Interactions among Subunits of the Oligosaccharyltransferase Complex*

The mammalian oligosaccharyltransferase (OST) is an oligomeric complex composed of three membrane proteins of the endoplasmic reticulum: ribophorin I (RI), ribophorin II (RII), and OST48. In addition, sequence homology between the Ost2 subunit of the yeast OST complex and Dad1 (defender against apoptotic death) suggests that Dad1 may represent a fourth subunit of the mammalian OST complex. In attempts to elucidate the structural organization of this complex, we have studied the interactions among its subunits. Using the yeast two-hybrid system, we have shown that the luminal domains of RI and RII (RIL and RIIL, respectively) interacted with the luminal domain of OST48 (OST48L), but no direct interaction was observed between RIL and RIIL. These results were confirmed by biochemical assays. Deletion analyses using the yeast two-hybrid system showed that subdomain of RIL or RIIL adjacent to the respective transmembrane domains interacted with OST48L. Of the three equal length subdomains of OST48L, the one at the N terminus and the one next to the transmembrane domain interacted with RIL. None of these three subdomains of OST48L interacted with RIIL. The yeast two-hybrid assay also revealed affinity between the cytoplasmically located N-terminal region of Dad1 and the short cytoplasmic tail of OST48, thus placing Dad1 firmly into the OST complex. In addition, we found a homotypic interaction between the cytoplasmic domains of RI, which may play a role in the formation of the oligomeric array formed by components of the translocation machinery.

During their translocation into the lumen of the rough endoplasmic reticulum, polypeptides made on membrane-bound polysomes may be cotranslationally modified by N-glycosylation (1,2). During this process, the oligosaccharyltransferase catalyzes the transfer of high mannose oligosaccharides, which are preassembled on lipid-anchored dolicholpyrophosphate moieties, to certain asparagine residues facing the lumen of the ER 1 (3). The mammalian OST was first isolated by incubating high salt-extracted dog pancreas rough microsomes with nonionic detergents, followed by purification on sucrose density gradients and by ion exchange chromatography (4). It was found that the OST forms an oligomeric complex that sediments in a sucrose gradient as a 10 S particle. This complex was composed of three integral membrane proteins, of which ribophorin I (RI) (5,6) and OST48 (7) are type I transmembrane proteins with most of their polypeptide chains facing the lumen of the ER (Fig. 1). The third member of the complex, ribophorin II (RII), also has a large luminal domain, but the disposition of the transmembrane and cytoplasmic domains is less clear: although this region bears three hydrophobic domains (6,8), only the one most proximal to the N terminus is of sufficient length and hydrophobicity to function as a typical transmembrane domain and RII may, therefore, also have a type I disposition, like RI and OST48 (Fig. 1). The OST complex was later purified from other mammalian species (9,10), as well as from yeast (11). The yeast OST complex contains apparently six subunits (11)(12)(13)(14)(15)(16), and the mammalian one included originally only three subunits (4). On the basis of homology to the yeast OST proteins, it was possible to identify another mammalian protein as a potential OST component. Dad1 (defender against apoptotic death) was originally cloned by Nakashima et al. (17) as a gene with a temperature-sensitive point mutation that triggers apoptosis at the non-permissive temperature. On the basis of its 40% sequence identity with the yeast Ost2 protein, a membrane protein that copurified with the yeast OST complex (11,16), Dad1 represented a possible fourth subunit of the mammalian OST complex. This conclusion was most recently confirmed when Dad1 was shown to copurify with RI, RII, and OST48 (18). Dad1 is a small hydrophobic protein with a cytoplasmically located N terminus and up to three transmembrane domains (2) (Fig. 1). As is true for the other subunits of the OST, the precise role of Dad1 in N-glycosylation is not known.
In native rough microsomal membranes, components of the OST complex such as the ribophorins are in close proximity to membrane-bound ribosomes, since they can be chemically cross-linked to the 60 S ribosomal subunit (19). Furthermore, site-specific antibodies directed against the cytoplasmic domain of RI, but not those against the luminal domain of the protein, inhibited the targeting to microsomal membranes of ribosome-nascent chain complexes formed in a cell-free translation mixture (20). These results suggest that the ribophorins are also adjacent to the polypeptide translocation site. This conclusion is in agreement with recent studies, showing that during the purification of the Sec61p complex, in which dog pancreas rough microsomes were extracted with certain detergents at different ionic strength, the subunits of the OST complex, RI and RII and OST48, remained tightly associated with the Sec61p complex (21). The Sec61p complex, which is composed of an ␣, ␤, and ␥ subunit, is a key component of the translocation apparatus since it not only forms the translocation pore but has also ribosome binding activity (21,22).
It may be expected that the enzymatic activity of the OST complex is dependent on the integrity of the oligomeric assembly of its subunits. The long range goal of our studies is to understand the contribution of each OST protein to the overall function of the complex, which is likely to be related to the spatial relationship among the components of OST complex.
We present here the first information about specific interactions among the OST subunits. Using the yeast two-hybrid approach in conjunction with in vitro binding studies, we demonstrated that the luminal domain of OST48 interacts with the luminal domains of RI and RII, but we detected no direct interactions between RIL and RIIL. Furthermore, we found that the cytoplasmically located N-terminal region of Dad1 interacts with the cytoplasmic tail of OST48 and that cytoplasmic domains of RI interact with each other in a homotypic fashion.

EXPERIMENTAL PROCEDURES
Yeast Vectors and Plasmid Construction-The yeast vectors pEG202 and pJG4-5 used in the yeast two-hybrid assay were obtained from Drs. Russ Finley and Roger Brent (Massachusetts General Hospital, Boston) (23).
The rat RI and RII cDNAs were cloned in this laboratory by Harnik-Ort et al. (5) and Pirozzi et al. (8), respectively. The dog OST48 cDNA was a gift of Dr. Reid Gilmore (University of Massachusetts Medical School, Worcester) (7). Full-length human Dad1 cDNA was amplified from a HeLa cDNA library (provided by Dr. Li Zhang, New York University Medical Center, New York) using the following primers for PCR: sense strand primer, 5Ј-CCC GAA TCC TCG GCG TCG GTA GTG TCT-3Ј; and antisense strand primer, 5Ј-CCC CTC GAG TCA GCC AAC AAA GTT CAT GAC-3Ј. To make either bait or prey plasmids encoding the constructs shown in Table I, we generated cDNAs encoding these domains by PCR using full-length cDNAs as templates and the respective primers that provided also EcoRI and XhoI restriction sites. All fragments were ligated into the EcoRI and XhoI sites of pEG202, downstream of the LexA DNA binding domain, and into the same sites of pJG4-5, downstream of B42/HA. Transformation of bacteria and preparation of plasmids were done according to standard molecular biology techniques (24). To generate the construct encoding a single copy of the cytoplasmic domain of OST48 (OST48C 1 , amino acids 406 -414), two oligonucleotides (sense strand oligonucleotide 5Ј-AATTC CAC ATG AAG GAG AAG GAG AAA TCT GAC TGA GGA TCC C 3Ј and antisense strand oligonucleotide 5Ј TCGAG GGA TCC TCA GTC AGA TTT CTC CTT CTC CTT CAT GTG G 3Ј) were synthesized. Since OST48C 1 is encoded by only 27 nucleotides, it is impossible to check the insert through enzyme digestion of the cloning sites. We, therefore, introduced a BamHI restriction site after the stop codon. To generate three repeats of OST48C (OST48C 3 ), synthetic oligonucleotides (sense strand oligonucleotide 5Ј-AATTC (CAC ATG AAG GAG AAG GAG AAA TCT GAC) 3 TGA C-3Ј and antisense strand oligonucleotide 5Ј-TCGAG TCA (GTC AGA TTT CTC CTT CTC CTT CAT GTG) 3 G) were made. Sense strand and antisense strand oligonucleotides were phosphorylated and annealed as described previously (24) and ligated into same sites of pEG202 and pJG4-5 as described above. To subclone OST48L into pAA, a plasmid that contains LEU2 as a selectable marker (gift of Dr. Eric Chang, New York University, New York) (25), the PCR-generated cDNA encoding OST48L was inserted in frame between NheI and SalI sites of pAA. All constructs were confirmed by sequencing on the 373 DNA sequencer (ABI, Foster City, CA) according to the manufacturer's protocols.
Yeast Transformation-The yeast strain EGY48 (MATa ura3 his3 trp1 6LexAop-LEU2), provided by Drs. Russ Finley and Roger Brent (23), was used for transformation. The techniques used to apply the yeast two-hybrid system were previously described (26). The yeast transformants were grown at 30°C for 3 days on SD plates that contain 2% dextrose and a mixture of amino acids that lacks histidine and tryptophan as well as uracil. The colonies were then transferred to X-gal plates that contain 2% galactose and 1% raffinose, an amino acid mixture that lacks histidine, tryptophan, and leucine as well as uracil, and allowed to grow further at 30°C for 3 days. For the bridging experiment (Fig. 2), strain EGY48 was transformed with pSH18-34, pEG202-RIIL, and pJG4 -5-RIL, followed by an additional transformation with pAA-OST48L. The colonies were grown on an SD plate without leucine to select pAA-OST48L and then transferred to X-gal plates. The rate of color development was evaluated on 3 consecutive days of incubation of the respective X-gal plates at 30°C. When the blue reac-tion product could be detected at day 1, a rating of "ϩϩϩ" was assigned. If 2 or 3 days were required to obtain a color reaction, "ϩϩ" or "ϩ", respectively, was assigned. Photographs of X-gal plates shown in Figs. 2, 4, and 5 were taken at day 3.
Liquid assays were done for each two-hybrid test as described previously (27). The units were calculated according to the following equation: U ϭ 1000 ϫ A 420 /(t) ϫ (v) ϫ A 600 ; where t is reaction time (min), v is culture volume used in the assay (ml), A 600 indicates cell density at the start of the assay, and A 420 indicates combination of absorbance by 0-nitrophenol and light scattering by cell debris (28). For the ϩϩϩ and ϩϩ ratings, a good qualitative correspondence with the liquid assay data was observed; for the ϩ rating, values were sometimes at or barely above background. If the X-gal plate assay gave consistently blue colonies within 3 days we considered this a positive result, even if the result from liquid assay was ambiguous. The repression assay was done as described previously (29).
Western Blot Analysis-The expression of different B42/HA fusion proteins encoded by the prey plasmids was analyzed by Western blot analysis. For this purpose, yeast cells transformed with the prey plasmid were grown to an A 600 of 1.0 in 3 dropout media (ura-, his-, and trp-) containing 2% galactose and 1% raffinose, and then 5 ml of yeast cells were lysed with 0.25 N of NaOH and 1% ␤-mercaptoethanol and subjected to trichloroacetic acid precipitation. Samples were dissolved in 200 l of SDS loading buffer and boiled for 5 min, and 20 l of sample was resolved by SDS-PAGE, followed by electrophoretic transfer to nitrocellulose membranes. All blots were preincubated in 2% nonfat dried milk and 0.1% Tween 20 prepared in phosphatebuffered saline. The monoclonal antibody 16B12, (BAbCO, Berkeley, CA) was used for detection of HA-tagged fusion proteins, and the polyclonal antibody against OST48 was employed to probe for the expression of OST48L.
In Vitro Binding Assay-A cDNA encoding OST48L with an upstream EcoRI site and a downstream XhoI site was generated from either pEG202-OST48L or pJG4-5-OST48L through digestion with EcoRI and XhoI. This cDNA was then cloned in frame downstream of the glutathione S-transferase (GST) coding region in pGEX-EF (gift of Dr. David Ron, New York University Medical Center, New York). The GST or GST-OST48L fusion proteins was expressed by induction of transformed bacterial cultures with isopropylthio-␤-D-galactoside. Bacteria were lysed through a freeze-thaw cycle and sonication, and GST or GST-OST48L in the lysate was immobilized on glutathione-Sepharose beads (Pharmacia Biotech Inc.). The cDNAs encoding RIL and RIIL were made by PCR using pJG4-5-RIL or pJG4-5-RIIL as templates. The T3 polymerase binding site was introduced by using the following forward primer, GCAATTAACCCTCACTAAAGGGTACCATGGATG-TGCCAGATTATGC. The cDNAs corresponding to RIL and RIIL were transcribed in vitro using T3 polymerase and translated in a reticulocyte lysate system (Promega, Madison, WI) using [ 35 S]methionine to label the proteins. To test for the interactions between OST48L and RIL or RIIL, about 30 g of GST or GST-OST48L immobilized on 15 l of glutathione-agarose beads (1:1 suspension in phosphate-buffered saline) was incubated with 150,000 cpm of [ 35 S]methionine-labeled RIL or RIIL at room temperature for 30 min and then washed four times with phosphate-buffered saline containing 1 mM phenylmethylsulfonyl fluoride, 0.5% Triton X-100, and 1 mM dithiothreitol. The bound proteins were eluted with 5 mM free glutathione in 50 mM Tris (pH 8.0) and mixed with SDS loading buffer. Samples were analyzed by SDS-PAGE followed by autoradiography. To quantify the intensity of bands on autoradiographs, they were scanned and analyzed using the NIH Image software. Fig. 1 shows a schematic representation of four mammalian OST subunits. To detect potential interactions among the luminal domains of these OST subunits, we subcloned the cDNAs encoding RIL 1-415 , RIIL 1-516 , or OST48L 1-385 (see Fig. 1) into the cloning sites of plasmids pEG202 or pJG4-5. Yeast cells of strain EGY48 transformed with the ␤-galactosidase reporter plasmid pSH18-34 were then transformed with different pairs of plasmids, such as pEG202-OST48L/pJG4-5-RIL, pEG202-OST48L/pJG4-5-RIIL, and pEG202-RIIL/pJG4-5-RIL, to express LexA and B42 fusion proteins. As detected on the X-gal plate shown in Fig. 2A, interactions between OST48L and RI or RII resulted in blue yeast colonies, indicating that OST48L interacts with both RIL and RIIL. Since the expression levels of the fusion proteins containing RIL and RIIL were similar (Fig. 2B), the rate of color development indicated that the interaction between OST48L and RIIL is stronger than that between OST48L and RIL. This difference was confirmed by the liquid ␤-galactosidase assay (see Table I). Similar results were obtained when the inserts were exchanged between the two plasmids (pEG202-RIL/pJG4-5-OST48L and pEG202-RIIL/pJG4-5-OST48L) (data not shown). These results ruled out the possibility that OST48L by itself contains a transcriptional activating domain, causing a false positive result. In contrast, no interactions could be detected between RIIL and RIL ( Fig. 2A; Table I). To substantiate this negative result, a repression assay was performed (29), demonstrating that the LexA-RIIL fusion protein encoded by the bait plasmid, pEG202-RIIL, did in fact enter the nucleus and bind to the LexA operator, thus blocking the synthesis of the lacZ gene product (data not shown). These observations suggested that OST48L serves as a bridge between RIL and RIIL. To examine this hypothesis more directly, we subcloned OST48L into pAA, a plasmid that encodes neither a DNA binding domain nor a transcriptional activating domain (25), and co-expressed OST48L with the fusion proteins LexA-RIIL and B42-RIL in yeast cells. As expected, the co-expression of OST48L with LexA-RIIL and B42-RIL led to the activation of the lacZ gene, indicating that OST48L forms a bridge between RIL and RIIL ( Fig. 2A). A value of 9 units of ␤-galactosidase activity was found for the latter triple transformation experiment which is significantly above background but clearly lower than those in experiments where OST48L was cotransformed with RIL or RIIL. To ascertain that the B42/HA fusion proteins with RIL and RIIL were expressed at similar levels, the respective yeast cell lysates were analyzed by Western blotting using the monoclonal antibody 16B12, which is directed against the HA tag (Fig. 2B). In a similar fashion, the expression of OST48L in yeast cells transformed with the pAA plasmid containing the OST48 cDNA was confirmed using the polyclonal antibody directed against OST48 (Fig. 2B).

Interactions of the Luminal Domains of Both RI and RII with the Luminal Domain of OST48 Were Detected by the Yeast Two-hybrid System-
The Interactions between Different Subunits of the OST Complex Can Be Confirmed Biochemically-To confirm the interactions between OST48L and RIL or RIIL by an independent approach, we established a biochemical assay. In this assay, GST or a GST-OST48L fusion protein expressed in the bacteria (Fig. 3A) was incubated with [ 35 S]methioninelabeled RIL or RIIL which was synthesized in an in vitro transcription/translation system (Fig. 3B). After extensive washing with buffer containing 0.5% Triton X-100 and 1 mM dithiothreitol, bound proteins were eluted with free glutathi-  figure (Fig. 2A). OST48L was also co-expressed with the LexA-RIIL and B42-RIL fusion proteins (Fig. 2A). The Rate refers to the rate of color development (for details, see "Experimental Procedures"). B, fusion proteins B42-HA-RIL, B42-HA-RIIL, and OST48L were expressed in yeast. Equivalent amounts of yeast lysates were separated by SDS-PAGE followed by Western blot analysis using either the monoclonal antibody (16B12) against the HA epitope or a polyclonal antibody against OST48. The arrows indicate the position of molecular mass markers (in kDa). one and analyzed by SDS-PAGE, followed by autoradiography (Fig. 3C). Quantitative analysis of autoradiographs showed that about three times as much labeled RIL and RIIL bound specifically to GST-OST48L compared with GST alone, thus confirming the interactions revealed in the yeast twohybrid system ( Fig. 2A). Although only about 1% of the labeled RIL and RIIL added to the incubation mixture bound specifically to GST-OST48L, this result was highly reproducible (three independent experiments). It is not entirely surprising that the level of specific binding is low, since the proper folding of the in vitro synthesized products as well as of the GST-OST48L fusion protein made in bacteria is a prerequisite for this assay to function.
Characterization of Subdomains within RIL, RIIL, and OST48L Involved in Interactions among the OST Subunits-To determine which regions within RIL and RIIL are responsible for the observed interactions with OST48L, we again used the yeast two-hybrid assay and expressed fusion proteins with subdomains of RI and RII. Initially, we tested for the interaction of OST48L with either the N-terminal halves of RIL and RIIL (RI 1-207 and RII 1-258 , see Fig. 1) or the C-terminal halves of RIL and RIIL (RI 208 -415 and RII 259 -516 , see also Fig. 1). We found that the C-terminal halves of RIL (RI 208 -415 ) and RIIL (RII 259 -516 ) are mainly responsible for the interactions with OST48L (Fig. 4A, b and f; Table I), whereas the N-terminal halves do not interact with OST48L (Fig. 4A, a and e; Table I).
To define even more narrowly the interacting domains within RI 208 -415 or RII 259 -516 , we generated the deletion mutants RI 208 -311 , RI 312-415 , RII 259 -387 , and RII 388 -516 , (see Fig. 1). Results shown in Fig. 4A revealed that RI 312-415 (d) and RII 388 -516 (h), which are adjacent to the respective transmembrane domains of RI and RII, have high affinity for OST48L, whereas essentially no interactions were observed between OST48L and RI 208 -311 (c) or RII 259 -387 (g) (see also Table I). In the set of liquid assays where OST48L was expressed as a LexA fusion protein, the background was rather high, and ␤-galactosidase activity units of up to 18 were observed.
To localize the regions within OST48L that are responsible for interacting with RIL and RIIL, we divided OST48L into three subdomains of equal length, OST48 1-128 , OST48 129 -256 , and OST48 257-385 (Fig. 1), and tested them for interactions with RIL and RIIL in the yeast two-hybrid system. We observed that the one-third portion at the N terminus of OST48L (OST48 1-128 , i) and another portion next to the transmembrane domain of OST48L (OST48 257-385 , k) interacted with RIL, but no interaction was seen between the middle region (OST48 129 -256, j) and RIL ( Fig. 4A; Table I). On the other hand, none of the three subdomains of OST48L interacted with RIIL (data not shown), suggesting that either the RIIL-binding sites within OST48L were destroyed by the deletions or that the entire structure of OST48L is required for its interaction with RIIL. In these experiments, Western blot analysis using anti-HA tag antibody showed that subdomains of RI, RII, and OST48 fused with B42/HA were expressed at comparable levels (Fig. 4B, for details, see also legend to Fig. 2B).
The Yeast Two-hybrid System Allowed Us to Detect Interactions between the Cytoplasmic Domain of OST48 and the Nterminal Tail of Dad1, as Well as Homotypic Interactions of the Cytoplasmic Domains of RI-We have also tested whether the cytoplasmic domains of RI, RII, OST48, and Dad1 interact with each other. The cDNAs encoding the cytoplasmic domains of RI and RII (RIC and RIIC, respectively) and the cytoplasmically located N-terminal domain of Dad1 (Dad1 1-28 ) were subcloned into pEG202 and pJG4-5 at the same sites as described above. Since the cytoplasmic domain of OST48 is only 9 amino acids long, we initially inserted a cDNA encoding three repeats of this domain (OST48C 3 ) into pEG202 and pJG4-5. By expressing pairwise combination of these constructs in the yeast twohybrid system, we found that OST48C 3 interacted with Dad1 1-28 (Fig. 5A, a; Table I). The possibility that OST48C 3 by itself contains a transcriptional activating domain, leading to a false positive result, was ruled out by exchanging the inserts between the two plasmids (data not shown). We found consistently the ␤-galactosidase activity on the X-gal plate higher (ϩϩ, Fig. 5A) than expected from the results of the liquid assay (10 units, Table I). An interaction of apparently similar strength was obtained when only one copy of the cytoplasmic domain of OST48 (OST48C 1 ) was tested for interaction with Dad1 1-28 (see Table I). Using the yeast two-hybrid system, apparently no interaction was observed between RIC and Dad1 1-28 or between RIC and OST48C 3 , since results in both the X-gal plate assay and the liquid assay were negative ( Fig.  1, 5A, b and c; Table I). Similarly, RIIC does not interact with RIC, Dad 1-28 , or OST48C 3 (data not shown). That the fusion protein LexA-RIIC encoded by pEG202-RIIC did enter the nucleus was demonstrated by the repression assay since the expression of LexA-RIIC in yeast cells blocked synthesis of the lacZ gene product (data not shown). We also tested for homotypic interactions between the cytoplasmic domain of RI by co-expressing RIC in the yeast cells as LexA and B42/HA fusion proteins (Fig. 5A, d; Table I). A weak but consistently positive reaction was observed on the X-gal plate. Based on the rate of color development, the homotypic interaction is not as strong as the interaction between OST48C 3 and Dad1 1-28 (compare Fig.  5A, a-d). As in previous experiments, we demonstrated that the B42-HA fusion proteins with OST48C 3 , RIC, RIIC, and Dad1 1-28 were expressed at comparable levels by performing Western blots (Fig. 5B). DISCUSSION The yeast two-hybrid system has been used successfully to characterize interactions of nuclear and/or cytoplasmic proteins and the cytoplasmic or extracellular domains of membrane proteins from a variety of organisms (30 -33). In the experiments described here, the yeast two-hybrid system was employed for the first time to detect interactions between protein domains that are normally within the ER lumen, which is topologically equivalent to the extracellular space. In contrast to the cytosol and the nucleoplasm, the lumen of the ER is characterized by high calcium concentrations. It is also the site where RIL and RIIL are normally N-glycosylated (1). Since the yeast two-hybrid test functions in the nucleoplasm of yeast cells, which is not a high calcium environment, it appears that interactions between the luminal domain of OST48 and those of RI and RII are not dependent on the presence of high calcium levels. Likewise, it appears that RI and RII, which are glycosylated in the native state, but not in the yeast two-hybrid system, do not need to be glycosylated to interact with OST48. Evidence that N-glycosylation of RII may not be required for its proper oligomerization and functioning is provided by the fact that, in vivo, only about half of the molecules are glycosylated (6).
As mentioned in the Introduction, the oligosaccharyltransferase activity was first purified from canine pancreas rough microsomal membranes, and initially three subunits RI, RII, and OST48 were identified as part of an oligomeric complex (4). Later the corresponding yeast enzyme was isolated and shown to contain Ost1p, Swp1p, Wbp1p, and Ost2p (which are homologous to the mammalian OST subunits RI, RII, OST48, and Dad1, respectively) as well as two additional polypeptides (Ost3p and Ost5p). The first four subunits are essential for viability of haploid yeast cells and are required for oligosaccharyltransferase activity as demonstrated by in vivo and in vitro assays (11-13, 16, 34). It was observed that Ost2p, a subunit of the yeast complex, showed 40% sequence identity with the mammalian protein Dad1 as described before. Dad1 was characterized as an integral membrane protein that suppresses apoptotic cell death and a specific temperature-sensitive mutation in the corresponding gene triggered apoptosis at the non-permissive temperature (17). Taken together, these findings suggested that Dad1 is a fourth subunit of the mammalian OST complex, which was recently confirmed by biochemical means (18). It appears therefore that underglycosylation of certain proteins caused by the temperature-sensitive point mutation in the dad1 gene results in apoptosis. The notion that inhibition of N-glycosylation may indeed lead to apoptosis was recently demonstrated by treating human promyelocytic HL-60 cells with tunicamycin. It was shown that this drug, which interferes with the synthesis of dolichol-linked oligosaccharide, triggers the apoptotic death of these cells (35). In the present study we provide evidence that Dad1 interacts with OST48 via cytoplasmically exposed domains, which strongly supports the notion that Dad1 is indeed a fourth subunit of the mammalian oligosaccharyltransferase complex.
The finding that the cytoplasmic domains of RI interact with each other in a homotypic fashion is not completely unexpected. Our previous analysis of the conformation of the RI sequence revealed that the cytoplasmic portion of RI is likely to form ␣-helical domains (5), and helical wheel analysis provided evidence for two short heptad repeats (amino acids 527-537 and 572-583). The formation of leucine zipper-like structures may, therefore, interconnect adjacent oligosaccharyltransferase complexes. This type of interaction may be part of the mechanism that integrates the OST complex into a proteinaceous network formed by the components of translocation apparatus. The existence of such a proteinaceous network has been inferred from our previous results showing that the components of the translocation apparatus, including membrane-bound ribosomes, form large oligomeric arrays that define the domain of the rough endoplasmic reticulum and are physically excluded from small transport vesicles that bud from the transitional elements of the ER (1, 19, 36). Since the oligosaccharyltransferase complex is located in the immediate vicinity of the protein translocation channel (19,20) and tightly bound to the Sec61p complex (21), interactions among these components would provide a mechanism to segregate the components of the translocation apparatus, including the OST complex, to the rough domain of the ER (36). Since the cytoplasmic domain of RI is not conserved in its yeast homologue Ost1, the homotypic interaction between RIC may be a unique feature of mammalian cells.
Our findings concerning the interactions between the mammalian OST subunits, OST48 and RII, are also supported by genetic tests and chemical cross-linking studies performed on the yeast OST. te Heesen et al. (12) have shown that overexpression of SWP1 results in the suppression of the temperature-sensitive phenotype of the WBP1-2 mutation and that Swp1p can be cross-linked to Wbp1p when yeast membrane fractions, solubilized with Triton X-100, are treated with a reversible cross-linking reagent (12). A sequence comparison of RII with its yeast homologue Swp1p reveals that the Swp1p represents a truncated form of RII lacking about two-thirds of the N-terminal portion of the luminal domain (13). The conserved luminal domain of Swp1p is homologous to the subdomain of RIIL (RII 388 -516 ) that is responsible for the interaction with OST48L. By analogy to our findings on the mammalian oligosaccharyltransferase complex, this portion of Swp1p is expected to interact with the luminal domain of Wbp1. Support for the observed interaction between Dad1 and OST48 was also obtained for the yeast homologues. Silberstein et al. (16) demonstrated that the OST2 gene is a suppressor of the WBP1-2 mutation and that overexpression of Ost2p increases the stability of Wbp1p and Swp1p (16). These results suggest a direct physical interaction among Wbp1, Swp1, and Ost2, which is analogous to the interactions that occur among the mammalian homologues OST48, RII, and Dad1. Although the sequence homology between the mammalian OST subunits and their yeast counterparts is not very high (22-28% identity for RII, OST48, and RI and 40% for Dad1) (2), the near neighbor relationships between the subunits of the OST complexes of these phylogenetically distant species appear to be well conserved. This may be expected, since the reactions catalyzed by the yeast or the mammalian enzyme complexes are identical. A more detailed structural analysis of both OST complexes may, in fact, allow us to identify functionally important domains within the respective subunits.