared as described (Nakayama et al., 1992) by centrifuging cell lysates at 10,000 g for 20 min and by centrifuging the resulting supernatant at 100,000 g for 1 h. The high speed pellet contained the insoluble membrane fraction. Yeast proteins were separated by SDS-PAGE, and immunoblots were carried out mainly as described (Lussier et al., 1995a) . Briefly, blots were treated in TBST buffer (10 mM Tris, pH 8.0, 150 mM NaC1, 0.05% Tween 20, 5% nonfat dried milk powder) and subsequently incubated with affinity-purified anti-Kre2p antibodies in the same buffer. After antibody binding, membranes were washed in TBST and a second antibody directed against rabbit immunoglobulins and conjugated with alkaline phosphatase, was then added. The blots were again washed and proteins detected using an enhanced chemiluminescence procedure (Amersham Canada, OakviUe, Ontario). Analysis of [35S]methionine-labeled proteins was as described (Cooper and Bussey, 1989) . Briefly, cells were grown to an OD600 of 0.4-0.45 in selective medium (YNB) and then labeled with Trans35S (100 ~Ci; ICN Biochemicals Inc., Montr6al, Canada) for 10 min at 30°C. Yeast cells treated with tunicamycin (10 ~g m1-1) were preincubated before radiolabeling for 30 min at 30°C. Tunicamycin is a hydrophobic analogue of UDP-N-acetylglucosamine that blocks the addition of N-acetylglucosamine to dolichol phosphate, the first step in N-linked oligosaccharide formation. Chase conditions were achieved by addition of L-methionine and L-cysteine to a final concentration of i mM. txl,2-mannosyltransferase activity assays were performed essentially as described (Lewis and Ballou, 1991; Hhusler and Robbins, 1992) . S. cerevisiae cells ($86 background) were grown in selective medium to an OD600 of 0.8-0.9 and lysed with glass beads in the presence of protease inhibitors. High speed pellet fractions including Golgi and vacuolar membranes were prepared by centrifuging cell lysates at 1,1300 g for 20 min and by centrifuging the resulting supernatant at 100,000 g for 1 h at 4°C. Incubation mixtures contained 50 mM Hepes, pH 7.2, 10 mM MnCI2, 0.1% Triton X-100, 0.2 mM GDP-[lgC]mannose, and 10 mM C~-D-methylmanno-pyranoside as an acceptor (Sigma Chemical Co.) and 10 p,1 of membrane fraction in a total vol of 20 tzl. Reaction mixtures were incubated for 30--60 min at 30°C and then passed through a resin (AG1-X4; Bio-Rad Laboratories, Hercules, CA) column to remove unreacted GDP-mannose. Neutral products were eluted with 1 ml of water and radioactivity was measured. Control assays were conducted in which the saccharide acceptor was omitted and counts obtained in these assays were subtracted from values obtained in assays made with the sugar acceptor. Enzymatic activities (see Fig. 9 ) are expressed as percentages of specific activity (mmol/h/mg of membrane protein) for each chimeric protein. Yeast cells were grown in selective medium to an OD600 of 0.5-0.7. and immediately fixed by the addition to cultures of potassium-phosphate (pH 6.5) to 0.1 M, and formaldehyde to 3.7%. After gentle agitation for 30 rain, cells were pelleted and resuspended in a fixation buffer containing 0.1 M potassium-phosphate (pH 6.5) and 4.5% formaldehyde. Cells were further fixed for another 30 rain. Formaldehyde-treated ceils were washed with 0.1 M potassium-phosphate (pH 6.5) buffer, resuspended in a solution of IOO mM Tris-HC1, pH 8, 25 mM DTr, 5 mM EDTA, and 1.2 M sorbitol and incubated for 10 min at 30°C with gentle agitation. After washing of the fixed cells, cell walls were removed by treatment with Zymolyase 100T (ICN Biochemicals) at a final concentration of 200 ixg/ml in 0.1 M potassium phosphate (pH 6.5) 1.2 M sorbitol for 20-30 min at 30°C. Fixed spheroplasted cells were washed in 0.1 M potassium-phosphate (pH 7.4) 1.2 M sorbitol and resuspended in the same buffer. Cells were subsequently adsorbed to poly-L-lysine coated microscope slides, permitted to stand for 10 min and washed with PBS. Slides were then immersed in -20°C methanol for 1-6 min, and then for 20450 s in -20°C acetone, depending on yeast strains. Treated slides were air-dried and used immediately or stored at 4°C until needed. Fixed mounted cells were incubated with primary antibodies diluted in PBS containing 0.5 mg/ml BSA for 1 h at 25°C or overnight at 4°C. Anti-Kre2p Ab was used at dilutions of 1:25-1:100. For dual-labeling experiments involving Kre2p and medialor late Golgi markers, the influenza hemagglutinin virus epitope (sequence YPY-DVPDYA) was inserted by oligonucleotide-directed mutagenesis directly at the COOH-terminal domain of Mnnlp and in the region corresponding to the Kexlp luminal domain. Epitope-tagged Mnnlp and Kexlp were detected with the 12CA5 monoclonal antibody (Kolodziej and Young, 1991) . The latter was used at dilutions ranging from 1:250--1:1,000. mAb 13Dll which recognizes the 60-kD subunit of the yeast vacuolar membrane H+-ATPase (Kane et al., 1992) was used at dilutions of 1:10-1:25 as a vacuolar marker for colocalization studies with Kre2p chimeric proteins. Fluorescence signals were obtained by subsequent incubation of treated cells with rhodamine X sulfonyl chloride (Texas red)-conjugated goat anti-rabbit IgG (1:50-1:200) and FfTC-conjugated goat anti-mouse IgG (1:50-1:200) which were used as secondary antibodies. Nuclei and mitochondria were visualized by staining with 4',6-diamidino-2-phenyl-indole (DAPI). Cells were examined with an epifluorescence microscope (Axiophot; Carl Zeiss, Inc., Thornwood, NY), and photographed with film (T-Max 400; Eastman Kodak Co., Rochester, NY). The extent of colocalization of Kre2p with Mnnlp or Kexlp was scored by quantitating in a given dual localization experiment 30 cells that contained clearly defined signals representing 200-250 individual fluorescent punctate spots for each antigen. Compiled data revealed that Mnnlp gave rise to approximately the same number of punctate spots per cell as Kre2p and 75% of the punctiform fluorescent spots from Texas red and FITC overlapped. Kexlp gave rise to ~10% fewer punctate spots per cell compared with Kre2p-associated signals, and 65% of the punctiform fluorescent spots emanating from both proteins did not overlap. Finally, the intracellular localization of chimeric protein KKP was quantitatively scored by examining 750 individual ceils containing clearly defined signals. Punctiform fluorescence different from nucleus, ER, or vacuole was defined as Golgi localization. Vacuolar localization was determined by colocalization with the 60-kD vacuolar membrane H+-ATPase subunit. The product encoded by the KRE2 gene is a predicted type II membrane protein of 442 amino acid residues with a short cytoplasmic NH2 terminus, a hydrophobic transmembrane region, and a large luminal enzymatic domain containing one potential N-glycosylation site ( Fig. 1 A) . To identify and subsequently analyze Kre2p, a specific rabbit antiserum was raised, affinity purified, and used to detect antigen-antibody complexes by Western blotting of total-cell protein extracts of yeast strains harboring a KRE2 disruption or containing a KRE2 multicopy plasmid to facilitate immunological detection. The affinity-purified Ab detected Kre2p as a 59-60 kD protein in the KRE2overexpressing strain that was absent from the kre2::TRP1 strain ( Fig. 1 B) , and not detected by preimmune antiserum (data not shown). The in vivo-produced Kre2p is ~8.0 kD larger than the 51.5-kD molecular mass predicted from the D N A sequence. The possible integral membrane protein nature of Kre2p was examined by using reagents extracting cytoplasmic, vesicle-enclosed, and peripheral membrane proteins but leaving intact tightly associated membrane proteins. Ceils expressing KRE2 at high levels were lysed with glass beads and treated with sodium carbonate or urea. After high-speed centrifugation of the treated cell lysates, the distribution of Kre2p in the pellet and supernatant fractions was assessed by Western immunoblotting. As can be seen in Fig. 1 C, Kre2p was found only in membrane pellet fractions. The difference between the expected (51.5 kD) and observed molecular mass of Kre2p ( Fig. 1 B) is likely to be at Kre2p is oriented as a type II membrane-anchored protein, a topology characteristic of all isolated glycosyltransferases (Shaper and Shaper, 1992; Kleene and Berger, 1993; Gleeson et al., 1994) . Kre2p consists of a short amino-terminal cytoplasmic domain, a hydrophobic transmembrane domain, and a large carboxy-terminal luminal catalytic domain. The catalytic domain is linked to the transmembrane domain by a polypeptidic "stem" region. The stem region is generally thought to be devoid of secondary structure. (B) Immunological detection of Kre2p. Yeast total-protein extracts from kre2::TRP1 in SEY6210 and from the same strain expressing KRE2 from the multicopy plasmid YEp352 (Hill et al., 1986) , were immunoblotted with affinity-purified anti-Kre2p polyclonal antibodies (see Materials and Methods). The molecular mass standards are shown in kilodaltons. (C) Kre2p is membrane associated. SEY6210 cells overexpressing KRE2 were lysed, and cell debris was removed by centrifugation. The resulting crude homogenate was aliquoted, and fractions were rendered 0.1 M with Na2CO 3 or 1.6 M with urea. After 30 min, each fraction was centrifuged at 100,000 g for 1 h and after SDS-PAGE, Kre2p was detected by immunoblotting with the anti-Kre2p Ab in pellet (P) and supernatant (S) fractions. least partly due to N-glycosylation, since Kre2p is expected to act in the secretory pathway and the protein possesses a single site for N-glycosyl attachment in its predicted luminal domain (Asna97-Gln-Thr) (see Fig. 6 ). To test for Kre2p N-glycosylation, yeast cells were [35S]methionine labeled in the presence or absence of tunicamycin, an inhibitor of N-glycosylation. Immunoprecipitation and S D S -P A G E analysis of labeled cell lysates revealed that Kre2p was N-glycosylated (see Fig. 3 , lanes 1 and 2), with the position of the sole N-glycosylation attachment site at A s n 197 being consistent with a type II topology for Kre2p. The molecular mass of Kre2p in the presence of tunicamycin is about 54 kD, still 2.5 kD larger than its pre-dicted molecular mass. The membrane nature of Kre2p or other posttranslational modifications could explain this discrepancy. To establish possible Golgi modifications that Kre2p might acquire, and to attempt to assess to which Golgi subcompartments it had been exposed, the size of the Kre2plinked oligosaccharide chains was examined in specific N-linked glycosylation mutants. The mnn strains carry mutations at various loci resulting in glycosylation defects ( Fig. 2 ; Ballou, 1990; Ballou et al., 1990) . The N-linked carbohydrates from an mnn9 strain lack the outer chain. The mnnl strain produces glycoproteins with N-linked chains lacking terminal c~l,3-mannosyl residues (Fig. 2) . Wild-type, mnnl, and mnn9 mutant yeast strains carrying KRE2 were [35S]methionine labeled for 10 rain and chased for 45 min, and the extent of glycosylation of immunoprecipitated Kre2p was examined (Fig. 3) . The size of Kre2p produced in a wild-type strain after a 10-min radiolabeling is ~57 kD (major band, lane 2). After an additional 45-min chase, an apparent increase of molecular mass from 57 to ~59-60 kD was seen (Fig. 3 , lanes 2 and 5). This apparent 2-3 kD increase in mass suggests that Kre2p may undergo a post-ER modification not involving extensive outer chain elaboration. The molecular mass of Kre2p produced in an mnn9 strain after a 10-min pulse (Fig. 3, lanes 2 and 4) and a 45 min chase (lanes 5 and 7) was identical to the mass of Kre2p from a wild-type cell, indicating that Kre2p does not receive an outer chain oligosaccharide. We tested whether the time-dependent additional carbohydrate modification required the MNN1encoded Golgi al,3-mannosyltransferase, an enzyme that acts on both core and outer chains ( Fig. 2 ; Ballou et al., 1990; Graham and Emr, 1991; Graham et al., 1994; Yip et al., 1994) . After a short pulse, Kre2p synthesized in a mnnl strain is approximately of wild-type size (Fig. 3, lane 3) . After a 45-min chase, however, the mnnl-produced Kre2p was ~1-2 kD smaller than the wild-type protein (lanes 5-7) indicating that Kre2p is normally exposed to a Golgi compartment where the Mnnlp cxl,3-mannosyltransferase adds at least three mannose residues to the Kre2p N-glycosyl core. Mnnlp therefore contributes to most of the observed Kre2p post-ER modifications (Fig. 2) . GlcNAc2 carbohydrate structure. In other cases, core structures may be extended by an outer chain of variable size (up to 200 mannose residues) that is composed of a backbone of al,6-mannosyl residues to which are attached branched al,2-and cd,3-mannosyl side chains. Recent evidence suggests that the initiating Ochlp al,6-mannosyltransferase (Nakayama et al., 1992 ) defines a very early Golgi compartment that appears to be distinct from a subsequent early Golgi compartment responsible for cxl,6-mannosyl side chains elongation (Gaynor et al., 1994) . The ~1,2-and al,3mannosyl side chain modifications are predicted to occur in the medial-Golgi. The possible S. cerevisiae N-linked oligosaccharide structures are shown (adapted from Ballou, 1990 and Ballou et al., 1990) . Arrows depict [31,4, al,6, cd,2, and cd,3 linkages between mannoses of the core and outer chain. X = 10 on average. The mannose units not present in the mnn9 mutation are indicated. The Mnnlp-terminal al,3-mannosyltransferase is responsible for the addition of all al,3-mannosyl residues in a medial-Golgi compartment (Ballou, 1990; Graham et al., 1994; Yip et al., 1994) . [] represents mannose residues that are thought to be added to the Kre2p core oligosaccharide in a MNNl-dependent manner (see Fig. 3 ). Fully glycosylated proteins are then transported to a late Golgi compartment which includes the proteolytic enzymes (Kex2p, Kexlp, and DPAP A) responsible for maturation of secreted protein precursors. Glycoproteins that are not retained in the Golgi complex can be subsequently directed to the vacuole through an endosomal compartment or to the cell surface where they can be (a) incorporated into the plasma membrane, (b) secreted and retained in the periplasmic space/cell wall, or (c) secreted extracellularly. Both the function of Kre2p as a txl,2-mannosyltransferase and its slow receipt of MNNl-dependent modifications imply localization to a Golgi compartment. Kre2p subcellular localization was examined by indirect immunofluorescence and showed a punctate pattern of fluorescent signals indicative of a Golgi association (Fig. 4 C) . Between 3 and 14 structures per cell can be seen, depending on individual cells and the plane of focus. Punctiform fluorescence was never seen with the anti-Kre2p antibody in kre2:: TRP1 cells (Fig. 4 A) . The Kre2p signal distribution did not overlap with nuclei or mitochondria viewed by D N A staining with DAPI ( Fig. 4 D) . To attempt to define in which Golgi compartment Kre2p is localized, its intracellular localization was compared with that of a known medial-Golgi protein, the mannosyltransferase Mnnlp (Ballou, 1990; Graham et al., 1994; Yip et al., 1994) , by double-label immunofluorescence (Fig. 5) . Quantitative evaluation of punctate spot distribution (described in Materials and Methods) revealed that 75% of the Kre2p corresponding Texas red-labeled structures coincided with the FITC-labeled Mnnlp fluorescence spots. The location of Kre2p was also compared with that of a late Golgi protein, the carboxypeptidase Kexlp (see Figs. 2 and 5; Dmochowska et al., 1987; Bussey, 1989, 1992) . A quantitative scoring of punctiform fluorescence distribution indicated that 65% of the Kre2p corresponding structures did not coincide with the Kexlp fluorescence patches. To examine the basis of localization of a yeast Golgi glycosyltransferase, an analysis of Kre2p noncatalytic domains was made. The roles of the cytoplasmic NH2 terminus, TMD, and luminal stem region of Kre2p in Golgi targeting were tested by constructing chimeric proteins in which Kre2p-specific segments were substituted with the corresponding domains of the DAP2 or PH08 gene products (Fig. 6) . DAP2 encodes the vacuolar dipeptidyl aminopeptidase B (Roberts et al., 1989) , DPAP B, and is a type II integral membrane glycoprotein that lacks apparent vacuolar targeting information. No individual domain of DPAP B was shown to be required for its transport to the vacuole besides a nonspecific hydrophobic TMD (Roberts et al., 1992) . PH08 encodes a vacuolar alkaline phosphatase. It is also a type II membrane protein (Kaneko et al., 1987; Klionsky and Emr, 1989; Nothwehr et al., 1993) that is thought to be transported to the vacuole by default (Nothwehr et al., 1993) . Substitutions designed with vacuolar proteins were used to avoid potential problems with cryptic Golgi-targeting sequences. Indirect immunofluorescence detection of chimeric proteins was undertaken to determine their intracellular location(s). Initially, the role of the membrane-spanning domain of Kre2p was assessed with fusion protein KDKK (Fig. 6) , in which the Kre2p TMD was removed and substituted with that of DPAP B. KDKK was found, like wild-type Kre2p, to be localized to small punctate structures of the yeast Golgi (Fig. 7) . To exclude the possibility that the DPAP B TMD is able to function as a Kre2p TMD in the context of chimeric protein KDKK by sharing some unknown common features not apparent at the amino acid level, a chimeric protein consisting of Kre2p with the Pho8p membrane-spanning domain was made. As was the case with KDKK, KPKK was also found to be localized to the Golgi Hill et al., 1986) and the influenza hemagglutinin virus epitope-tagged MNN1 on multicopy plasmid YEp352 (URA3; Hill et al., 1986) or epitope-tagged KEX1 gene (see Materials and Methods) under an ADH1 gene promoter on centromeric-based plasmid pRS316 (URA3; Sikorski and Hieter, 1989) were fixed, spheroplasted, attached to polylysine-coated glass slides, and then incubated with affinity-purified anti-Kre2p and 12CA5 mAb (Kolodziej and Young, 1991) . Fluorescence signals were obtained by subsequent incubation of treated cells with Texas red-conjugated goat anti-rabbit IgG and FITC-conjugated goat antimouse IgG which were used as secondary antibodies. Arrows indicate colocalization of punctiform fluorescence. apparatus (Fig. 7) indicating that the Kre2p TMD is dispensable for Golgi retention. The function of the cytoplasmic domain of Kre2p was examined with chimeric protein D K K K which possesses the cytoplasmic NH2 terminus of Kre2p replaced by that of D P A P B. D K K K could not be visualized in SEY6210 (data not shown) suggesting possible degradation of this protein in the yeast vacuole as found with other mistargeted Golgi proteins (Cooper and Bussey, 1992; Roberts et al., 1992; Wilcox et al., 1992; Nothwehr et al., 1993) . To Figure 6 . Schematic representation of chimeric proteins used in the analysis of Kre2p Golgi targeting. The Kre2 protein is composed of 442 amino acid residues. The length of the different Kre2p domains are as follows: NH2 terminus, 11 residues; TMD, 19 residues; stem, ~87 residues and the catalytic domain is ~325 amino acid residues long. DPAP B is an 841-amino acid residue type II vacuolar membrane protein with cytoplasmic NH2-terminal and membrane-spanning domains of 29 and 16 amino acid residues, respectively, and a luminal domain consisting of 796 amino acid residues. The vacuolar alkaline phosphatase Pho8p is a 566-amino acid residue type II membrane protein. The length of the different Pho8p domains are as follows: NHz terminus, 33 residues; TMD, 26 residues; and the luminal domain is 507 amino acid residues long. The different chimeric constructions were made by replacing or fusing specific Kre2p domains with those of vacuolar DPAP B and Pho8p. Chimeric proteins are represented (not to scale) as an assembly of different domains which are distinguished by a letter specifying their origin. K, D, and P, respectively, denote Kre2p, DPAP B, and Pho8p. MR represents the two remaining amino acid residues in the cytoplasmic NH2 terminus of Kre2p after truncation. Putative N-glycosylation sites are indicated (Y). For details on constructions, see Materials and Methods. test this possibility, D K K K was expressed in strain $86 which contains mutations in the major vacuolar hydrolases and was found to be exclusively mislocalized to the yeast vacuole as shown by colocalization with a vacuolar membrane marker (Fig. 8) . In contrast, when wild-type Kre2p was expressed from a 2p~-based plasmid in strain $86, all positive cells showed a punctate pattern of fluorescence. However, 15% of these stained cells also showed fluorescence associated with the vacuole (data not shown). These vacuolar stained cells appeared to be expressing high levels of Kre2p as they displayed high fluorescence levels. The Kre2p cytoplasmic tail NH2 terminus was deleted to further explore its role in targeting. Chimeric construct MR/KKK lacks the Kre2p cytoplasmic tail, with the exception of the initiating methionine codon and the Arg residue lying directly adjacent to the TMD. MR/KKK was Figure 7 . Golgi localization of chimeric proteins KDKK, KPKK, and KD-K by indirect immunofluorescence. Yeast ceils (SEY6210) carrying a KRE2 disruption (kre2::TRP1) and containing the different chimeric constructions on YEp352 were fixed, spheroplasted, attached to polylysine-coated glass slides, and incubated with anti-Kre2p antibodies followed by a subsequent incubation with a Texas red-conjugated secondary Ab and DAPI. found to be mislocalized to the vacuole (Fig. 8) , again demonstrating that the cytoplasmic NH2-terminal domain of Kre2p is required for correct Golgi localization. Some ER retention of this MR/KKK protein was seen in ~10% of positive cells, indicative of an increased transit time through this organelle. The role of the luminal amino acid residues flanking the Kre2p TMD was examined with fusion protein KD-K. It comprises the Kre2p NH2 terminus, the membrane-spanning domain of DPAP B, the Kre2p stem region lacking the first 36 amino acid residues following the TMD, and the catalytic domain of Kre2p. KD-K was found to be localized in the Golgi complex (Fig. 7) suggesting that neither the deleted stem region nor the TMD is necessary for Golgi retention in the context of a Kre2p catalytic domain. In parallel with the localization studies, all chimeric proteins were assessed functionally. They were verified to be membrane associated and properly glycosylated (data not shown), showing that they possess a type II membrane orientation (see Fig. 6 ). In addition, their in vitro and in vivo enzymatic activities were assayed since low levels of mannosyltransferase activity would imply abnormal secondary structure of a particular chimera and possible mislocaliza-tion in the secretory pathway. The in vitro cd,2-mannosyltransferase activity was assayed in cell-membrane fractions. The specific activity of each chimeric protein is displayed in Fig. 9 , where KRE2 expressed in a kre2A strain was arbitrarily determined to be 100%, a value just slightly lower than that of wild-type mannosyltransferase activity from a genomic copy of KRE2. KDKK, KPKK, DKKK, as well as MR/KKK appear to be fully active. KD-K displays intermediate levels of mannosyltransferase activity suggesting that the stem region from position 31 to 66 is necessary for folding or proper catalytic activity. The capacity of chimeric proteins to function in vivo in Golgi-localized mannosylation was assessed by using a K1 killer toxin sensitivity assay (Fig. 9 ). K1 killer yeast strains secrete a small pore-forming toxin that requires a cell-wall receptor for function (Bussey, 1991) . KRE2 null mutations lead to shorter mannose chains on cell-wall mannoproreins disrupting the toxin receptor and leading to resistance (Hill et al., 1992; H/~usler et al., 1992) . Yeast cells containing different chimeric proteins were assayed for killer toxin sensitivity and comparisons were made between a sensitive wild-type strain, a resistant kre2 null strain, and the null strain containing plasmids bearing wild-type KRE2 or different chimeric constructions. As can be seen in Fig. 9 , the kre2 null mutant is toxin resistant, showing no killing zone, but the same strain containing the KRE2 gene, or expressing KDKK or KPKK displays a large clear killing zone of ,-~20 mm similar to wild-type cells consistent with correct Golgi localization. Chimeric protein KD-K has also been shown to be localized to the Golgi complex but has a reduced zone size (15 mm) likely a consequence of its reduced mannosyltransferase activity. The reduced killing zone (10 mm) of cells expressing DKKK which has wild-type enzymatic activity in vitro indicates that it is not correctly retained in the Kre2p Golgi subcompartment. MR/KKK has also been shown to be localized to the vacuole, but displays an almost wild-type zone size (17.5 mm). Its catalytic domain appears to have a normal conformation as it possesses wild-type mannosyltransferase activity, but it is partially retained in the ER (see above). Thus, MR/KKK while slowly passing through the secretory pathway en route to the vacuole, is likely able to correctly mannosylate cell-wall proteins. Finally, these results indicate that for Kre2p the default compartment for mislocalization is the vacuole, a conclusion previously reached for late Golgi membrane proteins (Cooper and Bussey, 1992; Roberts et al., 1992; Wilcox et al., 1992; Nothwehr et al., 1993) and some ER membrane proteins (Gaynor et al., 1994; Hill and Stevens, 1994) . To determine which regions of Kre2p are sufficient to target a reporter protein to the Golgi complex, chimeric proteins KKP and KKKP were constructed (see Fig. 6 and Materials and Methods for details). KKP consists of the Kre2p cytoplasmic and membrane-spanning domains fused to the Pho8p luminal domain. KKKP contains the Kre2p NH2 terminus, TMD, and a segment encompassing the first 36 amino acid residues of the stem domain fused to and containing the different chimeric constructions on YEp352 were fixed, spheroplasted, attached to polylysine-coated glass slides, and incubated with anti-Kre2p antibodies which was followed by a subsequent incubation with a FITC-coupled anti-rabbit secondary Ab and DAPI. Colocalization of vacuolar localized chimeric proteins with the yeast vacuolar membrane H÷-ATPase was achieved by coincubation of fixed cells with mAb 13Dll (Kane et al., 1992) with subsequent incubation of treated cells with Texas red-conjugated anti-mouse secondary antibodies. the Pho8p luminal domain. Both chimeric proteins were first verified to be membrane associated and properly glycosylated (data not shown), indicative of a correct type II membrane orientation (Fig. 6) . Chimeric protein KKKP was found, like wild-type Kre2p, to be typically localized to punctiform Golgi structures (Fig. 10) . However, when the partial stem domain was removed, the majority of the resulting chimeric protein (KKP) was found to be localized in the vacuole. The intracellular localization of chimeric protein KKP was quantitatively scored (see Materials and Methods); 70% of positive cells displayed a vacuolar signal and 26% of the cells had vacuolar and punctate fluorescent signals (typical results are shown in Fig. 10) , whereas 4% showed exclusively punctiform fluorescence. Thus, the Kre2 protein region encompassing the NH2 terminus, TMD, and partial stem domain is sufficient for full Golgi retention of a vacuolar protein. These results were extended biochemically by examining the processing kinetics of both chimeric proteins. The PhoSp alkaline phosphatase is activated in the vacuole (Klionsky and Emr, 1989; Nothwehr et al., 1993) where the catalytic luminal domain is processed by the proteinase A (Ammerer et al., 1986; Woolford et al., 1986) . KKKP was found not to be cleaved, whereas in contrast, the majority of KKP was processed in accord with its mostly vacuolar location (data not shown). Lastly, to assess if the Kre2p TMD is required for proper Golgi retention in the context of an heterologous luminal domain, we constructed fusion protein KPKP in which the Kre2p TMD of KKKP was removed and substituted with that of PhoSp (see Fig. 6 ). Golgi retention of this fusion protein would indicate that the Kre2p NH2 terminus cytoplasmic and stem domains are sufficient to achieve correct Golgi intracellular localization. Immunofluorescence studies of KPKP revealed that this chimeric protein was retained in the ER (data not shown), and thus uninformative. Kre2p is part of the yeast glycosylation machinery and was anticipated to be localized in a medial-Golgi compartment where N-and O-modified proteins receive al,2-1inked mannose residues (Fig. 2) . Evidence presented here is consistent with this expectation. We have demonstrated that in situ detection of Kre2p by indirect immunofluorescence reveals intracellular punctiform staining typical of the yeast Golgi, in agreement with Chapman and Munro (1994) . Kre2p is an al,2-mannosyltransferase that receives an Mnnlp-dependent ctl,3 modification, the next sequential step in the orderly glycosylation of yeast proteins. Thus, Kre2p must reach a Golgi compartment harboring the Mnnlp cd,3-mannosyltransferase (Fig. 2) . These ed,2and al,3-1inked activities could be colocalized, or, if they are situated in two distinct and consecutive cisternae, Kre2p would have to reach the Mnnlp compartment and then be retrieved to its own preceding organelle. Kre2p and Mnnlp have been shown to colocalize to a considerable extent, placing Kre2p in or close to the Mnnlp medial-Golgi compartment. Consistent with this assignment, previous biochemical studies provide evidence that Kre2p is localized in a compartment distinct from the late Golgi. Immunoisolated late Golgi organelles containing the endoproteinases Kexlp, Kex2p, and DPAP A were shown to be devoid of Kre2p (Bryant and Boyd, 1993) . Our immunocytochemical colocalization results also indicate that Kre2p is not in the same compartment as Kexlp. The 35% of Kre2p and Kexlp punctiform fluorescence signals that do colocalize could be due to two or more stacked cisternae that are seen in ~40% of all cisternae in a given yeast cell (Preuss et al., 1992) . However, the possibility remains that proteins from the medial-Golgi (Kre2p) and from the late Kre2,~ 31% Figure 9 . Activity of chimeric proteins. Enzymatic activity was measured in vitro by ed,2-mannosyltransferase assays of membrane preparations, and in vivo mannosylation of cell-wall proteins was assessed by verifying K1 killer toxin sensitivity in a wild-type strain and a kre2 null mutant containing KRE2 and chimeric constructs expressed from the centromeric-based plasmid pRS316 (Sikorski and Hieter, 1989) . Enzymatic activity is expressed as percentages of specific activity (mmot/h/mg) for each chimeric protein. A kre2 null strain expressing KRE2 from pRS316 arbitrarily represents 100% specific activity. Concentrated K1 killer toxin was spotted on a lawn of ~,106/ml cells from a fresh culture of each strain (see Materials and Methods). Golgi (Kexlp) are not always in mutually exclusive cisternae and their intracellular distribution may partially overlap. This can be the case in mammals where a pair of trans-Golgi and medial-Golgi glycosyltransferases showed substantial overlap (Nilsson et al., 1993) . The Kre2p/Mnnlp compartment remains to be precisely defined but some evidence suggests it could be identical to, or overlap with, compartments containing the Golgi type II membrane N-glycoprotein guanosine diphosphatase, Gdalp (Abeijon et al., 1989 (Abeijon et al., , 1993 . Subcellular fractionation experiments indicated that Gdalp and Kre2p activities cosediment (Abeijon et al., 1989) , whereas Gdalp and the late Golgi marker, Kex2p, are in distinct compartments (Bowser and Novick, 1991) . Mnnlp also cosediments with Gdalp and was found in fractions different from those containing Kex2p . These results are in accord with our immunofluorescence data indicating that Kre2p and Mnnlp are in the same Golgi compartment. The apparent double role of Kre2p in O-and N-linked glycosylation Hill et al., 1992; Lussier et al., 1995b) is most simply explained if both types of modification are carried out in the same Golgi compartments. The observations that Mnnlp also acts on O-and N-linked chains (Ballou, 1990; Herscovics and Orlean, 1993) , and that the O-glycosyltransferase Kre2p is itself N-glycosylated, are consistent with a common set of compartments for these modifications. However, more complex scenarios could be envisaged with specific subcompartments for O-and N-glycosylation. Kre2p is predicted to act in a functionally and genetically defined compartment where various cd,2-mannosyl residues are added to oligosaccharide chains. This compartment(s) is also likely to include other ~l,2-mannosyltransferases responsible for N-linked outer chain elaboration (Fig. 2) . A family of putative Kre2p-like mannosyltransferases has been identified (Hill et al., 1992; Bussereau et al., 1993; Lussier et al., 1993; Mallet et al., 1994) , and some members have been shown bioehemically to be Golgi mannosyltransferases (Lussier et al., 1995b) . Immunolocalization results from the different chimeric proteins demonstrate that the Kre2p NH2-terminal cytoplasmic domain was required for localization to the Golgi complex. Chimeric proteins lacking the 11-amino acid residue cytoplasmic domain are fully active as a mannosyltransferase in vitro but fail to be properly retained and are mislocalized to the vacuole. A trivial explanation for the localization of D K K K to the vacuole is that the cytoplasmic domain of DPAP B contains a cryptic vacuolar targeting signal that overrides the Golgi-targeting sequence of Kre2p. Arguing against this view is the demonstration that the NH2-terminal domain of DPAP B is not necessary for vacuolar targeting (Roberts et al., 1992; Nothwehr et al., 1993) . To ensure that mislocalization of DKKK was not caused by some previously unrecognized vacuolar targeting signal of the NH2-terminal domain of DPAP B, chimeric protein MR/KKK was constructed in which the Kre2p cytoplasmic tail was deleted. As with DKKK, MR/ KKK was mislocalized to the vacuole, showing that the NH2-terminal domain of Kre2p is necessary for proper Golgi targeting. Chimeric proteins K D K K and KPKK could not be distinguished from wild-type Kre2p, permitting the conclusion that the Kre2p TMD in the context of a native protein is not necessary for correct Golgi localization. Chimeric construct KD-K which comprises the Kre2p cytoplasmic tail, the DPAP B membrane-spanning domain, a Kre2p partial stem region, and the Kre2 protein mannosyltransferase domain was correctly localized to the Golgi complex. This indicates that the first 36 amino acid residues of the stem region of Kre2p are dispensable for Golgi targeting. The results obtained with this fusion protein again show that the Kre2p TMD is not required for Golgi retention. The fact that in the KD-K construct, the Kre2p TMD and part of the stem region are not present, argues that in the context of a protein containing a Kre2p luminal catalytic domain the cytoplasmic tail of Kre2p is sufficient to correctly target such a chimera to the Golgi complex, un- Figure I0 . Localization of chimeric proteins KKKP and KKP. Yeast cells carrying a PH08 disruption (pho8::TRP1) in a SEY6210 background and containing the different chimeric constructions on YEp352 were fixed, spheroplasted, attached to polylysine-coated glass slides, and incubated with anti-Pho8p antibodies followed by a subsequent incubation with a FITCcoupled anti-rabbit secondary Ab and DAPI. Colocalization of vacuolar localized chimeric proteins with the yeast vacuolar membrane H+-ATPase was achieved by coincubation of fixed cells along with mAb 13Dll (Kane et al., 1992) followed by incubation of treated cells with Texas red-conjugated anti-mouse secondary antibodies. Arrows point to Golgi punctiform structures. less some remaining luminal sequence also plays some role in retention. To analyze further the targeting role of the three nonenzymatic domains of Kre2p and to remove the possible additional complexity of a targeting mechanism involving the mannosyltransferase portion of Kre2p, combinations of noncatalytic domains of Kre2p were tested to assess what region of Kre2p was sufficient to target a Pho8p reporter protein to the Golgi complex. As opposed to the results obtained with fusion protein KD-K, all three noncatalytic domains of the Kre2 protein were found to be required for full Golgi retention of the Pho8p luminal region (KKKP). These results are in agreement with those of Chapman and Munro (1994) who found that a fusion protein containing the NHz terminus, the membrane-spanning domain, and a partial stem region of Kre2p linked to a reporter protein was retained in the Golgi complex. The mostly vacuolar localization of KKP demonstrated that the first part of the Kre2p stem region which is not required for retention in the context of a Kre2p catalytic domain (KD-K), is necessary, in combination with the Kre2p cytoplasmic tail and TMD for the targeting of the Pho8p catalytic portion to the Golgi complex (KKKP). Taken together, the intracellular localizations of chimeric proteins KD-K, KKP, and KKKP suggest that the Kre2p luminal domain does play a role in Golgi localization. Data obtained with chimeric proteins KDKK, KPKK, KD-K, and KKP clearly show that the Kre2p TMD is dispensable for Golgi retention and not sufficient to retain a reporter protein in the yeast Golgi complex. Our results are, however, at variance with conclusions reached in a recent study of Chapman and Munro (1994) and we address this below. In a first set of experiments, Chapman and Munro (1994) tested the role of the Kre2p TMD with a chimeric protein consisting of a full-length fusion of Kre2p/Mntlp linked to the SUC2-encoded invertase by a small region containing a Kex2p-dependent cleavage site (chimeric protein MMMI). The mislocalization of this protein was expected to result in production of soluble invertase when passing through the Kex2p-containing late Golgi compartment. This fusion protein was inferred to be in the Golgi complex as no invertase was secreted from the cell. When the TMD of Kre2p was replaced with that of Pho8p (MPMI) in this protein, it appeared to be mislocalized to some extent, as some invertase was secreted from the cell. This result was interpreted to indicate a role for the Kre2p TMD in Golgi retention. However, the intracellular localization of this chimeric protein was not visualized by immunofluorescence. We replaced the Kre2p TMD in the native protein with two different nonrelated TMD's, and showed that both proteins (KDKK and KPKK) were Golgi localized by using indirect immunofluorescence microscopy. In addition, we tested that the proteins had the right conformation by measuring their glycosylation in vivo, and their mannosyltransferase activity in vitro and in vivo. These proteins (KDKK and KPKK) behaved phenotypically as wild-type Kre2p in our tests with no indication that the Kre2p TMD in the context of a native protein is necessary for proper retention. A possible explanation of the Chapman and Munro (1994) results is that the Kre2p TMD is somehow necessary for Golgi retention in the context of those large heterologous proteins perhaps because some retention property of the luminal domain of Kre2p has been disrupted in the invertase fusion. Unfortunately, the ER retention of chimeric protein KPKP did not permit resolution of this question. Second, Chapman and Munro (1994) reported that a Pho8 chimeric protein containing a Kre2p TMD (chimeric protein PKP) slowed vacuolar processing of the luminal portion of Pho8p. They inferred from their results that PKP was at least partially localized in the Golgi complex and was retained by the Kre2p TMD. Again, the PKP chimeric protein was not localized intracellularly. This is clearly different from our results with Pho8p chimeras, which show that the Kre2p cytoplasmic NHz-terminal and transmembrane domains (KKP) are not sufficient for proper Golgi retention and the presence of the Kre2p cytoplasmic tail, TMD, and stem region is required for full retention of a Pho8p reporter protein (KKKP) in the yeast Golgi. To resolve this apparent inconsistency, we obtained the PKP construct and, on examination of its cellular location, found that it was exclusively localized to the vacuole displaying no punctiform fluorescence (data not shown). This finding is entirely consistent with our localization results with KKP, and offers no support for the Chapman and Munro conclusion that the Kre2p TMD will at least partially retain a heterologous protein in the yeast Golgi. The membrane-spanning domain plays a crucial role in the localization of resident animal Golgi glycosyltransferases. In contrast to Kre2p, alterations of the TMD of a certain number of mammalian glycosyltransferases results in some mislocalization of the modified protein (Munro, 1991; Aoki et al., 1992; Teasdale et al., 1992; Burke et al., 1994; Gleeson et al., 1994) . The hydrophobic TMD of several mammalian glycosyltransferases appears to be both necessary and sufficient to confer complete Golgi retention (Nilsson et al., 1991; Aoki et al., 1992; Shaper and Shaper, 1992; Teasdale et al., 1992; Gleeson et al., 1994; Low and Hong, 1994) whereas in other cases, some additional flanking sequences are required to achieve effective Golgi targeting (Munro, 1991; Burke et al., 1992; Shaper and Shaper, 1992; Dahdal et al., 1993; Burke et al., 1994; Gleeson et al., 1994; Low and Hong, 1994) . The exact nature of these NHz-terminal and stem sequences remains the be precisely determined but do not appear to involve the whole of each domain but mainly the short stretches of hydrophilic residues directly flanking the lipid-embedded TMD (Munro, 1991; Burke et al., 1992; Shaper and Shaper, 1992; Dahdal et al., 1993; Gleeson et al., 1994; Low and Hong, 1994) . It has been postulated that the targeting of glycosyltransferases carrying a TMD sorting signal could be due to interactions between the membrane-spanning domain and compartment specific membrane lipids (Pelham and Munro, 1993) . There is no evidence that Kre2p is retained by such a mechanism, since in the series of constructs we have devised, the Kre2p TMD is not implicated in retention. In this lipid interaction model, the length of the membranespanning domain is important for the proper sorting of animal glycosyltransferases (Munro, 1991; Pelham and Munro, 1993) . In the case of KDKK the TMD of DPAP B is three amino acid residues shorter than the Kre2p TMD, and in the case of KPKK the TMD of Pho8p is seven amino acid residues longer than the Kre2p TMD, yet both are retained in the Golgi complex. None of the studied animal glycosylation enzymes appear to need an intact cytoplasmic domain to achieve Golgi retention (Munro, 1991; Aoki et al., 1992; Colley et al., 1992; Shaper and Shaper, 1992; Teasdale et al., 1992; Dahdal et al., 1993; Burke et al., 1994; Gleeson et al., 1994; Low and Hong, 1994) . Kre2p, thus, constitutes the first demonstration of an eukaryotic glycosyltransferase requiring a short cytoplasmic amino-terminal domain for correct intracellular localization. The 11-amino acid residue-targeting domain of Kre2p seemingly parallels that of the late yeast Golgi enzymes Kexlp, Kex2p and DPAP A (Cooper and Bussey, 1992; Wilcox et al., 1992; Nothwehr et al., 1993) . It was shown for DPAP A that a 10-amino acid region within the cytoplasmic domain was both required and sufficient for proper retention (Nothwehr et al., 1993) , and for Kex2p, a 27-amino acid region was found to be essential (Wilcox et al., 1992) . The accurate Golgi targeting of Kex2p and DPAP A is clathrin dependent and aromatic residues are thought necessary for this retention process (Wilcox et al., 1992; Nothwehr et al., 1993; Wilsbach and Payne, 1993) . While unrelated at a sequence level with Kex2p and DPAP A, the amino-terminal domain of Kre2p contains a single aromatic phenylalanine residue (MALFLS-KRLLR) which was mutated to an alanine, an alteration showing no effect on targeting and, therefore, providing no evidence for clathrin-based retention of Kre2p (Lussier, M., A.-M. Sdicu, T. Ketela, and H. Bussey, unpublished results) . The Mnnlp al,3-mannosyltransferase is the only other yeast glycosyltransferase where targeting has been examined. Mnnlp targeting appears different from that of Kre2p since a mutant Mnnlp lacking its NH2-terminal cytoplasmic tail is properly localized to the Golgi complex . Correct retention of Mnnlp is clathrin dependent but, contrary to Kex2p and DPAP A, does not seem to be mediated by a direct interaction through its cytoplasmic tail. The reason for the Kre2p requirement for its cytoplasmic tail remains unclear. It could be clathrin dependent but not involving an aromatic residue, or through some other process. The Gdalp type II Golgi membrane protein (Abeijon et al., 1989; Abeijon el al., 1993) has been shown to be properly localized in a strain lacking clathrin heavy chains (Seeger and Payne, 1992; Wilsbach and Payne, 1993) . The Mnnlp, Kre2p, and Gdalp enzymes appear to be in close proximity or even in the same Golgi compartment (this work; Abeijon et al., 1989; Graham et al., 1994) . This raises the possibility that specific Golgi membrane proteins showing a similar compartmental distribution may be retained by more than one mechanism. Our results indicate that although only the cytoplasmic NH2 terminus has been shown to be required for Golgi retention, no single domain is able to specify correct Golgi localization. A chimeric protein including a Kre2p cytoplasmic tail and enzymatic domain (KD-K) was properly targeted. In contrast, a luminal reporter protein could only be properly retained in the Golgi complex by the Kre2p cytoplasmic tail, plus the membrane-spanning and a 36amino acid residue stem domain (KKKP). The entire Kre2p cytoplasmic tail and membrane-spanning domain were not sufficient to retain Pho8p in the Golgi complex (KKP). Therefore, it appears that a combination of Kre2p topological domains is needed to achieve proper Golgi localization. The presence or requirement of more than one specific targeting signal in a given secretory pathway membraneassociated protein has been hinted at in a few cases recently. The TMD might not be the sole targeting motif present in particular animal glycosyltransferases as some cytoplasmic and luminal sequences appear to be involved in retention (Munro, 1991; Burke et al., 1992; Shaper and Shaper, 1992; Dahdal et al., 1993; Burke et al., 1994; Gleeson et al., 1994; Low and Hong, 1994) . The trans-Golgi network protein TGN38 appears to achieve proper Golgi localization using two nonoverlapping targeting domains: a tyrosine-based cytoplasmic retrieval signal and a retention signal found in its membrane spanning domain (Bos et al., 1993; Humphrey et al., 1993; Wong and Hong, 1993; Ponnambalam et al., 1994) . For the CGN Sed5p syntaxin, the intracellular localization is only partially specified by its TMD and an additional localization signal involving its cytoplasmic domain appears to be involved (Banfield et al., 1994) . Targeting studies on the p63 type II membrane protein indicate that the cytoplasmic, membrane-spanning and luminal domains are all necessary for proper ER-Golgi intermediate compartment localization (Hauri and Schweizer, 1992; Schweizer et al., 1993 Schweizer et al., , 1994 . Finally, the Wbpl type I membrane protein of the yeast ER possesses a cytoplasmic KKXX Golgi-to-ER retrieval motif. When this targeting motif is inactivated, Wbpl is still ER retained indicating that another targeting signal is present in the protein (Gaynor et al., 1994) . The fact that Kre2p requires more than one topological segment to achieve proper Golgi localization, emphasizes that there may be distinct, but not necessarily mutually exclusive, mechanisms of retention. Golgi proteins that possess a cytoplasmic targeting signal have been proposed to be sorted via a receptor-mediated mechanism (Wilsbach and Payne, 1993; Gleeson et al., 1994; Low and Hong, 1994) . Overexpression of the late Golgi yeast proteins Kexlp, Kex2p, and DPAP A leads to some vacuolar mislocalization, indicating saturation of the capacity of a receptor-based sorting process (Cooper and Bussey, 1992; Wilcox et al., 1992; Nothwehr et al., 1993; Wilsbach and Payne, 1993) . Kre2p overexpression also results in some mistargeting to the vacuole, and could be similarly explained. When wild-type Kre2p is expressed from a 21xbased multicopy vector in a background with mutations in the major vacuolar hydrolases, all positive cells display a punctiform pattern of fluorescence with 15 % of the stained cells also showing vacuolar fluorescence. This small percentage of doubly stained cells are likely expressing high levels of Kre2p as determined by the intensity of the fluorescence. In contrast, overexpression of animal glycosyltransferases does not bring about saturation of the retention mechanism (Munro, 1991; Aoki et al., 1992; Burke et al., 1992; Colley et al., 1992; Teasdale et al., 1992; Gleeson et al., 1994) . Interpreting our data in the simplest way, the Kre2p segment encompassing the NH 2 terminus along with the TMD and stem region could interact directly with a medial-Golgi localized receptor spanning both sides of the Golgi lipid bilayer. For Kre2p to be retained in the Golgi, an interaction between the putative receptor and the NH2-terminal domain is essential. Part of the presumed Kre2p receptor complex could be part of the Golgi extracisternal space matrix which in mammals was recently shown to bind medial-Golgi enzymes presumably through their cytoplasmic tails . Interestingly, examination of the cytoplasmic NHz-terminal domains of the six members of the KRE2 mannosyltransferase family (Lussier et al., 1993; Mallet et al., 1994) reveals that the Kre2p (MALFLSKRLLR) sequence resembles only that of Ktr4p (MRFLSKRILK; Mallet et al., 1994) where the sequence FLSKR(I/L)L(K/R) is conserved in both enzymes. This may imply a common recep-tor for the two proteins. Interaction of the presumed receptor protein with a chimeric construct lacking the Kre2p stem domain but including a reporter luminal domain (KKP) may be partial, and insufficient to retain the fusion protein in the Golgi complex. To achieve full Golgi retention of a reporter luminal protein, interaction of the postulated receptor with the three nonenzymatic domains would be required. In the case of KD-K, only partial interaction would occur with the putative receptor and the observed Golgi retention may now reveal an additional mechanism involving a segment of the Kre2p luminal region possibly by oligomerization/kin recognition that has been implicated in the retention of certain mammalian Golgi'membrane proteins. It has been proposed that protein oligomers are assembled through their TMD and/or luminal domains in a particular Golgi cisternae and because of their highorder structure are consequently excluded from entering forward-moving secretory vesicles (Weisz et al., 1993; Gleeson et al., 1994; Low and Hong, 1994; Nilsson et al., 1994; Schweizer et al., 1994; Yamaguchi and Fukuda, 1995) . Applying such a model to a functionally active Kre2 protein involves self-association or formation of heterooligomers between Kre2p and other medial-Golgi protein(s) either through the catalytic domain or possibly with some contiguous stem sequences retained in construct Overall, our results suggest that proper targeting of Kre2p in the yeast medial-Golgi may involve two different mechanisms. The saturability of the Kre2p retention system implies a receptor-based retention mechanism that could involve protein-protein interactions over a tail/TMD/ partial stem region. In addition, a second mechanism that may involve oligomerization would operate in some way through the Kre2p luminal domain. In conclusion, our observations on Kre2p targeting may point to differences in Golgi localization mechanisms between yeast and animal glycosyltransferases or could indicate that multiple retention or retrieval mechanisms are used to varying extents in both systems. Topography of glycosylation in yeast: characterization of GDP-mannose transport and lumenal guanosine diphosphatase activities in Golgi-like vesicles Guanosine diphosphatase is required for protein and sphingolipid glycosylation in the Golgi lumen of Saccharomyces cerevisiae PEP4 gene of Saccharomyces cerevisiae encodes proteinase A, a vacuolar enzyme required for processing of vacuolar precursors The yeast Ca2+-ATPase homologue, PMR1 is required for normal Golgi function and localizes in a novel Golgi-like distribution Golgi retention of a trans-Golgi membrane protein. Galactosyltransferase requires cysteine and histidine residues within the membrane-anchoring domain Isolation, characterization, and properties of Saccharomyces cerevisiae mnn mutants with nonconditional protein glycosylation defects Revision of the oligosaccharide structures of yeast carboxypeptidase Y Localization of Sed5, a putative vesicle targeting molecule, to the eis-Golgi network involves both its transmembrane and cytoplasmic domains Yeast KRE genes provide evidence for a pathway of cell wall [3-glucan assembly TGN38 is maintained in the trans-Golgi network by a tyrosine-containing motif in the cytoplasmic domain Secl5 protein, an essential component of the exocytic apparatus, is associated with the plasma membrane and with a soluble 19.5S particle The K1 killer toxin: molecular and genetic applications to secretion and cell surface assembly. In Molecular Genetics of Yeast: A Practical Approach Immunoisolation of Kex2p-containing organelles from yeast demonstrates colocalisation of three processing proteinases to a single Golgi compartment The transmembrane and flanking sequences of 131,2-N-acetylglucosaminyltransferase I specify medial-Golgi localization Medial-Golgi retention of N-acetylglucosaminyltransferase I A 12.8 kB segment on the right arm of chromosome II from Saccharomyces cerevisiae including part of the DUR12 gene, contains five putative new genes K1 killer toxin, a pore-forming protein from yeast The functioning of the yeast Golgi apparatus requires an ER protein encoded by ANP1, a member of a new family of genes affecting the secretory pathway The signal anchor and stem regions of the [3-galactoside ~x2,6-sialyltransferase may each act to localize the enzyme to the Golgi apparatus Characterization of the yeast KEX1 gene product: a carboxypeptidase involved in processing of secreted precursor proteins Yeast Kexlp is a Golgi-associated membrane protein: deletions in a cytoplasmic targeting domain result in mislocalization to the vacuolar membrane Specific sequences in the signal anchor of the 13-galactoside a-2,6-sialyltransferase are not essential for Golgi retention Yeast KEX1 gene encodes a putative protease with a carboxypeptidase B-like function involved in killer toxin and alpha factor precursor processing Functional compartments of the yeast Golgi apparatus are defined by the see7 mutation Localization of components involved in protein transport and processing through the yeast Golgi apparatus Signal-mediated retrieval of a membrane protein from the Golgi to the ER in yeast Studies on the transformation of intact yeast cells by the LiAc/SS-DNA/PEG procedure Targeting of proteins to the Golgi apparatus Compartmental organization of Golgi-specific protein modification and vacuolar protein sorting events defined in yeast sec18 (NSF) mutant Clathrin-dependent localization of a-1,3 mannosyltransferase to the Golgi complex of S. cerevisiae The ER-Golgi intermediate compartment Glycosylation in Saccharomyces cerevisiae: cloning and characterization of an a-l,2-mannosyltransferase structural gene Yeast glycoprotein biosynthesis: MNT1 encodes an c~-l,2-mannosyltransferase involved in O-glycosylation Glycoprotein biosynthesis in yeast Yeast/E. coli shuttle vectors with multiple unique restriction sites Yeast KRE2 defines a new gene family encoding probable secretory proteins, and is required for the correct N-glycosylation of proteins Vma21p is a yeast membrane protein retained in the endoplasmic reticulum by a di-lysine motif and is required for the assembly of the vacuolar H+-ATPase complex Localization of TGN38 to the trans-Golgi network: involvement of a cytoplasmic tyrosine--containiug sequence Assembly and targeting of peripheral and integral membrane subunits of the yeast vacuolar H+-ATPase Structural characteristics of the PH08 gene encoding repressible alkaline phosphatase in Saccharomyces cerevisiae The molecular and cell biology of glycosyltransferases Membrane protein sorting: biosynthesis, transport and processing of yeast vacuolar alkaline phosphatase. EMBO (Eur Mot Epitope tagging and protein surveillance Protein glycosylation in yeast Rapid and efficient sitespecific mutagenesis without phenotypic selection Separation and characterization of two al,2-mannosyltransferase activities from Saccharomyces cerevisiae Protein trafficking in and out of the Golgi apparatus KTR2: a new member of the KRE2 mannosyltransferase gene family Protein O-glycosylation in yeast. The PMT2 gene specifies a second protein O-mannosyltransferase that functions in addition to the PMTl-encoded activity Functional characterization of the KRE2/MNT1 mannosyltransferase gene family Nucleotide sequence analysis of an 11.7kB fragment of yeast chromosome I1 including BEM1, a new gene of the WD-40 repeat family and a new member of the KRE2/MNT1 family Sequences within and adjacent to the transmembrane segment of a-2,6-sialyltransferase specify Golgi retention OCHI encodes a novel membrane bound mannosyltransferase: outer chain elongation of asparagine-linked oligosaccharides. EMBO (Eur Mot The membrane spanning domain of 13-1,4-galactosyltransferase specifies trans-Golgi localization Overlapping distribution of two glycosyltransferases in the Golgi apparatus of HeLa cells Kin recognition between medial Golgi enzymes in HeLa cells Membrane protein retention in the yeast Golgi apparatus: dipeptidyl aminopeptidase A is retained by a cytoplasmic signal containing aromatic residues Sorting of membrane proteins in the secretory pathway The TGN38 glycoprotein contains two nonoverlapping signals that mediate localization to the trans-Golgi network Characterization of the Saccharomyces Golgi complex through the cell cycle by immunoelectron microscopy Vesicle-mediated protein sorting Molecular analysis of the yeast VPS3 gene and the role of its product in vacuolar protein sorting and vacuolar segregation during the cell cycle Immunolocalization of Kex2 protease identifies a putative late Golgi compartment in the yeast Saccharomyces cerevisiae Structure, biosynthesis, and localization of dipeptidyl aminopeptidase B, an integral membrane glycoprotein of the yeast vacuole Membrane protein sorting in the yeast secretory pathway: evidence that the vacuole may be the default compartment Targeting, disruption, replacement, and allele rescue: integrative DNA transformation in yeast Molecular Cloning: A Laboratory Manual DNA sequencing with chainterminating inhibitors Characterization of a novel 63 kDa membrane protein. Implications for the organization of the ER-to-Golgi pathway Retention of p63 in an ER-Golgi intermediate compartment depends on the presence of all three of its domains and on its ability to form oligomers Selective and immediate effects of clathrin heavy chain mutations on Golgi membrane protein retention in Saccharomyces cerevisiae Enzymes associated with glycosylation A system of shuttle vectors and yeast host strains designed for efficient manipulation of DNA in Saccharomyces cerevisiae Isolation of a matrix that binds medial-Golgi enzymes A Golgi retention signal in a membrane-spanning domain of coronavirus E1 protein The transmembrane domain of N-glucosaminyltransferase I contains a Golgi retention signal The signal for Golgi retention of bovine 131,4-galactosyltransferase is the transmembrane domain Oligomerization of a membrane protein correlates with its retention in the Golgi complex Mutation of a tyrosine localization signal in the cytosolic tail of Kex2p protease disrupts Golgi retention and results in default transport to the vacuole Dynamic retention of TGN membrane proteins in Saccharomyces cerevisiae The 17-residue transmembrane domain of 13-galactoside ct2,6-sialyltransferase is sufficient for Golgi retention The SXYQRL sequence in the cytoplasmic domain of TGN38 plays a major role in the trans-Golgi network localization The PEP4 gene encodes an aspartyl protease implicated in the posttranslational regulation of Saccharomyces cerevisiae vacuolar hydrolases Golgi retention mechanism of 13-1,4-galactosyltransferase Cloning and analysis of the Saccharomyces cerevisine MNN9 and MNN1 genes required for complex glycosylation of secreted proteins We thank Dr. Annette Herscovics for critically reading the manuscript, the members of the Bussey laboratory for helpful comments and suggestions, Antony Cooper and Kathryn Hill for all their help and advice, Tom Stevens for kindly providing strains, plasmids, and antibodies, Anne Camirand for advice with the mannosyltransferase assays, Rowan Chapman for discussion and plasmids, V. Mackay and C. Yip for the MNN1 clone, Carole Smith and Guy I'Heureux for photographic work, and Diane Oki for assistance with manuscript preparation.Supported by an Operating Grant from the Natural Sciences and Engineering Research Council of Canada. M. Lussier was the recipient of a postdoctoral fellowship from the Medical Research Council of Canada.Received for publication 12 May 1995 and in revised form 27 July 1995.