A Six-membrane-spanning Topology for Yeast and Arabidopsis Tsc13p, the Enoyl Reductases of the Microsomal Fatty Acid Elongating System*

The very long chain fatty acids are crucial building blocks of essential lipids, most notably the sphingolipids. These elongated fatty acids are synthesized by a system of enzymes that are organized in a complex within the endoplasmic reticulum membrane. Although several of the components of the elongase complex have recently been identified, little is known about how these proteins are organized within the membrane or about how they interact with one another during fatty acid elongation. In this study the topology of Tsc13p, the enoyl reductase of the elongase system, was investigated. The N and C termini of Tsc13p reside in the cytoplasm, and six putative membrane-spanning domains were identified by insertion of glycosylation and factor Xa cleavage sites at various positions. The N-terminal domain including the first membrane-spanning segment contains sufficient information for targeting to the endoplasmic reticulum membrane. Studies of the Arabidopsis Tsc13p protein revealed a similar topology. Highly conserved domains of the Tsc13p proteins that are likely to be important for enzymatic activity lie on the cytosolic face of the endoplasmic reticulum, possibly partially embedded within the membrane.

The very long chain fatty acids are crucial building blocks of essential lipids, most notably the sphingolipids. These elongated fatty acids are synthesized by a system of enzymes that are organized in a complex within the endoplasmic reticulum membrane. Although several of the components of the elongase complex have recently been identified, little is known about how these proteins are organized within the membrane or about how they interact with one another during fatty acid elongation. In this study the topology of Tsc13p, the enoyl reductase of the elongase system, was investigated. The N and C termini of Tsc13p reside in the cytoplasm, and six putative membranespanning domains were identified by insertion of glycosylation and factor Xa cleavage sites at various positions. The N-terminal domain including the first membrane-spanning segment contains sufficient information for targeting to the endoplasmic reticulum membrane. Studies of the Arabidopsis Tsc13p protein revealed a similar topology. Highly conserved domains of the Tsc13p proteins that are likely to be important for enzymatic activity lie on the cytosolic face of the endoplasmic reticulum, possibly partially embedded within the membrane.
Cytosolic fatty acid synthases catalyze the de novo synthesis of the 16 or 18 carbons containing fatty acids that are further elongated to very long chain fatty acids by a microsomal enzyme system, the elongase. Very long chain fatty acids are essential components of cuticular waxes and seed triacylglycerols in plants and of several classes of lipids, including the sphingolipids. Sphingolipids function in eukaryotic cells as membrane structural components, in cell interactions with surroundings, and as bioactive molecules involved in signaling and cell regulation. Sphingolipids also interact with sterols to form lipid rafts, which are involved in trafficking of plasma membrane proteins, endocytosis, and protein stability at the cell surface (1)(2)(3).
The endoplasmic reticulum (ER) 2 -associated elongase system is composed of four distinct enzymes that sequentially catalyze condensation between a CoA-esterified fatty-acyl substrate and malonyl-CoA, a 3-ketoacyl-CoA reduction, a 3-hydroxyacyl-CoA dehydration, and a final enoyl-CoA reduction to yield a fatty acid that is two carbon units longer than the primer (4). Several of the genes encoding components of the elongase system were first identified in Saccharomyces cerevisiae (5)(6)(7)(8)(9) and later in plants and mammals (10 -16). Several studies indicate that the elongase proteins are organized in a complex within the ER (5,7,17). A complete understanding of the molecular mechanism and organization of the elongase complex will require structural analysis, a challenging prospect because of the intrinsic technical difficulties associated with the purification and crystallization of membrane proteins. However, in the absence of high resolution structural data, detailed topology models aid in the design and interpretation of structure-function studies of membrane proteins. Thus, as a first step toward elucidating the organization of the elongase complex we have undertaken the topological mapping of the component proteins. Here we present our studies on the topological organization of the S. cerevisiae and Arabidopsis thaliana Tsc13p proteins, the enoyl-CoA reductases of the elongase complex.
TSC13 was identified in a screen for mutants that suppress the Ca 2ϩ sensitivity of csg2⌬ cells (18). Csg2p is required for conversion of inositolphosphoceramide (IPC) to mannosylinositolphosphoceramide, and we discovered several years ago that the over-accumulation of IPC in the csg2⌬ cells results in calcium sensitivity (18,19). Furthermore, mutations that reduce inositolphosphoceramide (IPC) accumulation were found to suppress the Ca 2ϩ sensitivity, including mutations in the genes required for the very long chain fatty acids component of IPC (18,19). The mammalian (14) and plant (10,15) orthologs of Tsc13p have also been identified and characterized.
In the present study we provide evidence that the N and C termini of both the yeast and Arabidopsis Tsc13p proteins reside in the cytosol and that these proteins contain six membrane-spanning domains. Based on this model, two highly con-served and functionally critical residues are found to be associated with the membrane, raising the possibility that the active site of the enzyme is partially embedded within the lipid bilayer.
Disruption of TSC13 Gene and Construction of Tagged Yeast and Arabidopsis TSC13-Construction of the tsc13⌬ mutant and the pTSC13-316 and pMYC-TSC13-426 plasmids were described previously (7). The plasmid expressing the dual epitope-tagged Myc-Tsc13p-HA was generated by moving the 3ϫ-MYC-TSC13 cassette from pMYC-TSC13-426 to pRS425, introducing an in-frame SpeI site between codon 309 and the stop codon of the TSC13 gene by QuikChange mutagenesis (Stratagene) and ligating a SpeI-ended 3ϫ-HA cassette into the SpeI site.
To generate the 1-117-GFP fusion protein, the MYC-TSC13-425 plasmid containing the SpeI at position 309 was digested with SpeI and NheI to remove the fragment encoding amino acids 117-309 of Tsc13p. The remaining vector fragment was ligated with an XbaI-ended fragment encoding GFP (5), and the construct was verified by DNA sequencing. A yeast plasmid constitutively expressing the 3ϫ-HA-N-terminaltagged Arabidopsis gene, At3g55360 (designated hereafter as AtTSC13) under control of the ADH promoter, was generated as described previously (15).
TSC13-GC Topology Reporter Plasmids-The TSC13 gene fusion alleles with the invertase glycosylation cassette (Suc2A or GC) inserted were constructed in two steps. In the first step, in-frame SpeI restriction sites (encoding threonine and serine) were introduced at 12 positions (indicated in Fig. 1) within the TSC13 gene. For each mutagenesis reaction, the MYC-TSC13-425 plasmid was used as template, and a pair of mutagenic primers containing the SpeI site flanked by 12 nucleotides of homology to TSC13 was used. After mutagenesis and transformation into Escherichia coli, plasmids were isolated and screened for the presence of the SpeI site. In the second step the SpeI-ended GC (the 53amino acid domain from the Suc2p protein that has 3 sites for N-linked glycosylation) was inserted into the SpeI sites as well as into an in-frame NheI site at position 117. The DNA sequences of all the constructs were verified by sequencing. The GC cassette was inserted at seven different locations along the AtTSC13 gene ( Fig. 1) using the same strategy. Construction of TSC13-fXa Fusion Alleles-fXa cleavage sites were inserted using a pair of oligonucleotides encoding a repeat of the fXa protease recognition site (IEGR) that would anneal to leave overhanging ends compatible for ligation into the SpeI sites of pMYC-TSC13-425 (described above). The oligonucleotides were annealed by mixing, heating at 95°C for 5 min, and cooling on ice. The TSC13-fXa fusion alleles were verified by DNA sequencing.
Site-directed Mutagenesis-The tsc13 mutant alleles with charged residues substituted by alanine were generated by QuikChange mutagenesis (Stratagene). All mutants were confirmed by DNA sequencing.
Complementation Assay-The various Tsc13p-GC, Tsc13p-fXa, alanine substitution mutant, and epitope-tagged Tsc13p proteins were assessed for function by determining whether their expression would complement the lethality of the tsc13⌬ mutant. Plasmids expressing the modified tsc13 alleles were introduced into pTSC13-316-rescued tsc13⌬ mutant cells, and the transformants were tested for their ability to lose the URA3 ϩ -marked pTSC13-316 rescuing plasmid. Transformants containing plasmids expressing functional Tsc13p proteins were able to lose the URA3 ϩ -marked plasmid and were, therefore, able to grow on 5-fluoroorotic acid (FOA)-containing plates.
Determination of the Glycosylation Status of the Tsc13p-GC Fusion Proteins-For analysis of the glycosylation status of the Tsc13p-GC proteins, microsomes were prepared from cells grown in minimal medium (minus leucine) to ensure maintenance of the plasmid. The cells were pelleted, washed with water, resuspended at 2 ml/g wet weight, and lysed by bead beating in 50 mM Tris, 7.5, 1 mM EGTA, 1 mM ␤-mercaptoeth-

Sites of insertion of the GC cassettes into the yeast and Arabidopsis
Tsc13p. An alignment of the human (Human), plant (A. thaliana (At)), and yeast (S. cerevisiae (Sc)) Tsc13ps was generated using ClustalW (36) and used to predict locations within the protein that might accommodate insertion of the GC without disrupting function. The residues after which the GC was inserted are in green for functional proteins, blue for nonfunctional but stable proteins, or red for unstable proteins. The highly conserved residues are shaded black. The blue asterisks mark the residues identified as critical for function in this study, and the orange asterisk (Gln-81) marks a residue previously shown to be important for function (7). anol, 1 M aprotinin, 1 M phenylmethylsulfonyl fluoride, and 1 M leupeptin. Glass beads were added to the meniscus, and cells were lysed by four cycles (60 s each) of vortexing with cooling on ice between cycles. Unbroken cells, beads, and debris were removed by centrifugation (5000 ϫ g for 10 min), and the low speed supernatant was centrifuged at 100,000 ϫ g for 30 min at 4°C to provide the microsomal pellet. The pellet was resuspended in the same buffer, repelleted at 100,000 ϫ g for 30 min at 4°C, and resuspended at ϳ10 mg/ml protein in the same buffer containing 33% glycerol. The glycosylation status of the Tsc13p-GC proteins was analyzed as described previously (21,22) with slight modifications. In brief, 100 g of microsomal protein was suspended in 54 l of 20 mM Tris-HCl, pH 7.5, 5 mM Na 2 EDTA, 50 mM NaCl. Six l of 10ϫ denaturing buffer (5% SDS, 10% ␤-mercaptoethanol) were added, and the samples were incubated at 70°C for 10 min. Ten l of 0.5 M sodium citrate, pH 5.5, were added to the heated samples to bring the volume to 70 l, and the samples were split into two aliquots. One aliquot was treated with 5000 units of endoglycosidase H (Endo H; New England Biolabs), and the other aliquot was incubated with buffer at 37°C for 60 min and then further incubated at 70°C for 10 min with denaturing buffer (5% SDS, 10% ␤-mercaptoethanol). After resolution by 12% SDS-PAGE, the proteins were analyzed by immunoblotting.
Right-side-out Microsomal Vesicle Preparation and Cleavage with Factor Xa Protease-Preparation of microsomes for analyzing the fusion proteins containing the fXa cleavage sites was performed essentially as described previously (21,22) with minor modifications. Spheroplasts were generated and lysed with lysis buffer, and the homogenates were cleared of debris. The microsomal membrane fraction was recovered by centrifugation at 100,000 ϫ g for 30 min at 4°C, washed once with storage buffer (20 mM Tris-HCl, pH 7.5, containing 250 mM sorbitol, 50 mM potassium acetate, and 1 mM ␤-mercaptoethanol), and resuspended in storage buffer at 4 mg/ml protein. The fXa protease cleavage assay was performed as described (22), and the products were separated by SDS-PAGE and detected by immunoblotting.
Protease Protection Assay-Right-side-out microsomes were prepared from yeast expressing the HA-TSC13-MYC fusion protein. Protease protection assays were performed as described (23), and the samples were resolved by SDS-PAGE and analyzed by immunoblotting.
Cell Fractionation-Microsomes were prepared by bead beating (17) and incubated on ice for 1 h with an equal volume of buffer containing either 1 M sodium chloride, 0.4 M sodium carbonate, or 2% Triton X-100. After centrifugation of the samples at 100,000 ϫ g for 30 min at 4°C, the supernatant and the pellet fractions were collected and subjected to SDS-PAGE and immunoblotting.

Membrane Topology Predictions for Tsc13p-Tsc13p
, the enoyl reductase of the fatty acid elongation system, is an integral ER membrane protein. Determination of the topology of Tsc13p is an important step toward understanding both its mechanism and its organization within the elongase complex. As is often the case, various algorithms for predicting membrane-spanning segments suggested several possible topologies for Tsc13p. SOSUI, a program that takes into account the Kyte Doolittle hydropathy, amphiphilicity, amino acid charge, and the length of the protein (24), predicted a single transmembrane domain (TMD) located between amino acids 262-283. In contrast, six TMDs were predicted by both Localizome (86 -105, 125-143, 164 -184, 204 -226, 247-266, and 272-291) and HMMTOP (88 -112, 145-162, 169 -186, 203-220, 251-269, and 274 -291). Localizome uses hmmpfam to detect the presence of Pfam domains (25) and a prediction algorithm, Phobius, to predict the TM helices. The results are combined and checked against the TM topology rules stored in a protein domain data base called LocaloDom (26). HMMTOP (Hidden Markov Model for Topology Prediction) is based on the principle that topology of TM proteins is determined by the difference in amino acid distribution in various structural parts of these proteins rather than by specific amino acid composition (27,28). TMHMM, which is also based on a hidden Markov Model (29), predicts four TMDs (166 -188, 202-222, 243-265 and 269 -291). In the experiments described below, these different models for Tsc13p topology were experimentally evaluated.
Mapping the Orientation of the Tsc13p Termini by Protease Protection-To determine the orientation of the N-and C-terminal ends of Tsc13p, the N terminus was tagged by inserting a Myc epitope after the start codon, and the C terminus was tagged by inserting an HA epitope before the stop codon. To ascertain that insertion of the epitope tags did not alter the structure of Tsc13p, the function of the epitope-tagged protein was tested. TCS13 is an essential gene in S. cerevisiae (7), and thus, the plasmid carrying the wild-type TCS13 gene is required for survival of the haploid deletion strain. A LEU2-marked plasmid expressing Myc-Tsc13p-HA was transformed into the rescued tsc13⌬. FOA-resistant colonies that had lost the URA3marked TSC13 plasmid were recovered, indicating that insertion of the N-and C-terminal epitope tags did not disrupt the function of Tsc13p ( Fig. 2A). In addition, immunofluorescence localization of the Myc-Tsc13p-HA protein revealed perinuclear and peripheral ER staining (Fig. 2B) similar to that previously observed for wild-type Tsc13p (7).
To localize the N and C termini of Tsc13p with respect to the ER membrane, protease protection assays were conducted. Sealed right-side-out membrane vesicles were isolated by gently lysing yeast spheroplasts prepared from cells expressing Myc-Tsc13p-HA, and sensitivity of the protein to proteinase K in the presence or absence of detergent was examined. The vesicles were treated with proteinase K in the presence or absence of detergent, and the Myc-Tsc13p-HA was analyzed by immunoblotting with either anti-Myc or anti-HA antibodies. The immunoblots indicated that both the N-terminal Myc tag and the C-terminal HA tag were sensitive to proteinase K even in the absence of detergent as no protected immunoreactive protein fragments were detected in either case (Fig. 3). If the N terminus had been lumenal, a protected Myc fragment should have been observed, and similarly, if the C terminus were in the ER lumen, a protected HA-tagged fragment would have been seen. This indicates that both the N and C termini of Tsc13p are cytosolically oriented and that Tsc13p has an even number of membrane spanning domains. The integrity of the sealed rightside-out vesicles was confirmed by the observation that the ER lumenal Kar2p protein was insensitive to proteinase K until the vesicles were disrupted with detergent (Fig. 3). As has been previously reported, in the presence of detergent proteinase K clipped Kar2p to a smaller size but did not completely degrade it (23). The sidedness of the vesicles prepared using this procedure was previously confirmed by demonstrating that fXa protease cleavage sites inserted into the lumenal domains of the integral membrane protein, Lcb1p, were inaccessible to the fXa protease unless the vesicles were disrupted with detergent (22).
The Native Glycosylation Sites at Residues 38 and 255 of Tsc13p Are Not Modified-A well established approach to map the topology of ER membrane proteins takes advantage of the lumen specific glycosylation machinery. The glycosylation status of the two potential N-linked glycosylation sites (NX(S/T)) at residues 38 and 255 of Tsc13p was evaluated. As shown in Fig. 4A, treatment with Endo H did not alter the electrophoretic mobility of Myc-Tsc13p, demonstrating that the potential N-linked glycosylation sites in the native Tsc13p sequence are not glycosylated. The first predicted TMD lies between residues 86 and 110; therefore, the lack of modification of the potential glycosylation site at 38 is consistent with the protease sensitivity results and provides further evidence that the N terminus of Tsc13p is cytosolic. The lack of glycosylation at 255 indicates that this residue is also not in a lumenal loop that is accessible to the glycosylation machinery. Several of the hydropathy algorithms (discussed above) place residue 255 within a TMD, which would explain the lack of modification of this potential glycosylation site.
Analysis of Tsc13p-GC Fusion Proteins Reveals Several Membrane-associated Domains-To further investigate the topology of Tsc13p, a set of fusion proteins containing a glycosylation reporter cassette (GC) inserted in-frame at 13 positions along the length of the protein was constructed (Fig. 4A). The cassette consists of a 53-amino acid domain comprising residues 80 -133 of invertase (Suc2p) that contains three NX(S/T) sites for N-linked glycosylation. These sites are located far enough from the N-and C-terminal ends of the cassette to ensure that if the GC is inserted anywhere in the lumenal loop of a fusion protein, the acceptor sites will be sufficiently far from the membrane to be recognized by the glycosylation machinery (21). The locations within Tsc13p into which the cassettes were inserted were chosen based on their ability to distinguish the FIGURE 2. Myc-Tsc13p-HA is functional and ER localized. A, the tsc13⌬ mutant is unable to lose the pTSC13-316 plasmid and is, therefore, unable to grow on FOA. Introduction of the LEU2-marked plasmid expressing the epitope-tagged MYC-Tsc13p-HA protein (but not the empty vector) allowed the tsc13⌬ mutant to lose the URA3-marked pTSC13-316 plasmid and to grow on FOA. B, indirect immunofluorescence shows that Myc-Tsc13p-HA localizes to the ER. Fixed and permeabilized cells were incubated with anti-Myc antibody, and the primary antibody was detected using Cy3-conjugated goat anti-mouse as the secondary antibody. different topological models predicted by the hydropathy analyses. In addition, alignments of Tsc13p proteins from different species were used in an attempt to identify sites with low conservation across evolution that might be sufficiently flexible to tolerate an insertion without disrupting function (Fig. 1). This reporter cassette has been successfully used in determining the topology of other membrane proteins (21,22,30), and its insertion into lumenal loops between membrane-spanning domains has not been observed to interfere with their insertion or orientation.
The Tsc13p-GC fusion proteins were evaluated for function by testing their ability to restore viability to the tsc13⌬ mutant. Tsc13p-GC fusions with the Suc2p domain inserted at eight different locations were functional (Fig. 4A), indicating that these proteins retained their native conformation. Tsc13p-GC fusions with the cassette inserted at position 135 or 166 resulted in unstable proteins. Although insertion of the cassette at position 159 or 170 resulted in stable proteins, these fusion proteins were not functional (Fig. 4A). Furthermore, indirect immuno-fluorescence revealed that these two fusion proteins failed to localize to the ER (data not shown). The Tsc13p-GC fusion protein with the cassette at position 273 was also not functional (Fig. 4A), but this protein displayed normal ER localization (data not shown).
The Tsc13p-GC fusion proteins with the cassette inserted at 117, 126, 198, or 273 were glycosylated as indicated by increased electrophoretic mobility after Endo H treatment, and therefore, these regions of Tsc13p are localized in the lumen of the ER. Moreover, the electrophoretic mobilities of the fusion proteins with the GC located at position 7, 69, 226, 236, or 309 were not altered after treatment with Endo H, demonstrating that these regions of Tsc13p are located in the cytosol (Fig. 4A). Although the fusion proteins with the GC at 159 and 170 were not glycosylated because they were not functional and did not localize properly, it is not possible to definitively conclude that they are normally cytosolic. Although these analyses clearly reveal the presence of at least two lumenal loops in Tsc13p, it is not clear from these data whether the segment between 126 and 198 forms a single large lumenal loop or whether there are additional membrane associated domains within this region of Tsc13p.
Analysis of Tsc13p Topology by fXa Insertion-To further investigate the topology of Tsc13p, tandem fXa protease cleavage sites (IEGRIEGR) were inserted at several positions (31), and their accessibility to fXa protease in sealed right-side-out membrane vesicles was assessed. A tandem recognition sequence was inserted to increase the probability of cleavage by the fXa protease, which cuts on the C-terminal side of the arginine in the IEGR tetrapeptide. The Tsc13p-fXa fusion proteins were expressed in tsc13⌬ mutant cells, sealed right-side-out vesicles were prepared, and the sensitivity to fXa protease cleavage was assayed in the absence or presence of detergent. The integrity of the vesicle preparations was verified by showing that the lumenal ER protein, Kar2p, was accessible to proteinase K only in the presence of detergent (discussed above). The fXa sites inserted at positions 69, 226, or 236 were accessible to fXa protease whether or not Nonidet P-40 (Nonidet P-40) was present (Fig. 4B). This result is consistent with the lack of glycosylation of the GC cassettes inserted at the same positions and confirms the cytoplasmic orientation of these regions of the protein. Furthermore, the protease recognition sites at 117, 198, or 273 were only accessible to the protease when Nonidet P-40 was present (Fig. 4B), a result that is also consistent with the glycosylation experiments.
Taken together, the glycosylation and fXa protease cleavage results suggest that Tsc13p contains at least four membrane spanning domains located between amino acids 69 -117, 198 - Each circle denotes a single amino acid, the blue circles denote the residues where the inserted GC was not glycosylated, and the black filled circles denote the residues where the GC was glycosylated. The conserved residues that were substituted with alanine are marked as green circles if the mutation did not disrupt function or as red circles if the mutation did disrupt function. The three cysteines (at 90, 165, and 220) are marked as yellow circles, and the tryptophan residues are marked as purple circles. The Gln-81 residue that is mutated to Lys in a reduced function allele of TSC13 is indicated by the orange circle. The two intrinsic N-linked glycosylation sites (Asn-38 and Asn-255) that were not glycosylated are indicated as Ns. The residues predicted to lie at the ends of the TMDs are indicated. The uncertainty as to the location of TMD2 is indicated by the dashed line. The two highly conserved regions of Tsc13p are marked by the blue shading. 226, 236 -273, and 273-309 and that the N and C termini are cytosolic (Fig. 4C). This agrees with the Localizome and HMMTOP algorithms, which predicted the presence of TMDs between amino acids 88 -108, 204 -224, 248 -268, and 270 -290. It is likely that there are two additional TMDs located between 126 and 198, but this could not be demonstrated using these methods because the insertion of GC and fXa recognition sequences into this region of Tsc13p resulted in unstable, nonfunctional, and/or mislocalized proteins. The possibility that there are additional membrane domains between 126 and 198 is addressed below.
Evidence for Additional TMDs between Residues 126 and 198 of Tsc13p-To determine whether the segment of Tsc13p between residues 126 and 198 is a large lumenal domain or whether it contains additional membrane associated segments, fXa cleavage sites were inserted after residue 117 and after residue 200 such that this fragment could be cut out of Tsc13p and its membrane association could be assessed. The fXa-flanked segment of Tsc13p also had an HA tag inserted at 198 for immunodetection (Fig. 5A). This HA-tagged Tsc13p with the two fXa cleavage sites retained the ability to rescue the tsc13⌬ mutant, and therefore, it was assumed to adopt its native topology. Microsomes were prepared from the tsc13⌬ mutant cells expressing the tagged Tsc13p using a bead beating procedure that generates inverted vesicles. The segment between residues 117 and 200 was liberated from the protein with fXa protease, the microsomes were subjected to high speed centrifugation, and the pellets and supernatants were analyzed for the presence of the HA-tagged fragment (Fig. 5B). Four distinct bands were detected with the anti-HA antibody, the sizes of which were consistent with the full-length tagged Tsc13p, the products from cleavage at one or the other of the fXa sites, and the 117-200-amino acid fragment derived from cleavage of both fXa sites. After high speed centrifugation, the liberated 117-200 fragment of Tsc13p was found exclusively in the pellet. Furthermore, the 117-200 fragment of Tsc13p was found to be integrally associated with the membrane because 1% Triton X-100, but not 0.5 M NaCl or 0.1 M sodium carbonate, solubilized it (Fig. 5B). These results clearly showed that the 117-200 fragment of Tsc13p is not a lumenal loop, but rather, that it contains additional membrane-associated domains. Taken together, the data are, therefore, most consistent with the presence of six-membrane-spanning domains in Tsc13p (Fig. 5C).
The N-terminal Membrane-spanning Domain of Tsc13p Is Sufficient for ER Localization-Polytopic membrane proteins are initially targeted to the ER by a signal sequence that may include the first TMD, and subsequent TMDs often contain information that directs insertion and contributes to the topogenesis (32). To determine whether the first hydrophobic segment of Tsc13p contains sufficient information to direct membrane association, a chimeric protein with the N terminus including TMD1 of Tsc13p (amino acids 1-117, with a Myc tag after the first codon) was fused to GFP. When expressed in yeast, this chimera showed a typical ER localization pattern (Fig. 6A) similar to that of wild-type Tsc13p (Fig. 2B). The Tsc13p1-117-GFP chimeric protein fractionated with membranes and behaved as an integral membrane protein in that it could only be solubilized with detergent (Fig. 6B). Taken together, the fluorescence localization and solubilization studies show that the N-terminal 117 amino acids of Tsc13p, which contains the first TMD, has sufficient information to direct ER membrane targeting and association.
The Arabidopsis Tsc13p Is Also a Six-membrane-spanning Protein-Comparison of the hydropathy predictions of the S. cerevisiae Tsc13p with the A. thaliana orthologue AtTSC13, encoded by the At3g55360 gene that was previously shown to complement the yeast tsc13⌬ mutant and to be involved in very long chain fatty acids synthesis in plants (10,15), suggested that the proteins are likely to have the same topologies. To investigate the topology of the AtTSC13 protein, HA-AtTSC13 and several HA-AtTSC13-GC-fusion proteins were expressed in yeast, and their glycosylation was investigated. Neither of the two endogenous glycosylation sites located at Asn-76 and Asn-244 was glycosylated nor were the GCs inserted after amino acid 80, 159, 218, or 310. On the other hand, the GCs inserted after amino acids 122, 182, and 258 were glycosylated since their electrophoretic mobilities increased after Endo H treatment (Fig. 7A). These results are entirely consistent with the presence of six membrane-spanning domains in AtTSC13 and, thus, indicate that the Arabidopsis and yeast proteins have similar topologies (Figs. 1, 5C, and 7B).
Alanine Substitutions of Conserved Charged Residues in or near the TMDs of Tsc13p-A comparison of the amino acid sequences of Tsc13p homologs from Arabidopsis, Homo sapiens, and S. cerevisiae revealed the presence of several highly conserved amino acids including several charged residues near or within the predicted TMDs (Figs. 1 and 5C). To examine the physiological significance of several of these amino acids, alanine substitution mutants were generated and tested for their ability to complement tsc13⌬.
Despite their high conservation, several of these residues were not critical for function since the alanine substitution mutants, K76A, D77A, Y103A, H137A, E144A, H149A, and . After a second high speed centrifugation, 10 g of protein from the supernatant (S) and pellet (P) was subjected to 12% SDS-PAGE, and the chimeric protein was detected by immunoblotting with an anti-Myc antibody. E259A (Fig. 8A) and E91A (data not shown) were able to complement the tsc13⌬ mutant. On the other hand, the Y138A, K140A, and R141A mutants were not functional (Fig. 8A). Immunoblot analysis revealed that the Y138A mutant protein was unstable, but the K140A and R141A mutant proteins were present at similar levels to wild-type Tsc13p (data not shown).
The Lys-140 and Arg-141 residues are predicted to lie in or near the cytosolic end of TMD2 (Fig. 5). It was, therefore, of interest to determine whether their mutation altered the topology of Tsc13p. This was investigated by introducing the K140A and R141A mutations into the Tsc13p-GC reporter proteins with the GC at 117, 198, and 226. The results indicated that these mutations did not alter the membrane topology of Tsc13p (Fig. 8B). Thus, it appears that although the Lys-140 and Arg-141 residues are critical for function, they are not required for proper topogenesis of Tsc13p.
Tsc13p has three cysteine residues at 90, 165, and 220, of which Cys-165 is conserved. Individual substitution of each cysteine to serine did not disrupt function nor did the simultaneous substitution of all three residues as the triple cysteineless mutant was also able to complement tsc13⌬ (data not shown).

DISCUSSION
The topology of Tsc13p, the enoyl-CoA reductase of the ERassociated elongase system, was investigated by several approaches. Because the exact features of membrane proteins that dictate how they insert into membranes are not fully understood, the available algorithms for predicting membrane topologies serve only as guides for the actual topology. Various programs predicted different numbers of membrane-spanning domains (1, 4, or 6) for Tsc13p. Our results are most consistent with a six-membrane-spanning model which places both the N and the C termini of Tsc13p in the cytosol. We also evaluated the topology of the Arabidopsis Tsc13p ortholog and find that it has a similar membrane topology to that of the yeast Tsc13p.
The accessibility of the epitope tags at both ends of Tsc13p to proteinase K in sealed right-side out vesicles together with the absence of any protected epitope-tagged fragments suggested that both ends are cytosolically oriented. The absence of glycosylation at Asn-38, which is a potential site for glycosylation, provided further evidence that the N terminus is cytosolic. However, glycosylation sites can be obscured by their proximity to the membrane or by association with other proteins (33); to rule out the possibility that Asn-38 might reside in the lumen but escape glycosylation, we utilized the Suc2p glycosylation cassette. This 53-amino acid domain contains three N-linked glycosylation sites that are located far enough from the N-and C-terminal ends of the cassette to ensure their recognition by the glycosylation machinery. Cassettes inserted at the N (positions 7 and 69) and C (position 309) termini were not glycosylated, confirming the conclusion from the protease protection studies. Assignment of the C terminus to the cytosol is also consistent with two other studies. In an effort to determine the topology of the S. cerevisiae membrane proteome (30), a topological reporter cassette (Suc2p/His4C) was fused at the C terminus of many membrane proteins, including Tsc13p. The lack of glycosylation along with catalytic activity of the His4Cp domain indicated that the reporter at the C terminus of Tsc13p was cytosolic. In another study aimed at identifying interactions between yeast membrane proteins (34), Tsc13p with a split ubiquitin domain fused at the C terminus was found to interact with other appropriately tagged elongase proteins. Such an interaction is reflected by cleavage of the reconstituted ubiquitin by a cytosolic protease and, therefore, further confirms that the C terminus of Tsc13p is in the cytosol.
Our glycosylation and factor Xa results provide strong evidence that the regions of Tsc13p comprised of amino acids 85-110, 202-222, 243-269, and 274 -291 contain membrane  (AtTSC13). A, the glycosylation status of several AtTSC13-GC fusion proteins was assessed as described above for the yeast Tsc13p-GC fusion proteins. Briefly, microsomes were prepared from yeast expressing the indicated AtTSC13p-Suc2p fusion proteins, 10 g of microsomal protein (with or without Endo H treatment) were resolved by 12% SDS-PAGE and transferred to nitrocellulose, and the fusion proteins were detected using anti-HA antibodies. V, vector; WT, wild type. B, proposed six-membrane-spanning topology model for the A. thaliana TSC13 ortholog. The circles represent the amino acids after which the GC was inserted.  Fig. 5C) were mutated to alanine, and the mutant Tsc13p proteins were tested for function by the complementation assay (see "Experimental Procedures"). Wt, wild type. B, the glycosylation status of the Tsc13p-GC fusion proteins carrying the K140A and R141A mutations was assayed. Microsomes were prepared from yeast harboring Tsc13p-GC fusion proteins (GC after amino acid 117, 198, or 226) with either the K140A or R141A mutation as indicated. 10 g of microsomal protein (with or without Endo H treatment) was resolved by 12% SDS-PAGE, and Tsc13p proteins were detected by immunoblotting with anti-Myc. spanning domains. Because the majority of the topological reporter cassette-containing fusion proteins were capable of complementing the tsc13⌬ mutant, it is likely that these proteins are adopting the native Tsc13p topology. Moreover, the cytosolically oriented N-terminal 85-amino acid tail together with the first predicted membrane-spanning domain located between residues 85 and 110 contains sufficient information to target a cytosolic GFP to the ER, providing clear evidence for the presence of a transmembrane domain, TMD1, in this region.
Because all insertions between residues 126 and 198 resulted in mislocalization and/or disruption of function, it was unclear whether this segment of Tsc13p represented a long lumenal domain or whether, as predicted by several algorithms, there were additional membrane-spanning domains in this region. Because there is one TMD before residue 126 and three after residue 198 and both ends of Tsc13p are cytosolic, the presence of any additional TMDs in this region would require an even number, most likely two. We provided evidence for additional membrane-associated segments within this region by proteolytically cleaving it from the full-length Tsc13p and showing that the released fragment remained associated with the membrane fraction. The membrane association of this liberated fragment is apparently not a result of interaction with other membrane proteins because it was stable to high salt and pH but could be solubilized with detergent.
The inability to detect a lumenal loop between residues 126 and 198 leaves open the possibility that there are membraneembedded rather than membrane-spanning domains in this region of Tsc13p. However, because the analysis of the Arabidopsis ortholog of Tsc13p provides evidence for the presence of a lumenal loop in the region analogous to the yeast 126 to 198 segment (discussed below), we favor a topological model with six membrane-spanning domains (Fig. 5C). This topology is consistent with the predictions of the Localizome and HMMTOP prediction programs, although these programs differ in their prediction of the precise location of TMD2. Localizome places it between residues 125 and 143, and HMMTOP places it between residues 145 and 162. We have indicated the uncertainty about the location of TMD2 using the dashed line in Fig. 5C. However, it seems likely that TMD2 is located between 125 and 143 because the corresponding regions of the plant and mammalian orthologs are also predicted to contain a membrane-spanning domain. Furthermore, this topology would place the highly conserved domain that contains the functionally important Lys-140 and Arg-141 residues identified in this study near the cytosolic face of TMD2 where it could act in conjunction with the conserved cytosolic domain immediately preceding TMD1. This later domain contains the Gln-81 residue that when mutated to lysine reduces enoyl-reductase activity (7).
The plant and mammalian Tsc13p orthologs have hydrophilicity profiles that are similar to that of the yeast protein. As mentioned above, HMMTOP predicts six TMDs in Arabidopsis Tsc13p located in similar relative positions and with the same orientation as those in yeast except that the fifth predicted TMD of AtTSC13 is more N-terminal, suggesting that the third cytoplasmic loop is shorter and the third lumenal loop longer in comparison to yeast Tsc13p. Our topological analysis of the AtTSC13 protein expressed in yeast is fully consistent with this model. HMMTOP also predicts six TMDs and a very similar overall topology for the human Tsc13p ortholog.
Several recent studies have indicated that membrane spanning segments of proteins are often flanked by tryptophan residues because of their propensity to localize to the interface between the polar and hydrophobic layers of the phospholipid bilayer (35). It is, therefore, of interest to note that several of the predicted transmembrane domains of Tsc13p are flanked by tryptophan residues (purple circles, Fig. 5C). Although in many cases the Arabidopsis and mammalian Tsc13p orthologs lack a similarly positioned tryptophan residue (Fig. 1), there are often either tyrosines or phenylalanines (which also often flank membrane-spanning domains) located in equivalent positions near the ends of the predicted membrane-spanning domains.
Based on our topology model of Tsc13p, the three cysteine residues in the protein all lay near the cytosolic side of the membrane (yellow circles, Fig. 5C). We tested whether any of the cysteines were critical for function by substituting them with serine. Each of the single mutants and the triple cysteine-less mutant was functional, suggesting that these residues do not participate directly in catalysis. However, it should be emphasized that the assay for function was the ability to complement the tsc13⌬ mutant, and thus, we cannot conclude that these residues are not required for optimal activity of Tsc13p.
Little is known about the mechanism of Tsc13p or about its active site, and it bears no similarity with the well characterized enoyl-ACP reductase of the soluble fatty acid synthase system. Furthermore, although the C terminus of Tsc13p shares significant homology with steroid-5␣ reductase and both enzymes catalyze the reduction of a double bond that is ␣,␤ to a carbonyl group, Tsc13p lacks the predicted NADPH binding site that is present in steroid-5␣ reductase. In this study several conserved charged residues present in highly conserved regions (blue shading in Fig. 5C) as well as some conserved charged residues that are predicted to lie in the TMDs were substituted with alanine. We identified two residues, Lys-140 and Arg-141, that are critical for function but not for stability or topogenesis, whereas the Y138A mutation rendered Tsc13p unstable. Our topology studies suggest that these residues lie in or at the cytosolic face of TMD2. It is also worth pointing out that several of the insertions into the 126 -198 region of Tsc13p abolished function. Although it is tempting to speculate that these two highly conserved domains contribute to the formation of an active site at the cytosolic face of the ER membrane or possibly extending into the membrane, the precise functions of these highly conserved domains of the trans-2,3-enoyl-CoA reductases remain to be determined. In conclusion, our topology study has established the foundation for further structure-function studies on Tsc13p.