The topology of the ER-resident phospholipid methyltransferase Opi3 of Saccharomyces cerevisiae is consistent with in trans catalysis

Phospholipid N-methyltransferases (PLMTs) synthesize phosphatidylcholine by methylating phosphatidylethanolamine using S-adenosylmethionine as a methyl donor. Eukaryotic PLMTs are integral membrane enzymes located in the endoplasmic reticulum (ER). Recently Opi3, a PLMT of the yeast Saccharomyces cerevisiae was proposed to perform in trans catalysis, i.e. while localized in the ER, Opi3 would methylate lipid substrates located in the plasma membrane at membrane contact sites. Here, we tested whether the Opi3 active site is located at the cytosolic side of the ER membrane, which is a prerequisite for in trans catalysis. The membrane topology of Opi3 (and its human counterpart, phosphatidylethanolamine N-methyltransferase, expressed in yeast) was addressed by topology prediction algorithms and by the substituted cysteine accessibility method. The results of these analyses indicated that Opi3 (as well as phosphatidylethanolamine N-methyltransferase) has an N-out C-in topology and contains four transmembrane domains, with the fourth forming a re-entrant loop. On the basis of the sequence conservation between the C-terminal half of Opi3 and isoprenyl cysteine carboxyl methyltransferases with a solved crystal structure, we identified amino acids critical for Opi3 activity by site-directed mutagenesis. Modeling of the structure of the C-terminal part of Opi3 was consistent with the topology obtained by the substituted cysteine accessibility method and revealed that the active site faces the cytosol. In conclusion, the location of the Opi3 active site identified here is consistent with the proposed mechanism of in trans catalysis, as well as with conventional catalysis in cis.

The CDP-choline (Kennedy) pathway is the other main source of PC in eukaryotes. In mammals and fungi, the CDPcholine pathway depends on the supply of choline for net synthesis (8), whereas plants contain methyltransferases converting phosphoethanolamine to phosphocholine that serves as substrate (9). By converting SAM into S-adenosylhomocysteine (SAH), the PE methylation pathway impacts metabolism of sulfurcontaining amino acids (10), and its inactivation was recently shown to lead to accumulation of SAM and hypermethylation of histones and the major phosphatase PP2A in yeast (11).
Two main types of PLMT have been described to date. Several bacterial species harbor a soluble, cytosolic PLMT that synthesizes PC by transiently associating with membranes dependent on the anionic/zwitterionic phospholipid ratio (12). Eukaryotic PLMTs are integral membrane proteins located in the endoplasmic reticulum (ER) (7,13,14), from which the synthesized PC can be transported to other cellular compartments via various mechanisms (15).
Yeast Opi3 has recently been proposed to be capable of in trans catalysis at membrane contact sites, where membranes stay in very close apposition within a distance of 10 -30 nm (16). In trans catalysis entails the conversion of lipid substrate in one membrane by an enzyme located in a juxtaposed membrane and has been put forward as a mechanism of intermembrane regulation of lipid composition that avoids the energetically expensive efflux of lipids from a donor membrane and transport to the destination membrane (17).
Several lines of evidence support in trans catalysis by Opi3. Opi3 in isolated ER membranes (microsomes) from a strain lacking Cho2 (cho2⌬) was shown to methylate PMME accumulated in microsomes isolated from an opi3⌬ mutant (18). Plasma membrane-ER contact sites were shown to be required for PC synthesis in vivo by Opi3 using yeast mutant strains with reduced cortical ER (19,20). In trans catalysis by Opi3 is further supported by the ability of microsomal Opi3 to convert PMME in liposomes in vitro (20) and more recently by the demonstration that cell-free synthesized Opi3 in nanodisc membranes converts PMME supplied in enzyme-free nanodiscs (21).
A prerequisite for in trans catalysis by Opi3 is a cytosolic localization of its active site. So far, attempts to obtain crystal structures of phospholipid methyltransferases have not been successful (22). Shields et al. (13,23) investigated the membrane topology and the localization of the active site of human PEMT, and obtained results consistent with a cytosolic localization of both N and C terminus, as well as the active site.
Here the membrane topology of Opi3 is analyzed using a dual topology reporter (DTR) and by the substituted cysteine acces-sibility method (SCAM). These methods have recently been applied to analyze the membrane topology of a number of yeast lipid biosynthetic enzymes residing in the ER membrane including PI synthase and several acyltransferases (24 -26). The results show that Opi3 contains four transmembrane domains (TMs), with the fourth TM forming a re-entrant loop, and has an N-out C-in topology. Because the membrane topology obtained for Opi3 differs from the topology previously proposed for human PEMT (13), PEMT was expressed in yeast and analyzed by SCAM. The topology obtained for human PEMT was similar to that of Opi3. Based on a comparison with membrane methyltransferases of known structure (27,28), potential SAM-binding residues in Opi3 were predicted and tested. The active site of Opi3 was found to face the cytosol, consistent with in trans catalysis by Opi3.

Membrane topology of Opi3
Opi3 is a 206-aa membrane protein containing four hydrophobic regions according to the hydropathy plot and in agreement with the ⌬G app values for membrane insertion (Fig. 1A).  (38) according to Kyte and Doolittle (39), and ⌬G app values (blue line) for membrane insertion generated by TOPCONS. B, Opi3 topology predictions generated by algorithms: TOPCONS (topology 1), OCTOPUS (topology 2), SPOCTOPUS (topology 3), Philius (topology 4), SCAMPI (topology 5), and PolyPhobius (topology 6) using the TOPCONS server. The numbers indicate amino acid positions in the Opi3 sequence. Potential transmembrane segments and loops are indicated by Roman numbers and capital letters, respectively. Blue and red lines indicate luminal and cytosolic localization of loops, respectively.

Membrane topology of Opi3
TOPCONS algorithms suggest the presence of four or five membrane-spanning domains and a cytosolic localization of the C terminus, whereas the predicted localization of the N terminus varies (Fig. 1B). Three types of structure prediction emerge: (i) N-out, C-in following the positive-inside rule (the charge distribution in Opi3 and net charge calculations for the soluble domains predicted by PolyPhobius are shown in Fig.  S1A), localizing positively charged domain A and the C terminus cytosolically, whereas the other soluble domains localize to the ER lumen, generating a re-entrant loop in TM III (topologies 1-3 in Fig. 1B); (ii) N-in, C-in without any re-entrant loops (topologies 4 and 5 in Fig. 1B); and (iii) N-out, C-in topology with a re-entrant loop in TM IV that corresponds to aa stretch 138 -185 (topology 6 in Fig. 1B). Fig. S1B shows the apparent free energies for membrane insertion calculated for the TM domains predicted by the algorithms.
The predicted cytosolic localization of the C terminus was tested by attaching the SUC2-HIS4C DTR (29) to the C terminus of Opi3. If the C-terminal DTR localizes to the cytoplasm, the His4C fragment allows histidine auxotrophic cells to grow on histidinol, whereas DTR localization in the ER lumen results in a protein size shift upon Endo H treatment caused by N-glycosylation of the Suc2 fragment. Fig. 2A shows that the Opi3-DTR fusion protein is not Endo H-sensitive and allows growth on histidinol, demonstrating that the C terminus localizes to the cytosol.
To address the full topology of Opi3, SCAM was applied to microsomes (ER-derived vesicles), using ubiquitin-6-maleimido-caproic acid sulfo-n-succinimidyl ester (Ubi-EMCS) as membrane-impermeant cysteine-reactive probe that increases protein mass by 9 kDa for each cysteine labeled (24,25). Because microsomes have a right-side-out membrane orienta- . Serial dilution experiment testing the ability of STY50 cells expressing Opi3-DTR versus the empty vector control (EV) to grow on histidinol and histidine (right panels). B, complementation test of Opi3 single cysteine mutants. Cells from a cho2opi3 strain transformed with the pYES2.1 TOPO vector encoding V5-His 6 -tagged WT or mutant Opi3 as indicated were serially diluted on synthetic media lacking choline and containing galactose and raffinose (SGR cho Ϫ ) to induce expression from the GAL1 promoter or containing 1 mM choline and glucose (SD cho ϩ ) as a positive control and incubated for 3 days at 30°C. EV denotes the empty vector control. The cells expressing active Opi3 grow in medium containing galactose and lacking choline. C, the single cysteine in Kar2 is not accessible to Ubi-EMCS under SCAM conditions in the absence of detergent. Preparations of microsomes from two Opi3 single cysteine mutants were incubated with Ubi-EMCS in the presence and absence of 1% SDS. Ubiquitination was analyzed by Western blotting using anti-Kar2p. D, cysteine accessibility in Opi3 in the absence and presence of SDS was probed with Ubi-EMCS using microsomes from cho2opi3 cells expressing WT, cysteine-free (cys0), and single-cysteine Opi3-V5-His 6 constructs as indicated. Ubiquitination was analyzed by Western blotting using antibodies against the C-terminal V5 tag. E, SCAM results of Opi3 superimposed on the PolyPhobius topology prediction (topology 6 in Fig. 1B). Accessible (cytosolic) and inaccessible cysteines are indicated in red and black, respectively, and soluble domains are in green. The results are representative of two independent experiments.

Membrane topology of Opi3
tion, the probe reacts only with cysteines in cytosolic domains. Based on the topology predictions (Fig. 1B), a set of single cysteine Opi3 constructs was generated with the cysteine localized in each of the putative transmembrane and soluble domains and equipped with a C-terminal V5-His 6 tag. Single cysteine mutants were obtained by site-directed mutagenesis, replacing three of the four native cysteines (Cys-29, Cys-54, Cys-61, and Cys-89) by serine residues, or by substituting single amino acids by cysteine in a cysteine-free Opi3 background. Subsequently, the constructs were expressed in a cho2opi3 double deletion mutant. The ability of the plasmid-borne OPI3 alleles to complement the choline auxotrophy of the cho2opi3 double deletion mutant was used to test the activity of the single cysteine Opi3 constructs. Upon induction from the GAL1 promoter, almost all alleles supported growth of the cho2opi3 strain in choline-free medium, except for S115C, S154C, and the cysteine-free construct (Fig. 2B).
Impermeability of the microsomes to Ubi-EMCS under the reaction conditions used was verified by analysis of the soluble luminal protein Kar2 that contains a single cysteine. Only in the presence of detergent was the cysteine modification observed (Fig. 2C) in agreement with previous reports (24,25). Fig. 2D shows the results of the cysteine scanning of Opi3. Cys-54 (loop A), V138C (loop C), and Ala-197 (C-terminal part of Opi3) were accessible to reaction with Ubi-EMCS in the absence of detergent, indicating a cytosolic localization. In contrast, S4C (N-terminal part), T93C (putative loop B), and Cys-29, Cys-61, and G166C (in predicted transmembrane domains) ( Fig. 1) only reacted with Ubi-EMCS in the presence of detergent. S115C and S154C displayed similar behavior, but these results should be interpreted with caution, because both constructs were not functional (Fig. 2B). As expected, all Opi3 variants were modified by Ubi-EMCS in the presence of SDS, except for the cysteine-free construct. The N-out (luminal), C-in (cytosolic) topology is indicative for an odd number of transmembrane segments. The localization of loop C in the cytosol indicates that putative transmembrane segment IV forms a re-entrant loop or is divided into two small transmembrane domains, as predicted by the PolyPhobius algorithm (Fig. 2E, compare with topology 6 in Fig. 1B).

Membrane topology of human PEMT
The N-out, C-in Opi3 topology deduced from the SCAM data is at odds with the N-in, C-in topology reported previously for human PEMT (13). PEMT and Opi3 show a high level of sequence conservation (56% sequence similarity and 41% identity according to the EMBOSS Needle server (30)) and corresponding predicted membrane topologies (Fig. S1A). Therefore, we decided to extend the SCAM analysis to human PEMT heterologously expressed in yeast. The ability of PEMT expressed from the GAL1 promoter to suppress the choline auxotrophy of cho2opi3 yeast cultured on choline-free medium containing galactose (SGR cho Ϫ ; Fig. 3A) demonstrates that the enzyme adopts a functional conformation and thus can be used for SCAM. SCAM analysis revealed that mutants T39C (loop A), A128C (loop C), and S193C (C terminus) reacted with Ubi- A, complementation test of PEMT single-cysteine mutants. The cells from a cho2opi3 strain transformed with the pYES2.1 TOPO vector encoding V5-His 6 -tagged WT or mutant PEMT as indicated were serially diluted on choline-free SGR medium to induce expression from the GAL1 promoter and on SD medium supplemented with 1 mM choline (positive control) and incubated for 3 days at 30°C. EV denotes the empty vector control. B, cysteine accessibility in PEMT in the absence and presence of SDS was probed with Ubi-EMCS using microsomes from cho2opi3 cells expressing WT, cysteine-free (cys0), and the single cysteine PEMT-V5-His 6 constructs as indicated. Ubiquitination was analyzed by Western blotting using antibodies against the C-terminal V5 tag. The dashed lines in the second panel serve to indicate that the S77C construct was analyzed on a separate gel and that the scan of its blot was spliced to correct an error that occurred during loading of the gel. C, SCAM results of PEMT superimposed on the PolyPhobius topology prediction ( Fig. S1 and topology 6 in Fig. 1B). Accessible (cytosolic) and inaccessible cysteines are indicated in red and black, respectively, and soluble domains are in green. The results are representative of two independent experiments.

Membrane topology of Opi3
EMCS reflecting a cytosolic localization, whereas Y7C (N terminus), S77C (loop B), and S146C (loop C) together with Cys-53, V101C, D139C, and T166C did not, indicating a luminal or transmembrane localization (Fig. 3B). The results obtained for mutants T39C, Cys-53, and T166C should be interpreted with caution because the growth of strains expressing these alleles is similar to the strain harboring the empty vector control (Fig.  3A). The cysteine accessibility scanning of the PEMT and Opi3 constructs yielded consistent results (Figs. 2E and 3C), indicating that both enzymes share a membrane topology consistent with that predicted by PolyPhobius ( Fig. 1B and Fig. S1A).

Localization of the active site of Opi3
To identify amino acids involved in the catalytic cycle of Opi3, the amino acid sequence of Opi3 was screened for conserved motifs using HHpred software. Three significant hits were found: a putative isoprenyl cysteine carboxyl methyltransferase from Metanosarcina acetivorans (M.a. ICMT, 4a2n, E value 4.5E-21) (27), an isoprenylcysteine carboxyl methyltransferase from Tribolium castaneum (T.c. ICMT, 5v7p, E value 3.5E-20) (28), and ⌬ 14 -sterol reductase from Methylomicrobium alcaliphilum (4QUV, 2.7E-15) (31). Here we focus on the similarity between Opi3 (and PEMT) and the two ICMTs because they also use SAM as methyl donor. A relatively high level of similarity is observed in the C-terminal region (aa 98 -206 in Opi3), as shown in the alignment with the corresponding sequences of human PEMT, M.a. ICMT, and T.c. ICMT (Fig. 4A). We generated a model of the structure of the C-terminal part of Opi3 (aa 98 -206) based on the crystal structure of M.a. ICMT (aa 75-194) using the SWISS-MODEL homology modeling server. The amino acids forming hydrogen bonds with SAH the product of SAM demethylation) in M.a.
Even though the quality of the model is not very high (QMEAN Ϫ4.62) (33), the overall predicted structure of C-terminal Opi3 resembles that of M.a. ICMT. Fig. 4B shows a superposition of the modeled C-terminal part of Opi3 (aa 98 -206 of Opi3) on the structure of M.a. ICMT (aa 75-131) (27) and zooms in on the SAM-binding pocket, with the residues involved in SAH binding in ICMT and the corresponding conserved residues in Opi3 indicated. Positions of the PXY and EE motifs in Opi3 match with those in M.a. ICMT, suggesting involvement in hydrogen bonding to SAM/SAH. Alanine substitutions of Pro-149, Tyr-151, Glu-188, and Asp-128, which may correspond to Glu-108 and Gln-189 that contact SAM in M.a. ICMT and T.c. ICMT, respectively (27,28), abrogated Opi3 activity in vitro (Fig. 5A), and in vivo (Fig. 5B). Western blotting analysis showed that the impaired enzyme activity is not due to reduced expression levels of the Opi3 mutant proteins with the exception of the P149A mutant (Fig. 5C) that may be susceptible to proteolytic degradation. Interestingly, two conserved glycine residues (Gly-106 and Gly-108) in Opi3, which were implicated in SAM binding in PEMT (23), turned out not to influence the activity of Opi3 (Fig. 5, A and B), in agreement with their localization far from the putative SAMbinding site (Fig. 4). Importantly, the model of the C-terminal part of Opi3 shows a re-entrant loop formed by TM IV, consistent with the topology of Opi3 proposed based on the SCAM analysis and predicted by PolyPhobius (Fig. 2E). However, in contrast to the PolyPhobius prediction, helix IV would start at Pro-149 rather than Val-138, and accordingly the re-entrant loop would be centered around Gly-166 rather than Gly-153 as in the Poly-Phobius structure. We conclude that the active site of Opi3 is located at the cytosolic face of the ER membrane.

Discussion
The topology of integral membrane enzymes and the localization of their active sites is important for understanding their mechanism of action. For the occurrence of in trans catalysis, the localization of the enzyme's active site is of vital importance, because it determines the accessible lipid substrates. Here, Opi3, the phospholipid methyltransferase from yeast that has been proposed to convert lipid substrates in trans has been analyzed for membrane topology and localization of its active site.
Opi3 is a small (23 kDa) membrane protein insensitive to glycosylases like peptide:N-glycosidase and Endo H ( Fig. 2A and data not shown) and therefore devoid of glycosylation. Topology prediction algorithms yielded four or five transmembrane domains varying in orientation and different localizations of the N terminus (Fig. 1). The discrepancies between topology predictions may be due to the ambiguous character of amino acid stretch 138 -157 ( Fig. 1 and Fig. S1B). Because of its low hydrophobicity (and high membrane insertion energy, ⌬G app ϭ 7.87 kcal/mol), most of the topology prediction algorithms we tested place it in the ER lumen. PolyPhobius is the exception that predicts this sequence to be in the membrane. Another potential reason for the discrepancies is the net negative charge of soluble domains B and C (Fig. S1A), resulting in a joint localization to the ER lumen and a predicted longer TM domain III forming a re-entrant loop (Fig. 1, topologies 1-3). Human PEMT shares 41% sequence identity with Opi3; therefore it is not surprising that its predicted topology is very similar (Fig. S1A).

Membrane topology of Opi3
The data obtained by SCAM analysis of Opi3 and of functionally expressed human PEMT (Figs. 2D and 3B) combined with the modeling of the structure of the C-terminal half of Opi3 (Fig. 4B) yield a coherent picture of the arrangement of the TM helices in the membrane that is consistent with the topology predicted by PolyPhobius (Fig. 1B, topology 6) except for the position of the C-terminal helix-loop-helix. Both enzymes contain four TMs, of which the more extended TM IV forms an intramembranous turn or re-entrant loop and leaves the membrane with its C terminus at the cytosolic side. The existence of the re-entrant loop was experimentally proven by the Ubi-EMCS modification of V138C and A197C in Opi3 (Fig. 2D) and of A128C and S193C in PEMT (Fig. 3B), which flank TM IV. Compared with the PolyPhobius prediction (Fig. 1B), the model of the C-terminal structure (Fig. 4B) shifts TM IV ϳ10 amino acid residues closer to the C terminus (Fig. 4A), providing a much better fit with the hydrophobicity plot and the free energy for membrane insertion (Fig. 1A). Fig. 6 depicts the proposed topology of Opi3 based on the SCAM data (Fig. 2) and the structural model (Fig. 4).
The cytosolic localization of the C terminus of Opi3 was confirmed by DTR reporter analysis ( Fig. 2A). The accessibility to Ubi-EMCS of native cysteine Cys-54 in Opi3 and of T39C in PEMT localized in soluble domain A, and the inability to derivatize N-terminal cysteines in both proteins demonstrates the N-out topology. Remarkably, Cys-54 in WT Opi3 is not accessible to Ubi-EMCS. This may be due to a nondeleterious alteration of the structure of cytosolic loop A caused by the triple cysteine mutation in the Cys-54 mutant, resulting in the exposure of cysteine 54 to the probe.
Previous analysis of the topology of human PEMT by bioinformatics and an endoproteinase protection assay suggested that the enzyme spans the ER membrane four times with both the N terminus and the C terminus facing the cytosol (13). Except for the cytosolic localization of the C terminus, the previously proposed topology is entirely opposite to the topology presented here. We speculate that the absence of proteolytic cleavage by Lys-C in cytosolic domains A and C observed previously (13) is due to steric hindrance by the proximity of the membrane or the protein's fold. In a later study, PEMT was shown to exist as two isoforms, of which the longer (26 kDa) contained an N-terminal glycosylated domain, indicating an ER luminal localization of the N terminus (34), consistent with the present data.
Opi3 and PEMT belong to the family of ICMT orthologues also including reductases involved in ergosterol synthesis that share sequence motifs involved in the binding of the soluble cofactors SAM and NADPH. The conserved motifs PXY and EE were shown to be involved in binding SAH in the crystal structures of two ICMTs (27,28,32). In the predicted structure of Opi3's C-terminal part, these motifs attain virtually identical positions as in M.a. ICMT (Fig. 4B). Mutation of these residues abrogates Opi3 activity (Fig. 5), indicating that PXY and EE are involved in SAM binding in Opi3. Importantly, additional residues in ICMT proteins located in the most C-terminal cytosolic loop contact SAM (Fig. 4A, blue highlighted residues). Mutation of Asp-128, the potential topological counterpart of Glu-108 that contacts SAM in M.a. ICMT (27), to alanine abrogated Opi3 activity, supporting the validity of the structural model.
Opi3 harbors a GXG motif at the same position of TM3 as human PEMT, which was shown to be required for PEMT activity and SAM binding by site-directed mutagenesis (23). Interestingly, mutation of the GXG motif in Opi3 did not affect Opi3 activity. According to our Opi3/PEMT topology model, the GXG motif resides within the membrane quite distant from the PXY/EE motifs, suggesting that the deleterious effects of mutating GXG in PEMT may be indirect.
In the proposed model of Opi3 topology, the SAM-binding residues are located at the membrane-cytosol interface (Figs. 4B and 6), as expected based on the membrane peripheral localization of the methyl-accepting headgroups of PMME and PDME. The conservation of the C-terminal part corresponding to TM domain IV, soluble domain C, and the C-terminal stretch including the SAM-binding residues between different organisms (Bacteria, Archaea, yeast, insects, and mammals) and different enzymes (reductases and methyltransferase) indicates that this is a universal fold/motif involved in binding of adenosyl-containing cofactors to integral membrane enzymes converting membrane resident substrates (sterol, prenylated protein, and phospholipid) (31-33). The localization of Opi3's active site is consistent with the proposed mechanism of in Figure 6. Proposed membrane topology of Opi3. The cartoon is based on SCAM analysis and the model of the structure of the C-terminal part of Opi3 (Fig.  4). The positions of the cysteines analyzed are indicated in blue. Those marked with asterisks do not complement the choline auxotrophy of a cho2opi3 mutant (Fig. 2B). The proposed SAM-interacting amino acid residues are depicted in yellow rectangles. Alanine substitutions of Asp-128, Pro-149, Tyr-151, and Glu-188 inactivate Opi3; in case of Pro-149, we cannot exclude that inactivation is due to a decreased protein expression level (Fig. 5).

Membrane topology of Opi3
trans catalysis, as well as with conventional catalysis in cis. Further insight into the catalytic mechanism awaits elucidation of a high-resolution structure containing bound lipid substrate.

Strains, media, and culture conditions
Yeast strains and plasmids used are listed in Table 1. Yeast strains were cultured in YPD medium (1% yeast extract, 2% peptone, and 2% glucose) or glucose-based synthetic defined (SD) medium under aerobic conditions at 30°C (35). For expression of Opi3 and PEMT under the GAL1 promoter from the pYES2.1 TOPO vector, 2% galactose and 1% raffinose were used as carbon sources in SGR medium lacking uracil. Growth of Opi3 and PEMT single cysteine mutants was tested by serial dilution on synthetic defined medium plates containing 75 M inositol, lacking or supplemented with 1 mM choline (36), after preculture to mid-log phase in glucose-based SD medium containing choline and inositol.

Bioinformatics
Protein sequence alignments were performed using the Uni-Prot ALIGN server (37) or EMBOSS needle (51) as indicated. Hydrophobicity plot was generated by ExPASy ProtScale (38) using the Kyte and Doolittle scale (39), and topology predictions were generated using TOPCONS (40). ⌬G values for membrane insertion were calculated using DGPRED (41). The 3D structure of the Opi3 C terminus was predicted using the HHpred platform (42) and SWISS-MODEL (43).

Molecular biology
The sequences of all primers used in this study are listed in Table S1. All PCRs were performed using Phusion polymerase (Thermo Fisher). PCR products were purified using Wizard SV gel and PCR clean-up system (Promega). Plasmids were isolated from Escherichia coli using QIAprep spin miniprep kit (Qiagen).
The OPI3 gene was PCR-amplified using genomic DNA from BY4742 and primers A-for and A-rev and cloned into the pGEM-T easy vector (Promega) according to the manufacturer's instructions. For expression in yeast, the Opi3 ORF was PCR-amplified from the pGEM-T-OPI3 plasmid using primers B-for and B-rev, and the intron-free PEMT ORF from plasmid hPEMT-pCI (13), kindly provided by Dr. D. Vance, was amplified using primers C-for and C-rev. Both PCR products were ligated into the pYES2.1 TOPO vector (Invitrogen) in frame with the C-terminal V5-His 6 tag according to the manufacturer's instructions generating pYES2.1 TOPO-OPI3 and pYES2.1 TOPO-PEMT. Single amino acid substitutions in Opi3-V5-His 6 and PEMT-V5-His 6 were introduced by PCRmediated site-directed mutagenesis using the QuikChange kit (Agilent Technologies) and the primers listed in Table S1. Plasmids were multiplied in E. coli DH5a cells and transformed into the cho2opi3 strain (44).
DTR was added to the C terminus of Opi3 by homologous recombination of an insert containing the OPI3 gene with the linearized plasmid pJK90-OST4 carrying the OST4-HA-SUC2-HIS4C construct (a kind gift of Dr. A. Conzelmann (45)). The insert was generated using genomic DNA and the primers D-for and D-rev. Plasmid pJK90-OST4 was linearized with SmaI, and transformed with the insert in a 1:10 molar ratio into STY50 cells. The generated plasmid pJK90-OPI3 carried the construct OPI3-HA-SUC2-HIS4C under the constitutive triosephosphate isomerase promoter. Transformants were selected on Ϫura plates. The pJK90-EV plasmid lacking OST4-HA-SUC2-HIS4C was prepared by reverse PCR of pJK90-OST4 plasmid and primers E-for and E-rev. PCR product was phosphorylated using T4 polynucleotide kinase (NEB), digested with DpnI (NEB), and circularized using T4 DNA ligase (NEB).

Preparation of microsomes
cho2opi3 cells expressing Opi3 or PEMT constructs were grown to log phase (A 600 1.0) in SGR medium. The cells were harvested by centrifugation for 10 min at 3600 ϫ g, washed in 1 mM EDTA (pH 7.4), and subjected to subcellular fractionation (46). Briefly, the cells were incubated at 30°C for 10 min at 2 ml/g of wet cells in 0.1 M Tris-HCl, pH 7.4, 10 mM DTT, spun down at 3600 ϫ g for 5 min, and washed in buffer S (1.2 M sorbitol, 50 mM Tris-HCl, pH 7.4). The cells were resuspended at 3 ml/g of wet cells in buffer S supplemented with 0.5 mg of Zymolyase (Amsbio, 100 kU/mg) per g of cells and incubated at 30°C for 30 min. Spheroplasts were diluted with ice-cold buffer S and collected by centrifugation at 3600 ϫ g for 5 min at 4°C. Pelleted spheroplasts were resuspended in ice cold 0.6 M sorbitol, 10 mM Tris-HCl, pH 7.4, containing protease inhibitors (cOmplete protease inhibitor mixture, Roche; 1 tablet/50 ml), and homogenized by 10 strokes with a Dounce homogenizer on ice. Unbroken cells and nuclei were removed by centrifugation at 3600 ϫ g for 5 min at 4°C.
The cell homogenate (supernatant) was centrifuged at 12,000 ϫ g for 15 min to remove mitochondria. The microsomal fraction was isolated by ultracentrifugation of the supernatant at 100,000 ϫ g for 1 h at 4°C. The pellet was resuspended in buffer A (50 mM Tris-HCl, pH 7.5, 0.6 M sorbitol), flash-

Membrane topology of Opi3
frozen as aliquots in liquid nitrogen, and stored at Ϫ80°C. Aliquots were thawed only once. Microsomal protein concentration was determined using the BCA assay (Pierce) in the presence of 0.1% SDS and using BSA as a standard.

Determination of protein topology
Synthesis of ubiquitin-EMCS and cysteine scanning-Ubi-EMCS was prepared according to Pagac et al. (24). Briefly, 2.3 mM ubiquitin (Sigma) and 3.35 mM sulfo-EMCS (6-maleimidocaproic acid sulfo-n-succinimidyl ester; Thermo Fisher Scientific) dissolved in 50 mM sodium phosphate, pH 7.4, were mixed in a ratio 3:2 (v/v), respectively, in a total volume of 500 l and incubated for 40 min at room temperature on a rotary wheel. The product was washed and concentrated in the same buffer using a Centricon centrifugal filter unit (3000 molecular weight cutoff). Sorbitol was added to 0.6 M, yielding a final concentration of Ubi-EMCS of 4.5 mM (assuming 100% reaction yield), and aliquots were snap-frozen and stored at Ϫ20°C. For cysteine scanning, microsomes (15 g of microsomal protein) were mixed with 45 nmol of Ubi-EMCS in 0.1 M sodium phosphate, pH 7.4, 0.6 M sorbitol, 5 mM MgCl 2 with or without 1% SDS in a total reaction volume of 50 l and incubated for 1 h on ice. For each series of reactions, a freshly thawed aliquot of Ubi-EMCS was used. The reaction was quenched by adding 5 l of 1 M DTT and incubating for 5 min at room temperature. Next, SDS-PAGE loading buffer was added, and samples were incubated at 56°C for 10 min. Proteins (15 g) were separated on 15% SDS-polyacrylamide gels and analyzed by Western blotting using anti-V5 monoclonal antibodies (Thermo Scientific), or anti-Kar2p.
Dual topology reporter analysis-STY50 yeast cells transformed with pJK90-EV or pJK90-OPI3 encoding Opi3-DTR were serially diluted on SD Ϫura and SD Ϫura Ϫhis plates containing 6 mM histidinol and incubated for 4 days at 30°C. To examine glycosylation of the DTR constructs, heavy microsomes were isolated from cells cultured on SD Ϫura medium. Microsomes corresponding to 10 g of total membrane protein were prepared according to the manufacturer's instructions prior to incubation with 1000 units of Endo H (NEB) for 2 h at 37°C with gentle agitation. The samples were separated on an 8% SDS-polyacrylamide gel and visualized on Western blotting using anti-HA antibodies (Abcam).

Measurement of methyltransferase activity
Enzyme activity of Opi3 and PEMT was assayed as the amount (nmol) of [ 3 H]methyl group from S-adenosyl-L-[methyl-3 H]methionine ([ 3 H]SAM) incorporated into chloroform soluble material per min per mg protein (14,18). Large unilamellar vesicles (LUVs) consisting of 1,2-dioleoyl-sn-glycero-3phosphoethanolamine-N-methyl (PMME; Avanti Polar Lipids) served as lipid substrate and were prepared by hydrating a lipid film in buffer A followed by 10 freeze-thaw cycles (in dry ice/ ethanol and a water bath at 30°C) and 20ϫ extrusion through a 200-nm pore-size Whatman nucleopore track-etched polycarbonate (hydrophilic) membrane (GE Healthcare) using the Avanti mini-extruder set. A 100-l reaction mixture containing microsomes (0.05 mg protein/ml), LUV (0.4 mM PMME), and 1 mM SAM (Sigma) containing 5000 dpm [ 3 H]-SAM/nmol (10 Ci/mmol; PerkinElmer Life Sciences) in buffer A was incubated at 30°C in a shaking thermomixer (650 rpm) for 10 min. The reaction was stopped by adding 375 l of CHCl 3 :methanol:0.1 M HCl, 10:5:1 (v/v/v), followed by lipid extraction (47). The incorporated [ 3 H]methyl was quantitated by liquid scintillation counting. The phospholipid concentrations of the PMME stock solution in chloroform, and the LUV preparations were determined according to Rouser et al. (48).