Physical interactions between the Alg1, Alg2, and Alg11 mannosyltransferases of the endoplasmic reticulum

Transcription

1 Glycobiology vol. 14 no. 6 pp. 559±570, 2004 DOI: /glycob/cwh072 Advance Access publication on March 24, 2004 Physical interactions between the Alg1, Alg2, and Alg11 mannosyltransferases of the endoplasmic reticulum Xiao-Dong Gao 2, Akiko Nishikawa 1, and Neta Dean 1 1 Department of Biochemistry and Cell Biology, Institute for Cell and Developmental Biology, State University of New York, Stony Brook, NY , and 2 Research Center for Glycoscience, National Institute of Advanced Industrial Science and Technology, Tsukuba Central 6, 1-1 Higashi, Tsukuba , Japan Received on January 26, 2004; revised on March 2, 2004; accepted on March 2, 2004 The early steps of N-linked glycosylation involve the synthesis of a lipid-linked oligosaccharide, Glc 3 Man 9 GlcNAc 2 -PPdolichol, on the endoplasmic reticulum (ER) membrane. Prior to its lumenal translocation and transfer to nascent glycoproteins, mannosylation of Man 5 GlcNAc 2 -PP-dolichol is catalyzed by the Alg1, Alg2, and Alg11 mannosyltransferases. We provide evidence for a physical interaction between these proteins. Using a combination of biochemical and genetic assays, two distinct complexes that contain multiple copies of Alg1 were identified. The two Alg1-containing complexes differ from one another in that one complex contains Alg2 and the other contains Alg11. Alg1 self-assembles through a C-terminal domain that is distinct from the region required for its association with Alg2 or Alg11. Missense mutations affecting catalysis but not Alg1 protein stability or assembly with Alg2 or Alg11 were also identified. Overexpression of these catalytically inactive alleles resulted in dominant negative phenotypes, providing genetic evidence for functional Alg1-containing complexes in vivo. These data suggest that an additional level of regulation that ensures the fidelity of complex oligosaccharide structures involves the physical association of the related catalytic enzymes in the ER membrane. Key words: ALG1, ALG2, ALG11/endoplasmic reticulum/ mannosyltransferase/n-glycosylation/s. cerevisiae Introduction Glycosylation is an essential modification important for protein folding and stability. The early steps that lead to synthesis of an asparagine (N )-linked glycan start in the encoplasmic reticulum (ER) and involve synthesis of a lipidlinked oligosaccharide precursor that is subsequently transferred to selected asparagines on nascent glycoproteins (for review see Herscovics and Orlean, 1993; Kornfeld and Kornfeld, 1985; Tanner and Lehle, 1987). Synthesis of this lipid-linked oligosaccharide, Glc 3 Man 9 GlcNAc 2 -PP-Dol, 1 To whom correspondence should be addressed; Neta.Dean@stonybrook.edu begins on the cytosolic face of the ER, where seven sugars (two N-acetylglucoseamines and five mannoses) are added sequentially to dolichyl phosphate on the outer leaflet of the ER, using nucleotide sugar donors (Abeijon and Hirschberg, 1992; Perez and Hirschberg, 1986; Snider and Rogers, 1984). After a ``flipping'' or translocation step, the last seven sugars (four mannoses and three glucoses) are added within the lumen of the ER, using dolichol-linked sugar donors (Burda and Aebi, 1999). Once assembled, the oligosaccharide is transferred from the lipid to nascent protein in a reaction catalyzed by oligosaccharyltransferase. After removal of terminal glucoses and a single mannose, nascent glycoproteins bearing the N-linked Man 8 GlcNAc 2 core can exit the ER to the Golgi, where this core may undergo further carbohydrate modifications. The structure and assembly pathway of the lipid-linked oligosaccharide has been highly conserved among almost all eukaryotes, as expected in light of its pivotal role during glycoprotein maturation and exit from the ER (Helenius and Aebi, 2001). Since the pioneering work of Robbins and colleagues, who initially isolated and characterized yeast mutants defective in lipid-linked oligosaccharide assembly (asparagine linked glycosylation or alg mutants) (Huffaker and Robbins, 1982, 1983), many of the genes affecting lipidlinked oligosaccharide synthesis in the ER have now been identified (for review see Burda and Aebi, 1999). With few exceptions, mutations in any one of the ALG genes result in accumulation of defined oligosaccharide intermediates. This phenotype is consistent with the notion that the lipidlinked oligosaccharide biosynthetic pathway involves a series of steps that lead to the synthesis of a branched and structurally complex oligosaccharide. Furthermore there is good evidence to support the hypothesis that this ordered assembly arises at least in part from differences in the substrate specificities of the various glycosyltransferases that participate in these reactions (Burda et al., 1999). During this ordered assembly, an important distinction can be made between the reactions that occur on the cytosolic face of the ER and those that occur in the lumen. A deletion of any gene affecting synthesis or translocation of Man 5 GlcNAc 2 -PP-Dol on the cytosolic face of the ER leads to death or an extremely severe phenotype. For instance, alg1d, alg2d, and alg11d mutants, required for addition of the first, third, and fourth or fifth mannose, respectively, of Man 5 GlcNAc 2 -P-P-Dol are inviable or grow very slowly (Albright and Robbins, 1990; Cipollo et al., 2001; Jackson et al., 1993). In contrast, mutations in genes affecting lumenal sugar additions result in little if any growth phenotypes (Burda et al., 1999). These observations suggest that although N-linked glycosylation is an essential Glycobiology vol. 14 no. 6 # Oxford University Press 2004; all rights reserved. 559

2 X.-D. Gao et al. modification, the cytosolic Man 5 GlcNAc 2 core structure is the minimal intermediate whose synthesis and translocation across the membrane is required for the successful execution of N-linked glycosylation. The Man 4 GlcNAc 2 core and earlier intermediates may be poor substrates for lumenal flipping or transfer to protein or, alternatively, for outer chain elongation by the Golgi mannosyltransferases. Despite the fact that many of the enzymes involved in lipid-linked oligosaccharide biosynthesis have been identified, the regulation and coordination of their activities is poorly understood. Here we show that three of the enzymes that participate in Man 5 GlcNAc 2 -PP-Dol synthesis, namely, Alg1, Alg2, and Alg11, exist in oligomeric protein complexes. Two distinct complexes that are distinguished by the presence of Alg2 or Alg11 can be identified, both of which contain Alg1. These results suggest that the physical association of the functionally related catalytic enzymes involved in lipid-linked oligosaccharide synthesis provides an additional level of regulation that assures proper construction of this complex oligosaccharide. Results The Alg1 protein exists in oligomeric complexes that contain Alg2 or Alg11 The Alg1 mannosyltransferase catalyzes addition of the first b1,4-linked ~ mannose on GlcNAc 2 -PP-Dol, producing ManGlcNAc 2 -PP-Dol (Couto et al., 1984). To investigate the physical properties of the Alg1 protein, a yeast strain was constructed whose chromosomal ALG1 locus was replaced with an allele that encodes an hemaglutinnin (HA)-tagged Alg1 protein. An ALG1-HA strain that solely produces Alg1-HA grows at rates comparable to the parental wild type and exhibits no detectable glycosylation abnormalities, suggesting that addition of the HA epitope to Alg1 does not alter its normal behavior (data not shown). To determine the native molecular weight of Alg1, we used gel filtration chromatography. Detergent extracts were prepared from this ALG1-HA strain, clarified by centrifugation at 100 kg, and fractionated over a Superose 12 column (see Materials and methods). Aliquots from each fraction were analyzed by sodium dodecyl sulfate±polyacrylamide gel electrophoresis (SDS±PAGE), followed by western blotting with anti-ha antibody to identify those fractions that contained Alg1-HA. Although its predicted molecular weight is 51.9 kda, Alg1-HA cofractionated broadly with protein markers between 150±400. The peak fractions containing Alg1-HA coeluted with markers of about 200± 300 kda, and only a small percentage of the total amount of Alg1 was detectable in fractions of a lower molecular weight (Figure 1). Though we considered that the interaction with detergent could alter the apparent molecular weight of Alg1 during gel filtration, together with the coimmunoprecipitation results presented later, these data suggested that the bulk of Alg1 exists in an oligomeric complex. Several observations prompted us to test the idea that Alg1 may interact with other mannosyltransferases that participate in lipid-linked oligosaccharide synthesis. First, synthesis of the lipid-linked oligosaccharide occurs very rapidly (Sharma et al., 2001). One explanation for these Apoferritin(443) α-amylase(200) BSA(66) 10 Fraction Alg1-HAp Fig. 1. Gel filtration chromatography of yeast extracts containing Alg1- HA. 1.0% digitonin extracts from an ALG1-HA yeast strain (XGY25) were prepared as described in Materials and methods. Samples were clarified by centrifugation at 100 kg and fractionated by Fast Performance Liquid Chromatography (FPLC) over a Superose 12 column. Aliquots of fractions were subjected to SDS±PAGE and western blotted with anti-ha antibodies. The panel above the western blot indicates the mobility of protein molecular weight markers run in parallel under identical conditions. rapid kinetics is that lipid-linked oligosaccharide biosynthesis is not a diffusion-limited process but rather involves the physical association of enzymes that catalyze each of the sequential mannose additions, which would in turn limit diffusion of the acceptor. Second, an increasingly large number of glycosyltransferases have been shown to function as homo- or hetero-oligomers (e.g., Jungmann and Munro, 1998; Marks et al., 1999; Sasai et al., 2001), suggesting a precedent for this mechanism in N-linked glycan synthesis. To explore this idea, we tested whether Alg1 interacts with any other mannosyltransferase known to catalyze lipid-linked oligosaccharide synthesis on the cytosolic face of the ER, namely, Alg2 and Alg11. Alg2 is required for addition of the third ~a1,6-linked mannose. Loss of ALG2 is lethal, and alg2 mutants accumulate Man 2 GlcNAc 2 -PP- Dol (Jackson et al., 1993). Alg11 is required for addition of the fourth and/or fifth ~a1,2-linked mannose to Man 3- GlcNAc 2 -PP-Dol. Loss of ALG11 results in a severe growth phenotype and temperature sensitive lethality (Cipollo et al., 2001). To determine if Alg2 or Alg11 interact with Alg1, we used a coimmunoprecipitation assay. Yeast strains were constructed that coexpressed different HA- and myc-tagged ALG genes. The chromosomal ALG1, ALG2, or ALG11 coding sequences were replaced with the corresponding myc-tagged alleles. In each of these strains, an additional copy of an HA-tagged triose phosphoisomerase (TPI) promoter-driven ALG1 gene was integrated in the chromosome (see Materials and methods). Strains solely expressing epitope-tagged ALG1, ALG2, or ALG11 display no obvious growth or glycosylation defects and each of the epitopetagged genes can complement the corresponding mutant allele, suggesting that these additional sequences do not affect normal protein function (data not shown). Extracts from each of these strains were prepared in buffer containing the nonionic detergent digitonin to maintain oligomeric interactions between membrane proteins, and these extracts were subjected to coimmunoprecipitation assays. Myc-tagged Alg1, Alg2, or Alg11 proteins were precipitated from these extracts with anti-myc antibody,

3 Interactions between the Alg1, Alg2, and Alg11 mannosyltransferases A Alg1-HA Alg11-myc Alg2-myc Alg1-myc ALG1-HA ALG2-HA ALG11-HA ALG1-HA Sec61 IP: Blot: α-myc α-ha and the immunoprecipitates were fractionated by SDS± PAGE. The relative steady state levels of myc-tagged Alg1, Alg2, and Alg11 proteins in the same extracts used for the immunoprecipitations were determined by western blot analysis of aliquots removed prior to immunoprecipitation and were found to be similar (Figure 2B). HA-tagged Alg1 that coprecipitated with myc-tagged Alg1, Alg2, or with Alg11 was detected by blotting with anti-ha antibody (Figure 2A). The result of this experiment demonstrated that Alg1 coprecipitated with itself, Alg2, and Alg11 (Figure 2A, lanes 1, 2, 3) at comparable levels. Similar amounts of Alg1-, Alg2-, and Alg11-myc were precipitated with anti-myc antibody, determined by stripping these blots of anti-ha antibody and reprobing with anti-myc antibody (Figure 2A, lanes 6±10). Coprecipitation of Alg1-HA occurred in the presence of several other nonionic α-myc α-myc α-ha α-sec61 Fig. 2. Alg1 coimmunoprecipitates with itself, Alg2, and Alg11. (A) Detergent extracts were prepared from yeast strains coexpressing ALG1-HA together with ALG11-myc (XGY35) (lanes 1 and 6), ALG1-myc (XGY33) (lane 2 and 7), ALG2-myc (XGY34) (lanes 3 and 8), or expressing just ALG1-HA (XGY25) (lanes 4, 5, 9, 10, 11, and 14), ALG2-HA (XGY26) (lane 12) or ALG11-HA (XGY27) (lane 13). Protein complexes were extracted from yeast cells by treatment with buffer containing 1.0% digitonin, clarified by centrifugation at 100 kg, and immunoprecipitated with anti-myc, anti-ha, or anti-wbp1 antibodies as indicated (IP). Precipitates were fractionated by 8% SDS±PAGE and immunoblotted with anti-ha or anti-sec61 antibodies to detect coprecipitated proteins. (B) An aliquot of the detergent extracts used for the coimmunoprecipitation assays were subjected to 8% SDS±PAGE and western blot analysis with anti-myc antibodies. detergents, including octylglucoside, Triton X-100, and CHAPS, though less efficiently than digitonin (not shown). Therefore we used buffers containing digitonin in all our further analyses of these protein complexes. Several controls demonstrated the specificity of these interactions. First, Alg1-HA did not coprecipitate in a control strain that does not coexpress ALG-myc tagged alleles (Figure 2A, lane 4) demonstrating that these interactions are dependent on the coexpression of Alg1 with Alg2 or Alg11. Second, we find no evidence that several other ER membrane proteins interact with Alg1, Alg2, or Alg11. These include Wbp1, an ER membrane protein that is a subunit of the oligosaccharyltransferase complex (Figure 2A, lane 5) and Sec61, a subunit of the ER translocon complex (Figure 2A, lanes 11±14), suggesting that the coprecipitation of Alg1 with Alg2 and Alg11 is not due to 561

4 X.-D. Gao et al. nonspecific hydrophobic interactions or aggregation of membrane proteins of the ER. Third, when extracts from strains expressing only HA-tagged Alg1 were mixed with extracts from strains expressing myc-tagged Alg2 or Alg11, no coprecipitation was observed, suggesting that these complexes are stable (data not shown). Fourth, these interactions are not an artifact due to protein overexpression because these immunoprecipitations were performed with proteins that are produced at physiological levels. Considered together, these results demonstrate that Alg1 is stably associated with itself, Alg2, and Alg11. Alg2 and Alg11 proteins do not interact with each other If Alg1, Alg2, and Alg11 proteins interact with one another in one large complex, as suggested by the results of the immunoprecipitation assay, all three proteins would be A Fraction Alg11p Alg2p Alg1p Load predicted to cofractionate when separated by gel filtration chromatography. To test this idea, digitonin extracts were prepared from a yeast strain (XGY32) that simultaneously expresses HA-tagged ALG1, ALG2, and ALG11 and size fractionated by FPLC. In this strain, each of the normal genes has been replaced by an epitope-tagged allele whose expression is controlled by the corresponding endogenous promoter. This strain grows normally even though it solely produces epitope-tagged Alg1, Alg2, and Alg1 proteins, further suggesting that their normal function is not altered by the presence of the C-terminal tags. Western blot analysis of extracts from this strain suggested that Alg1-HA, Alg2-HA, and Alg11-HA accumulated in a 2:1:1 ratio, respectively, demonstrating that at steady state, Alg1 is roughly twice as abundant as Alg2 or Alg11 (Figure 3, lane marked Load). Apoferritin(443) α-amylase(200) BSA(66) Fig. 3. Alg2 and Alg11 display distinct chromatographic profiles and do not interact with each other. (A) 1.0% digitonin extracts were prepared from a yeast strain coexpressing ALG1-HA, ALG2-HA, and ALG11-HA (XGY32). Samples were fractionated by FPLC over a Superose 12 column (see Materials and methods). Aliquots of fractions were subjected to 8% SDS±PAGE and western blotted with anti-ha antibodies. (B) Alg2 and Alg11 do not coprecipitate. Immunoprecipitations were performed as described in Materials and methods. Detergent extracts were prepared from strains coexpressing ALG2-myc and ALG1-HA (XGY23) (lanes 1 and 3) or ALG2-myc and ALG11-HA (XGY24) (lanes 2 and 4). Samples were either immunoprecipitated with anti-ha antibody, subjected to SDS±PAGE, and blotted with anti-myc antibody to detect coprecipitated myc-tagged protein (lanes 1 and 2) or directly fractionated by SDS±PAGE and subjected to western blot analysis with anti-ha antibodies to detect HA-tagged protein present in the extract (lanes 3 and 4). 562

5 Interactions between the Alg1, Alg2, and Alg11 mannosyltransferases Detergent extracts prepared from this strain were fractionated by gel filtration chromatography as described in Materials and methods. Each fraction was assayed for the presence of Alg1-HA, Alg2-HA, and Alg11-HA by western blot analysis with anti-ha antibody. As shown, Alg1 elutes broadly and the peak of Alg1, in fractions 27, 28, and 29, coincides with the peak of both Alg2 and Alg11 (Figure 3A). However, by this assay we found that the peak of fractions containing Alg11-HA (fraction 29) and Alg2-HA (fraction 28) did not precisely coincide with one another. Although its predicted molecular weight is 63 kda, the peak of Alg11-HA eluted with protein markers between 150 and 175. In contrast, the peak of Alg2-HA corresponded to a molecular weight of about 200 kda, although its predicted mass is 55.8 kda (Figure 3). The elution profiles of these two proteins were also distinct. The leading edge of fractions containing Alg2-HA leaned toward the larger molecular weight markers, upward of 400 kda. In contrast, the leading edge of fractions containing Alg11 leaned toward the lower-molecular-weight markers. Taken together, these data demonstrate that both Alg11 and Alg2 exist in oligomeric complexes, but these complexes do not display identical fractionation properties and may therefore be distinct from one another. A simple model to explain the different chromatographic behaviors of Alg2 and Alg11, as well as their stable interactions with Alg1, envisages two Alg1-containing complexes, one containing Alg2 and the other Alg11. If this model were correct, Alg2 and Alg11 would not be predicted to interact. To test this model, we used coimmunoprecipitation assays to examine the interaction between Alg2 and Alg11. Yeast strains were constructed that coexpressing myc-tagged ALG2 and HA-tagged ALG1 or ALG11(see Materials and methods). Digitonin extracts were prepared from these strains, Alg1-HA or Alg11-HA proteins were precipitated with anti-ha antibody, and precipitates were fractionated by SDS±PAGE and blotted with anti-myc to quantitate the amount of associated Alg2-myc. By this assay, we found no evidence for an interaction between Alg2 and Alg11, although, as shown, Alg2 associated stably with Alg1 (Figure 3B, lanes 1 and 2). This failure to detect an interaction between Alg2 and Alg11 was not due to a low level of Alg11-HA in these extracts. Western blot analysis of these extracts prior to immunoprecipitation demonstrated the levels of Alg11 and Alg1were comparable (Figure 3B, lanes 3 and 4). The lack of interaction between Alg2 and Alg11 observed by this coimmunoprecipitation assay was in agreement with their distinct chromatographic elution profiles (Figure 3). Together, these results suggest that both Alg2 and Alg11 interact with Alg1, but they do not interact with each other. Alg1, but not Alg2 and Alg11, forms homo-oligomers As a first step toward addressing the stoichiometry of each of these components in the oligomeric complexes, we examined the homomeric interactions between Alg1, Alg2, and Alg11 more closely. Specifically, we asked if Alg2 or Alg11 could self-assemble, as shown for Alg1 (Figure 2). To address this question, we used the immunoprecipitation assay, using extracts from yeast strains that coexpress HA- and IP: Western: anti-myc anti-ha anti-myc myc-tagged Alg2 or Alg11 proteins. Proteins in digitonin extracts prepared from these strains were precipitated with anti-myc antibodies and fractionated by SDS±PAGE; the ability of each of these proteins to homo-oligomerize was assayed by western blotting with anti-ha antibodies (Figure 4). By this assay, we found that although Alg1 homo-oligomerizes very efficiently, neither Alg2 nor Alg11 dimerize. The amounts of Alg11-HA or Alg2-HA that selfassembled were undetectable (Figure 4, compare lanes 1 with 2 and 3), despite the fact that western blot analysis demonstrated that Alg1, Alg2, and Alg11 proteins were found at similar levels in these extracts (Figure 4, lanes 4, 5, 6). These results demonstrate that although Alg1efficiently homooligomerizes, neither Alg2 and Alg11 can self-assemble. Identification of a C-terminal Alg1 domain required for homodimerization The data presented demonstrate that at least two distinct protein complexes share Alg1 as common member. To provide evidence for the biological relevance of these interactions, we sought mutations in ALG1 that differentially affect its ability to maintain protein±protein interactions or catalyze sugar transfer. The former types of mutations could help identify Alg1 protein-binding domains, whereas overexpression of the latter are predicted to interfere with mannosyltransferase activity in a dominant manner if these oligomeric interactions do indeed occur in vivo. The results described next document the existence of both such mutant alleles. - Alg11 Alg2 Alg1 Fig. 4. Alg1, but not Alg2 or Alg11, efficiently self-assembles. Detergent extracts prepared from yeast strains coexpressing both HA- and myctagged ALG1 (XGY20), ALG2 (XGY21), or ALG11 (XGY22) were immunoprecipitated with anti-myc antibodies, fractionated by SDS±PAGE, and analyzed by western blotting with anti-ha to detect coprecipitated HA-tagged protein (left) or directly fractionated by SDS±PAGEandsubjectedtowesternblotanalysiswithanti-mycantibodies to compare the level of myc-tagged protein present in the extract (right). 563

6 X.-D. Gao et al. Fig. 5. The C-terminus of Alg1 is required for homomeric but not heteromeric protein interactions. (A) Schematic diagram of Alg1, its predicted topology in the ER membrane, and the location of amino acids affected by mutations that were analyzed in this study. The alignment of this region from a small subset of PFAM GT1 members is shown. The amino acids of ALG1 that were altered after mutagenesis are highlighted by boldface. (B) Extracts were prepared from yeast expressing mutant alg1-1-ha (XGY30) or alg1-1-ha together with wild type ALG1-myc (XGY27), ALG2-myc (XGY28), or ALG11-myc (XGY29) and immunoprecipitated with anti-myc antibody. Precipitates were fractionated by 8% SDS±PAGE and subjected to western blot analysis with anti-ha antibody. (C) Aliquots of extracts containing mutant or wild type HA-tagged Alg1 protein, used in the immunoprecipitations shown in lanes 1 and 3 of B were analyzed directly by western blot with anti-ha antibodies. alg1-1 was identified previously as a temperature sensitive allele (Huffaker and Robbins, 1982). To determine the molecular basis for this effect, the alg1-1 locus was isolated by polymerase chain reaction (PCR) of genomic DNA from a mutant strain carrying this allele (see Materials and 564 methods). Sequence analysis demonstrated the presence of a single opal nonsense mutation affecting W434, resulting in the truncation of the C-terminal 16 amino acids of Alg1 in a domain predicted by several algorithms to face the cytoplasm (Figure 5A).

7 Interactions between the Alg1, Alg2, and Alg11 mannosyltransferases To determine if this C-terminal tail affects the ability of Alg1 to engage in protein±protein interactions, we constructed yeast strains that coexpress an HA-tagged alg1-1 allele with myc-tagged alleles of ALG1, ALG2, and ALG11 and used coimmunoprecipitation assays to compare the interactions among these proteins. This results of this experiment suggested that this C-terminal domain has a marked involvement in Alg1 protein±protein interactions. Loss of this C-terminal 16-amino-acid tail had a modest effect on the interaction between the truncated Alg1-1 protein and either Alg2 or Alg11 (Figure 5B, lanes 6 and 8). In contrast, homo-oligomerization with the full-length, wildtype Alg1 was completely abolished (Figure 5B, compare lane 1 with lanes 6 and 8). Alg1-1 accumulated at slightly lower (~ twofold decreased) steady state levels than the wild-type Alg1 protein, as shown by western blot analysis of whole cell extracts, shown in Figure 5C. This decreased level of Alg1-1-HA could explain the reduced amount of Alg2-myc and Alg11-myc that coprecipitated, but not the complete absence of interactions between Alg1-1 and Alg1 (Figure 5B, compare lane 1 with 3). These results demonstrate that sequences within the 16-amino-acid C-terminal tail of Alg1 are required for homo-oligomerization but not for its heteromeric interactions with either Alg2 or Alg11. Identification of catalytically inactivating mutations in ALG1 The model that these ER mannosyltransferase complexes function in vivo strongly predicts that overexpression of mutant alg alleles that affect sugar transfer but not protein± protein interactions should lead to a dominant negative phenotype. In contrast to Alg1, neither Alg2 nor Alg11 accumulated at high steady state levels when over expressed on a 2m plasmid, for reasons that are not apparent (data not shown). Therefore, to test the idea that these complexes are important in vivo, mutations were introduced into ALG1 that were predicted to affect catalysis. To obtain such mutations, we chose to alter amino acid residues that are (1) conserved among related retaining glycosyltransferases and (2) predicted to face the cytoplasm as would be required for utilization of the cytosolic GDP-mannose substrate (Figure 5A). Based on criteria used by several different databases, Alg1, Alg2, and Alg11 have structural features that place them as members of various glycosyltransferase families. All three proteins are members of the PFAM glycosyltransferase 1 group (GT1-PFAM accession number PF0053), whose members transfer nucleotide sugars to a variety of substrates, including polyprenol lipid carriers. In contrast only Alg2 and Alg11 are found in family 4 of the CaZY database, which notably excludes Alg1. The majority of members of both families contain a C-terminal EX 7 E domain, whose first glutamic acid residues has been proposed to participate directly in catalysis (Bourne and Henrissat, 2001) (see Figure 5A for an alignment of a small subset of members of these families). In the case of one family member, the bacterial a-mannosyltransferase AceA, this glutamic acid has been demonstrated to be critical for enzymatic activity (Abdian et al., 2000). Despite other similarities with members of this family, Alg1 notably lacks the EX 7 E, and instead, contains DX 6 D (D363-X6-D370) in this region. In addition, at the position corresponding to E278 of Alg1, most members of the PFAM GT1 and all members of CaZY family 4 contain a lysine that has been proposed to function in nucleotide sugar binding and which is also critical for AceA activity (Abdian et al., 2000). Residues in highly conserved regions that lie in the predicted C-terminal cytoplasmic domain correspond to E278, G310, D363, and D370 of Alg1 (see Figure 5A for a schematic diagram). Each D of the DX 6 D (D363 and D370) was replaced with alanine or residues that would alter the charge. In addition E278 was changed to the more highly conserved K and the conserved G310 and H356 were replaced with alanine. Each of these mutant alg1 alleles was first tested for complementation of the lethality caused by loss of ALG1 (Figure 6A). Each mutant allele was introduced on a CENcontaining plasmid into a strain in which the expression of ALG1 is controlled by the glucose-repressible GAL1 promoter and assayed for the ability to support growth after repression of ALG1 expression by growth in glucose containing media. Neither D363 nor D370 of the DX 6 D domain are essential because mutations that alter either of these residues complemented loss of ALG1, although it should be noted that the alg1-d363a mutant displayed a reduced growth rate (Figure 6A). Therefore, despite the conservation of the Alg1 DX 6 D domain with the EX 7 E catalytic domain of other PFAM glycosyltransferases, these results demonstrate that the aspartates of the Alg1 DX 6 D motif do not play an important role in catalysis. On the other hand, the E278K, G310D and H356Q mutant alleles completely failed to support growth on medium containing glucose, demonstrating that these are null alleles (Figure 6A). The null phenotypes resulting from these mutations may be a direct affect on Alg1-mediated catalysis of sugar addition or alternatively, an indirect affect on Alg1 protein stability or Alg1-mediated protein±protein interactions. To determine if these null mutations affect Alg1 protein levels or oligomeric properties, yeast strains were constructed that coexpress the normal myc-tagged ALG1, ALG2 and ALG11 alleles, along with each of these HAtagged mutant alg1 allele. As determined by western blotting of whole cell extracts, all of these mutant Alg1 proteins are produced at steady state levels that are indistinguishable from the wild-type Alg1 protein (Figure 6B), demonstrating that the observed null phenotype is not due to protein instability. Proteins extracts from these strains were precipitated with anti-myc antibodies and fractionated by SDS±PAGE; protein±protein interactions were assayed by western blotting with anti-ha antibodies (Figure 6C). In contrast to the alg1-1-encoded protein, the results of this experiment demonstrated that mutations that altered E278, G310, and H356 in Alg1, either alone (data not shown) or in combination do not affect the ability of Alg1 to homooligomerize or to hetero-oligomerize with Alg2 or Alg11 (Figure 6C). Together, these results suggest that we have identified mutations in ALG1 that affect catalysis but not protein±protein interactions. To determine the phenotypic consequences of their overexpression, these mutations, either singly or in combination, were introduced on 2m plasmids into an otherwise 565

8 X.-D. Gao et al. A YP+galactose GAL1pr-ALG1 YP+glucose CEN-plasmids alg1- H356Q alg1- ALG1 D370A alg1- alg1- E278K D363A alg1- G310D 30 o C 30 o C + hygromycin B B C Blotting: anti-ha ALG1-HA alg1-1-ha + ALG1-myc IP: anti-myc Blotting: anti-ha ALG1-HA alg1-e278k/h356q-ha alg1-g310d/h356q-ha alg1-d363a/h356q-ha alg1-e278k/h356q-ha alg1-g310d/h356q-ha alg1-d363a/h356q-ha ALG1-HA + ALG2-myc IP: anti-myc Blotting: anti-ha ALG1-HA alg1-e278k/h356q-ha alg1-g310d/h356q-ha alg1-e278k/h356q-ha normal strain. We observed no detrimental effect of singlecopy expression, but overexpression of both alg1-e278k/ H356Q and the alg1-g310d/ H356Q alleles displayed both temperature and hygromycin B sensitive growth (Figure 7). These phenotypes were not observed in a strain overexpressing the wild-type ALG1 gene, demonstrating that the dominant negative phenotype in each case is a direct result of the alg1 mutation. One interpretation of this dominant negative affect is that catalytically inactive Alg1 maintains - alg1-g310d/h356q-ha + ALG11-myc Fig. 6. Identification of catalytically inactive alg1 alleles that do not affect protein±protein interactions or Alg1 stability. (A) Identification of missense null alg1 alleles. Mutations were introduced into the ALG1 gene as described in Materials and methods. CEN-containing plasmids containing the ALG1 or a mutant alg1 gene were transformed into the GAL1p-ALG1 strain (XGY31) and streaked onto YPA medium supplemented with galactose or glucose. (B) Loss-of-function alg1 mutations that do not affect protein stability or homo-oligomeric interactions. Detergent extracts from strains coexpressing HA-tagged catalytically inactive mutant alg1 alleles together with myc-tagged ALG1, ALG2, or ALG11 (XGY20, XGY25, XGY27, XGY36, XGY37, XGY38, XGY44, XGY45, XGY47, XGY48Ðsee Table I) were immunoprecipitated with anti-myc antibodies, fractionated by SDS± PAGE, and subjected to western blot analysis with anti-ha antibodies. (C) Aliquots of extracts containing mutant or wild-type HA-tagged Alg1 protein, used in the immunoprecipitations shown in lanes 1, 3, 4, and 5 of B were analyzed directly by western blot with anti-ha antibodies o C alg1- E278K/H356Q ALG1 2µ plasmids alg1- D363A/H356Q alg1- G310D/H356Q Fig. 7. Overexpression of catalytically inactive alg1 alleles result in a dominant negative phenotype. Yeast strains (W303a) were transformed with 2m, high copy plasmids containing the ALG1 or the indicated catalytically inactive mutant alg1 alleles and grown on selective medium at 30 C, at 37 C, or at 30 C on selective medium supplemented with 50 mg/ml hygromycin B. the ability to oligomerize, but these Alg1 mutant± containing hetero-oligomeric complexes are nonfunctional. When overexpressed, the catalytically inactive Alg1 protein competitively inhibits its wild-type counterpart, which leads to the growth defects we observed. These results thus provide strong genetic evidence for the importance of Alg1-containing complexes in vivo. Discussion In this study we demonstrate that the Alg1, Alg2, and Alg11 mannosyltransferases that catalyze the earliest steps of lipid-linked mannosylation, leading to Man 5 - GlcNAc 2 PP-dolichol on the cytosolic face of the ER, display physical interactions among themselves. The native molecular weight of each of these proteins, far larger than would be predicted from their primary amino acid sequences, is consistent with their existence in oligomeric structures. Alg1 coimmunoprecipitates with both Alg2 and Alg11 when produced at physiological levels of expression. Using a combination of coimmunoprecipitation assays and native gel filtration chromatography, we found that the Alg11 and Alg2 proteins exist in at least two distinct complexes that have Alg1 as a common member. Alg1also has the ability to homo-oligomerize, but we find no evidence

9 Interactions between the Alg1, Alg2, and Alg11 mannosyltransferases that either Alg2 or Alg11 do so. These results imply that Alg1 may be present at two or more copies in each of these complexes, whereas Alg2 and Alg11 proteins are present at single copy. A strong prediction of our model that Alg1- containing complexes are of physiological relevance is that overexpression of catalytically inactive null alg1 alleles should result in inactive complexes. This prediction was borne out by the dominant negative phenotypes displayed by wild-type cells overexpressing mutant alg1 alleles that are apparently catalytically inactive, thus providing strong genetic evidence for the existence of Alg1-containing complexes in vivo. Mutational analysis of ALG1 identified several important residues that lie in positions conserved among related glycosyltransferases. Surprisingly, alteration of the aspartates of the DX 6 D, which aligns with the predicted EX 7 E catalytic motif of other Alg1-related glycosyltransferases (including Alg2 and Alg11), does not result in a severe growth or dominant negative phenotype (Figure 6). Thus despite its similarity with other members of the PFAM GT1 family of glycosyltransferases, Alg1 does not utilize this DX 6 D motif for catalysis. On the other hand, alteration of several other residues conserved in this family (E278, G310, H356) lead to complete loss of Alg1 function, without having an effect on protein stability or protein±protein interactions (Figure 6). Inexplicably, overexpression of either ALG2 or ALG11 does not lead to comparably high steady state, as is observed by ALG1 overexpression (data not shown). A test of the prediction that overexpression of catalytic alg2 and alg11 mutants should result in similar dominant negative phenotypes awaits the identification of experimental conditions that allow overexpression of these genes. How would the association of these proteins in oligomeric complexes benefit the cell? One obvious advantage is that such a nondissociating arrangement of enzymes that act to catalyze sequential reactions limits the distance through which the substrate must diffuse. In the case of the Alg1/ Alg2-containing complex, this arrangement could speed up the rate with which Man 3 GlcNAc 2 -PP-Dol is produced. This explanation is far less satisfying for the role of the Alg1/Alg11 complex. In this case, the benefits are less clear, because the reactions catalyzed by each of these two proteins, that is, addition of the first and the fourth or fifth mannose, are further removed from one another. One other possibility is brought to mind by the observation that Alg2 and Alg11 bear significant similarity to each other along their entire lengths (31% identity/48% similar/pˆ2.6e 06). On the basis of their structural similarities, both proteins are classified in the same glycosyltransferase family, according to the algorithms used by CAZy ( cnrs-mrs.fr/cazy/gt_intro.html) that are not based on predicted activity or substrate usage (Bourne and Henrissat, 2001). Alg2, Alg11, and YPL175W (involved in glycosylphosphatidylinositol anchor synthesis) are the only three Saccharomyces cerevisiae proteins that are members of this particular family, which notably excludes Alg1. The observation that Alg2 and Alg11 are structurally similar but interact with a common partner, Alg1, raises the possibility that this interaction may facilitate or even modify their respective activities. For instance, Alg1 may perform dual functions in both sugar transfer and in the recognition of dolichol-phosphate±derived substrates. An association between Alg1 with both the dolichol anchor as well as with Alg2 or Alg11 could facilitate their activities toward the lipid-linked oligosaccharide. The need for a dolichol recognition mechanism by glycosyltransferases of the ER has been appreciated for many years. A predicted membrane-embedded peptide sequence, conserved among several enzymes (including Alg1) was implicated in dolichol recognition (Albright et al., 1989). Although there is evidence that Alg1 does not require its transmembrane domain or even a dolichol-linked substrate for substrate recognition when expressed in Escherichia coli or in vitro (Revers et al., 1994), whether or not this membrane-spanning domain plays a role in the modulation of mannosyltransferase activity in vivo is unknown. Although the idea that an interaction with Alg1 may affect the activity of Alg2 and Alg11 is purely speculative, it suggests both a common mechanism for dolichol recognition by these glycosyltransferases and a possible advantage for their association. This idea also predicts that Alg1 may be associated with other as yet unidentified mannosyltransferases that catalyze addition of the second a-1,3-linked and perhaps the fourth a-1,2- linked mannose on the cytosolic face of the ER. Further work will be required to test the validity of such an idea. Materials and methods Yeast strains and media Yeast strains used in this study are listed in Table I. W303a (MATa ade2-1 ura3-1 his3-11 trp1-1 leu2-3, 112 can1-100) is the parental strain for the construction of strains that contain HA- or myc-tagged alleles of ALG1, ALG2, or ALG11 (strains XGY20±XGY30) and the GAL1p-driven ALG1 (XGY31). Strains XGY20±XGY30 contain HA- or myctagged alleles of ALG1, ALG2, or ALG11, driven by the TPI promoter, integrated at the ura3 or leu2 locus. In strain XGY31, ALG1 chromosomal promoter sequences (500 base pairs upstream the initiating ATG) were replaced by the GAL1 promoter. Strains XGY32±XGY35 were derived from SEY6210 (MATa ura3-52 leu2-3, 112 his3-d200 trp1- D901 lys2-801 suc2d9). XGY32 contains a replacement of the chromosomal ALG1, ALG2, and ALG11 loci with HA 3 - tagged alleles that are marked with the kanmx4, TRP1, or Saccharomyces pombe his5 gene, respectively. XGY33± XGY35 contain HA-tagged alleles of ALG1 driven by the TPI promoter and integrated at the ura3 locus and also a replacement of the chromosomal ALG1, ALG2, or ALG11 loci, respectively, with myc 13 -tagged alleles that are marked with kanmx4. Epitope tagging of chromosomal loci employed PCR-mediated recombination (Baudin et al., 1993) using a standard set of templates and gene-specific primers (Baudin et al., 1993; Longtine et al., 1998). Standard yeast media, growth condition, and genetic techniques were used (Guthrie and Fink, 1991). Construction of plasmids Standard molecular biology techniques were used for all plasmid constructions (Sambrook et al., 1989). psk ± ALG1-HA contains an epitope-tagged allele of 567

10 X.-D. Gao et al. Table I. Strains used in this study Strain Genotype W303a XGY20 XGY21 XGY22 XGY23 XGY24 XGY25 XGY26 XGY27 XGY28 XGY29 XGY30 XGY31 SEY6210 XGY32 XGY33 XGY34 XGY35 XGY36 XGY37 XGY38 XGY39 XGY40 XGY41 XGY42 XGY43 XGY44 XGY45 XGY46 XGY47 XGY48 XGY49 ALG1 that encodes Alg1 fused to three copies of the HA epitope at the C-terminus (Neiman et al., 1997). To construct the yeast integrative plasmid ptialg1-ha, an EcoRI/ SacI fragment from psk ± ALG1-HA containing the ALG1-HA open reading frame was cloned into a derivative of p ao (Gao et al., 2001), which places ALG1-HA expression under the constitutive TPI promoter. Linearization of this plasmid with XhoI within the URA3 gene targets integration at the ura3-1 locus in W303a. To generate prs- 305TiALG1-HA, a fragment containing TPI-ALG1-HA from ptialg1-ha was cloned into the LEU2, integrative plasmid prs305 (Sikorski and Hieter, 1989). Linearization of this plasmid with AflII within the LEU2 gene targets integration at the leu2 locus. Identical constructions, but containing the triple myc tag at the C-terminus are psk ± ALG1- myc, ptialg1-myc, and prs305tialg1-myc. A series of MATa ade2-1 ura3-1 his3-11 trp1-1 leu2-3,112 can1-100 As in W303a and TPIp-ALG1-HA 3 -URA3 TPIp-ALG1-myc 3 -LEU2 As in W303a and TPIp-ALG2-HA 3 -URA3 TPIp-ALG2-myc 3 -LEU2 As in W303a and TPIp-ALG11-HA 3 -URA3 TPIp-ALG11-myc 3 -LEU2 As in W303a and TPIp-ALG1-HA 3 -LEU2 TPIp-ALG2-myc 3 -URA3 As in W303a and TPIp-ALG11-HA 3 -URA3 TPIp-ALG2-myc 3 -LEU2 As in W303a and TPIp-ALG1-HA 3 -URA3 As in W303a and TPIp-ALG11-HA 3 -URA3 As in W303a and TPIp-alg1-1-HA 3 -URA3 TPIp-ALG1-myc 3 -LEU2 As in W303a and TPIp-alg1-1-HA 3 -URA3 TPIp-ALG2-myc 3 -LEU2 As in W303a and TPIp-alg1-1-HA 3 -URA3 TPIp-ALG11-myc 3 -LEU2 As in W303a and TPIp-alg1±1-HA 3 -URA3 As in W303a and GAL1p-ALG1-kanMX4 MATa ura3±52 leu2±3,112 his3-d200 trp1-d901 lys2-801 suc2d9 As in SEY6210 and ALG1-HA 3 -kanmx4 ALG2-HA 3 -Sphis5 ALG11-HA 3 -TRP1 As in SEY6210 and ALG1-myc 13 -kanmx4 TPIp-ALG1-HA 3 -URA3 As in SEY6210 and ALG2-myc 13 -kanmx4 TPIp-ALG1-HA 3 -URA3 As in SEY6210 and ALG11-myc 13 -Sphis5 TPIp-ALG1-HA 3 -URA3 As in W303a and TPIp-ALG1-myc-LEU2 TPIp-alg1-H356Q-HA-URA3 As in W303a and TPIp-ALG1-myc-LEU2 TPIp-alg1-E278K/H356Q-HA-URA3 As in W303a and TPIp-ALG1-myc-LEU2 TPIp-alg1-G310D/H356Q-HA-URA3 As in W303a and TPIp-ALG1-myc-LEU2 TPIp-alg1-D363A/H356Q-HA-URA3 As in W303a and TPIp-ALG1-myc-LEU2 TPIp-alg1-D370A/H356Q-HA-URA3 As in SEY6210 and ALG1-myc 13 -kanmx4 TPIp-alg1-E278K/H356Q-HA-URA3 As in SEY6210 and ALG1-myc 13 -kanmx4 TPIp-alg1-G310D/H356Q-HA-URA3 As in SEY6210 and ALG1-myc 13 -kanmx4 TPIp-alg1-D363A/H356Q-HA-URA3 As in SEY6210 and ALG2-myc 13 -kanmx4 TPIp-alg1-E278K/H356Q-HA-URA3 As in SEY6210 and ALG2-myc 13 -kanmx4 TPIp-alg1-G310D/H356Q-HA-URA3 As in SEY6210 and ALG2myc 13 -kanmx4 TPIp-alg1-D363A/H356Q-HA-URA3 As in SEY6210 and ALG11myc 13 -kanmx4 TPIp-alg1-E278K/H356Q-HA-URA3 As in SEY6210 and ALG11myc 13 -kanmx4 TPIp-alg1-G310D/H356Q-HA-URA3 As in SEY6210 and ALG11myc 13 -kanmx4 TPIp-alg1-D363A/H356Q-HA-URA3 epitope-tagged ALG2 and ALG11 plasmids, including pti- ALG2-HA, ptialg11-ha, ptialg2-myc, ptialg11- myc, prs305tialg2-myc, and prs305tialg11-myc, were constructed with exactly the same strategy as used to make the ALG1 derivatives already described. The alg1-1 open reading frame was amplified by PCR from the genomic DNA isolated from alg1-1 strain, PRY55 (from P. Robbins), and sequenced. This alg1-1 gene was epitope-tagged and cloned into p ao as described for TiALG1-HA, to generate ptialg1-1-ha. Single missense mutations in ALG1were generated by PCR mutagenesis using ptialg1-myc as a template and verified by DNA sequence analysis. This created myctagged alg1 alleles that produce Alg1 proteins that contain the following amino acids changes: E278K, G310D, H356Q, D363A, and D370A. These mutant alg1 alleles 568

11 Interactions between the Alg1, Alg2, and Alg11 mannosyltransferases were over expressed by placing each under control of the TDH3 promoter, in the 2m-containing plasmid, YEp352GAP. Preparation of detergent extracts Exponentially growing yeast cells were harvested at an OD 600 of 1±3 and converted to spheroplasts with lyticase (Poster and Dean, 1996). To prepare detergent extracts that were used for both FPLC analysis and the coimmunoprecipitation assays, spheroplasts from 5±6 OD units of cells were resuspended in 500 ml ice cold lysis buffer (150 mm NaCl, 10 mm HEPES-KOH [ph 7.5], 5 mm MgCl 2 ) that contained protease inhibitors and 1.0% digitonin as described (Gao and Dean, 2000). Extracts were clarified by centrifugation at 100 kg in a Beckman TLA 100 rotor prior to chromatographic fractionation or immunoprecipitation to remove protein aggregates. Coimmunoprecipitation and western immunoblotting Epitope-tagged proteins in digitonin extracts were immunoprecipitated with anti-ha 12CA5 monoclonal antibodies, anti-ha Y-11 rabbit polyclonal antibodies (Santa Cruz Biotechnology, Santa Cruz, CA), anti-myc 9E10 monoclonal antibodies, or anti-myc A-14 rabbit polyclonal antibodies (Santa Cruz Biotechnology) and protein A- Sepharose (Pharmacia, Uppsala, Sweden). After fractionation by 10% SDS±PAGE, immunoprecipitated proteins were detected by chemiluminescence (ECL, Amersham, Little Chalfont, U.K.) as described previously (Gao and Dean, 2000). Antibody combinations in immunoprecipitations used the rabbit polyclonal antibodies in the primary incubation and mouse monoclonal antibodies for subsequent blotting to eliminate cross-reactivity between the rabbit IgG used for the immunoprecipitation and the antimouse IgG±horseradish peroxidase used for chemiluminescence detection. Gel filtration chromatography Detergent extracts containing 1.0% digitonin were prepared as described from yeast expressing the HA-tagged ALG1, ALG2, or ALG11 alleles. After clarification at 100 kg, 200 ml extract were fractionated over a gel filtration column (Superose 12 HR 10/30, Pharmacia) equilibrated with 150 mm NaCl, 50 mm HEPES-KOH (ph 7.5), 5 mm MgCl 2, 1.0% digitonin, 1 mm phenylmethylsulfonyl fluoride using the FPLC system (Pharmacia). FPLC was performed at a flow rate of 0.2 ml/min, and 400 ml was collected for each fraction. Fractions were analyzed for the presence of HA-tagged Alg1, Alg2, and Alg11 by SDS± PAGE, followed by immunoblotting with HA-specific antibodies. For molecular weight determination, proteins that range from 29,000 to 700,000 MW (Pharmacia) were fractionated in parallel and detected after staining with Coomassie brilliant blue. Acknowledgments We thank Sabine Keppler-Ross for technical support and for help in some of the strain constructions. We also thank previous anonymous reviewers for their helpful comments on the manuscript. Abbreviations ER, endoplasmic reticulum; FPLC, Fast Performance Liquid Chromatography; HA, Hemaglutinnin; PCR, polymerase chain reaction; SDS±PAGE, sodium dodecyl sulfate±polyacrylamide gel electrophoresis; TPI, triose phosphoisomerase. References Abdian, P.L., Lellouch, A.C., Gautier, C., Ielpi, L., and Geremia, R.A. (2000) Identification of essential amino acids in the bacterial alphamannosyltransferase acea. J. Biol. Chem., 275, 40568± Abeijon, C. and Hirschberg, C.B. (1992) Topography of glycosylation reactions in the endoplasmic reticulum. Trends. Biochem. Sci., 17, 32±36. Albright, C.F. and Robbins, P.W. (1990) The sequence and transcript heterogeneity of the yeast gene ALG1, an essential mannosyltransferase involved in N-glycosylation. J. Biol. Chem., 265, 7042±7049. Albright, C.F., Orlean, P. and Robbins, P.W. (1989) A 13-amino acid peptide in three yeast glycosyltransferases may be involved in dolichol recognition. Proc. Natl Acad. Sci. USA, 86, 7366±7369. Baudin, A., Ozier-Kalogeropoulos, O., Denouel, A., Lacroute, F., and Cullin, C. (1993) A simple and efficient method for direct gene deletion in Saccharomyces cerevisiae. Nucleic Acids Res., 21, 3329±3330. Bourne, Y. and Henrissat, B. (2001) Glycoside hydrolases and glycosyltransferases: families and functional modules. Curr. Opin. Struct. Biol., 11, 593±600. Burda, P. and Aebi, M. (1999) The dolichol pathway of N-linked glycosylation. Biochim. Biophys. Acta, 1426, 239±257. Burda, P., Jakob, C.A., Beinhauer, J., Hegemann, J.H., and Aebi, M. (1999) Ordered assembly of the asymmetrically branched lipid-linked oligosaccharide in the endoplasmic reticulum is ensured by the substrate specificity of the individual glycosyltransferases. Glycobiology, 9, 617±625. Cipollo, J.F., Trimble, R.B., Chi, J.H., Yan, Q., and Dean, N. (2001) The yeast ALG11 gene specifies addition of the terminal alpha 1,2-Man to the Man 5 GlcNAc 2 -PP-dolichol N-glycosylation intermediate formed on the cytosolic side of the endoplasmic reticulum. J. Biol. Chem., 276, 21828± Couto, J.R., Huffaker, T.C., and Robbins, P.W. (1984) Cloning and expression in Escherichia coli of a yeast mannosyltransferase from the asparagine-linked glycosylation pathway. J. Biol. Chem., 259, 378±382. Gao, X.D. and Dean, N. (2000) Distinct protein domains of the yeast Golgi GDP-mannose transporter mediate oligomer assembly and export from the endoplasmic reticulum. J. Biol. Chem., 275, 17718± Gao, X.D., Nishikawa, A., and Dean, N. (2001) Identification of a conserved motif in the yeast Golgi GDP-mannose transporter required for binding to nucleotide sugar. J. Biol. Chem., 276, 4424±4432. Guthrie, C. and Fink, G.R. (1991) Guide to yeast genetics and molecular biology. Methods Enzymol., 194, 3±20. Helenius, A. and Aebi, M. (2001) Intracellular functions of N-linked glycans. Science, 291, 2364±2369. Herscovics, A. and Orlean, P. (1993) Glycoprotein synthesis in yeast. FASEB J., 7, 540±550. Huffaker, T.C. and Robbins, P.W. (1982) Temperature sensitive yeast mutants deficient in asparagine-linked glycosylation. J. Biol. Chem., 257, 3203±3210. Huffaker, T.C. and Robbins, P.W. (1983) Yeast mutants deficient in protein glycosylation. Proc. Natl Acad. Sci. USA, 80, 7466±7470. Jackson, B.J., Kukuruzinska, M.A., and Robbins, P.W. (1993) Biosynthesis of asparagine-linked oligosaccharides in Saccharomyces cerevisiae: the alg2 mutation. Glycobiology, 3, 357±