Comparative topology studies in Saccharomyces cerevisiae and in Escherichia coli. The N-terminal half of the yeast ABC protein Ste6.

Gene fusions have provided a strategy for determining the topology of polytopic membrane proteins in Escherichia coli. To evaluate whether this highly effective approach is applicable to heterologously expressed eukaryotic integral membrane proteins, we have carried out a comparative topological study of the eukaryotic membrane protein Ste6 both in bacteria and in yeast. Ste6, is an ATP binding cassette (ABC) protein, essential for export of the a-factor mating pheromone in Saccharomyces cerevisiae. The topogenic reporters, invertase in S. cerevisiae and alkaline phosphatase in E. coli, were fused to Ste6 at identical sites and the fusions were expressed in yeast and bacteria, respectively. The results obtained in both systems are similar, although more definitive in E. coli, and support the predicted six-transmembrane spans organization of the N-terminal half of Ste6. Thus, the topological determinants for membrane insertion of polytopic proteins in prokaryotic and in eukaryotic systems appear to be highly similar. In this study we also demonstrate that Ste6 does not contain a cleaved signal sequence.

We have recently used a gene fusion approach in Escherichia coli to examine the membrane topology of the eukaryotic multidrug resistance protein, mouse Mdr1 (1,2). Similar studies have been carried out with other eukaryotic integral membrane proteins expressed as gene fusions in E. coli, for instance the human ␤ 2 -adrenergic receptor (3) and the cyclic nucleotidegated ion channels (4). The proposal that it may be valid to carry out structural analysis of eukaryotic membrane proteins in E. coli stems from the assumption that the endoplasmic reticulum (ER) 1 membrane of eukaryotes and the cytoplasmic membrane of prokaryotes exhibit similar properties in regard to protein translocation, and that polytopic membrane proteins are assembled into their final membrane topology in both of these membranes. In order to explore this hypothesis, we have initiated a direct comparative examination of the topology of the N-terminal half of eukaryotic membrane protein Ste6, both in its native system, Saccharomyces cerevisiae, and in the heterologous prokaryotic system, E. coli.
Ste6 is a polytopic integral membrane protein which plays an essential role in secretion of the a-factor mating pheromone in S. cerevisiae (5)(6)(7). Ste6 belongs to a superfamily of transporters designated the ATP binding cassette (ABC) superfamily (8) or traffic ATPases (9). Members of the ABC superfamily include the multidrug resistance protein (Mdr) (10,11), the cystic fibrosis transmembrane conductance regulator (CFTR) (12), the TAP1/TAP2 peptide antigen transporter (13), the bacterial hemolysin exporter (HlyB) (14), and a variety of bacterial periplasmic permeases (15). Proteins in this family are modular in design and contain two homologous halves, each half containing a nucleotide binding domain located on the cytosolic face of the membrane and a membrane spanning domain (MSD) predicted to contain multiple transmembrane segments (TMs).
A general secondary structure model for Ste6 can be proposed based on hydropathy analysis. Such a hydropathy profile, generated according to the algorithim of Kyte and Doolittle (16) predicts that 6 TMs are present in the MSD of each half of Ste6, analogous to the secondary structure model proposed for the closely related protein, Mdr. While this model has not been experimentally addressed for Ste6, the topology of Mdr has been examined in an in vitro system (17)(18)(19), an in vivo heterologous expression system (1,2), and an in vivo eukaryotic system (20). The experimental data accumulated so far for Mdr indicates that its topology may differ from that predicted by its hydropathy profile, but notably, answers that emerge from the heterologous bacterial system provide a more clear cut picture of the topology than results obtained in vitro, suggesting the systems used for these analyses must be evaluated further.
To evaluate the use of E. coli for topology studies of eukaryotic polytopic membrane proteins and to specifically study the topology of Ste6, we have carried out a detailed comparative study using gene fusions. In E. coli we use the phoA-encoded enzyme alkaline phosphatase (AP) which is enzymatically active only when it is extracellular and thus can act as a topology sensor for the protein sequence to which it is attached (21,22). In S. cerevisiae we have used the SUC2-encoded enzyme, invertase (Inv). The extracellular form (but not the cytoplasmic form) of Inv is heavily glycosylated and only extracellular Inv is able to support growth of yeast on sucrose. The results of our topology study confirm the predicted six TM topology in the N-terminal half of Ste6 and importantly, demonstrate that Ste6 exhibits a similar pattern of topology when expressed in either S. cerevisiae or E. coli. In addition, we show in this study that contrary to a previous suggestion (23), Ste6 does not contain a cleavable signal peptide.  (27) in which a HindIII site is present at position 10 nucleotides upstream of the STE6 ATG start codon was used as the source of STE6; the Klenow-treated (in only the 5Ј end after partial digestion) 4.87-kilobase STE6-containing HindIII fragment from pSM579 (28) was ligated to pT7-5/lacY (29) that had been digested with BamHI, treated with Klenow, and then digested with HindIII. Subsequently, this construct was further modified by PCR-directed mutagenesis to create a NheI site just 5Ј to the STE6 stop codon and a BamHI site just 5Ј of the start codon of STE6. For STE6-SUC2 fusions, the full-length STE6 gene was subcloned from plasmid pTE/STE6 into pTESUC2, replacing the Inv signal sequence coding gene. The resulting plasmid, pTE/STE6-SUC2, contains a full-length STE6-SUC2 gene fusion, flanked by SalI and SacI with a unique NheI between STE6 and SUC2. The final DNA construct was sequenced through the fusion joint to confirm that the genes are joined in-frame. The hybrid encoded by this plasmid contains Ste6 in its entirety through its terminal codon Ser 1290 . pTE/STE6 -SUC2 was used to create a series of STE6-SUC2 fusions as follows: each fusion was made by PCR amplification of a specific 5Ј portion of the STE6 gene. All the PCR products were designed to contain a NheI site at their 3Ј end and another unique site at their 5Ј end. After digestion with these two restriction enzymes, PCR products were ligated into the vector pTE/STE6-SUC2 that had been digested with the same two enzymes. The plasmid products were analyzed by restriction enzyme digestion and sequencing of the cloned PCR amplified region, including the NheI junction.
Plasmid pRS/PGK containing amp r , TRP1, and the phosphoglycerate kinase (PGK) promoter, is described elsewhere (51). This shuttle vector (pRS/PGK) was used for expression of the gene fusions in yeast cells. After their construction, the STE6-SUC2 fusions were transferred from pTE/STE6-SUC2 to pRS/PGK as SalI-SacI fragments.
Plasmids pRS/PGK and pRS/SUC2 were used as negative and positive control plasmids, respectively. pRS/SUC2 was constructed by ligating the SalI-SacI SUC2 fragment from pTE/SUC2 into pRS/PGK digested with the same restriction enzymes. This plasmid encodes a fully functional precursor form of Inv, under the control of the PGK promoter.
For gene fusion studies in E. coli, plasmid pTE/STE6-phoA was constructed in which STE6 (from pTE/STE6) is placed under control of the lac promoter (Fig. 1A), replacing lacY in plasmid pT7-5/lacY-phoA (30) using the enzymes BamHI and NheI. This plasmid encodes the full-length Ste6-mature AP hybrid with a unique NheI site between STE6 and phoA, which lies in the same frame as it does between STE6 and SUC2 in pTE/STE6-SUC2. Using the unique NheI site and one of several other sites in the 5Ј sequence of STE6, the full-length STE6 gene in pTE/STE6-phoA was replaced by portions of STE6, obtained by digesting various STE6-SUC2 fusions with the same restriction enzymes.
To examine the possibility of signal sequence cleavage in Ste6 we used epitope-tagged STE6 constructs. Plasmid pSM498 (CEN LEU2 STE6-C-HA) contains a C-terminal triply iterated epitope from influenza hemagglutinin (HA) inserted in-frame at the C terminus of STE6. It was constructed as follows: 1) pSM192 (CEN URA3 STE6) (24) was subjected to oligonucleotide mutagenesis which resulted in the simultaneous removal of a native internal BamHI site (at residue 1002, codon 334) and the introduction of a novel BamHI site adjacent to the STE6 termination codon to yield plasmid pSM324. 2) The SalI-HindIII STE6 fragment from pSM324 was subcloned into pSM321, a modified version of the vector pRS315 (CEN LEU2) in which the BamHI, PstI, and SmaI polylinker sites are absent, generating pSM323. 3) Subsequently a BglII fragment containing the triply reiterated HA tag from pSM492 (27) was cloned into the C-terminal BamHI site of pSM323, yielding pSM498. Plasmid pSM1018 (CEN LEU2 STE6-N-HA) contains an Nterminal triply reiterated HA tag. It was constructed by a similar procedure to pSM498 except that in the first step, the novel BamHI site was inserted between codons 7 and 8 of STE6. The plasmid pSM322 (STE6 CEN LEU2) bears an untagged version of STE6 and was used as a control for specific immunoprecipitation.
Growth on Minimal Plates Containing Sucrose-Yeast cells expressing the various hybrids were grown in supplemented SD medium to saturation at 30°C. The cultures were washed once with phosphate/ citrate buffer, pH 5, and resuspended in the same buffer. Aliquots (20 l) of cells diluted 1:50 in the same buffer were spotted on sucrose plates, which contain 0.5% sucrose, Bacto-yeast nitrogen base without amino acids, and all amino acids except for those needed for selection. Plates were photographed after 24 h at 30°C.
Antibodies-Invertase from S. cerevisiae (U. S. Biochemical Corp.) was used to raise antibodies. Rabbit anti-Inv antibodies were prepared by injecting rabbits with Inv that was purified by SDS-PAGE. Carbohydrate-specific antibodies were removed by adsorbing the serum with S. cerevisiae DGY505 (suc2) cells. Serum was incubated with yeast cells at 4°C for 4 h and the incubation was repeated with new cells 3 times. The mouse monoclonal antibodies against the HA epitope were from Babco. Antibodies to AP were from 5 Prime 3 3 Prime Inc. Horseradish peroxidase-conjugated goat anti-rabbit antibodies were obtained from Jackson ImmunoResearch.
Western Blot and Endoglycosidase H Treatment-For Western blot analysis, saturated cultures were diluted 1:20 and grown in SD medium. Cells were harvested at stationary growth phase, washed twice with phosphate/citrate buffer, pH 5, containing 1 mM PMSF and 0.1% NaN 3 , and resuspended in the same buffer at OD 600 ϭ 25. Typically, 300 l of cells were mixed with glass beads and vortexed for 2 min followed by a 1-min centrifugation. The supernatant, which contains the cell extract was collected and the amount of protein in the samples was measured by the modified Lowry method (31) with SDS. Extracts containing equivalent amounts of proteins were boiled for 10 min in Endo H denaturing buffer. Each sample was split and half was subjected to treatment with 1000 units of Endo H f (BioLab) for 2 h at 37°C. The samples were then loaded on a 5-10% SDS-PAGE gel as specified. Transfer of the proteins to nitrocellulose membrane was carried out overnight in Tris glycine buffer containing 20% methanol. The membrane was incubated with preadsorbed anti-Inv antibodies. Horseradish peroxidase-conjugated goat anti-rabbit antibodies were used to detect the Inv-containing species by ECL (Amersham).
Metabolic Labeling in S. cerevisiae and Immunoprecipitation-For analysis of Ste6-Inv hybrids, overnight cultures propagated in SD medium lacking methionine were diluted one-tenth with fresh medium and grown for 4 h at 30°C. Cells (typically 5-ml cultures) were harvested, resuspended in 0.5 ml of fresh medium, and split into 2 aliquots. Tunicamycin (10 g/ml) was added to one aliquot, both samples were incubated at 30°C for 20 min to establish the tunicamycin block, and subsequently cells were labeled with 15 Ci of [ 35 S]methionine (1000 Ci/mmol) for 15 min. The labeled cells were mixed with lysis buffer (final concentrations: 0.24 M NaOH, 1% ␤-mercaptoethanol, 1 mM PMSF, and 0.2 M Trasylol) and proteins were precipitated by addition of trichloroacetic acid (final concentration 5%) on ice for 2 h. Precipitates were washed twice with acetone and dried at room temperature. The washed precipitates were resuspended in SDS buffer (10 mM Tris-HCl, pH 8, 1% SDS, 5 mM EDTA, 1 mM PMSF, and 0.2 M Trasylol) and placed on ice for 5 min. Ice-cold KI buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 2% Triton X-100, 5 mM EDTA, 1 mM PMSF, and 0.2 M Trasylol) (500 l) was added and after a 5-min incubation on ice, the insoluble material was pelleted by centrifugation at 4°C. 400 l were withdrawn and mixed with 300 l of ice-cold KI buffer. Samples were incubated for 30 min at 4°C with 15 l of pre-equilibrated protein A suspension to adsorb nonspecific labeled material. After centrifugation (2 min, 13,000 rpm), the pellet was discarded and the supernatant was incubated overnight with 15 l of anti-Inv antibodies at 4°C. Immunoprecipitation was accomplished by incubating the samples with 15 l of pre-equilibrated protein A suspension for 1.5 h at 4°C. Immunoprecipitates were washed twice with high salt buffer (50 mM Tris-HCl, pH 8, 1 M NaCl, 1% Triton X-100, 1 mM PMSF, and 0.2 M Trasylol) and once with ice-cold 10 mM Tris-HCl, pH 8, buffer containing 1 mM PMSF and 0.2 M Trasylol. Pellets were solubilized in sample buffer with 2% ␤-mercaptoethanol, incubated for 30 min at 50°C, and separated by SDS-PAGE, followed by autoradiography. Prestained molecular weight markers were used to estimate the molecular weight of the different hybrids.
For analysis of HA-tagged Ste6 constructs, growth of cells, metabolic labeling, immunoprecipitation, and SDS-PAGE were carried out essentially as described for Ste6 (27) except that mouse monoclonal anti-HA antiserum 12CA5 (ascites fluid, Babco, Richmond, CA) was used for immunoprecipitation.
Metabolic Labeling of E. coli and Immunoprecipitation-E. coli cells harboring the fusion plasmids, or the parental vector pT7-5 as a control, were grown, labeled, and immunoreacted with anti-AP antibodies as described (1,32). Immunoprecipitated material was extracted in 50 l of sample buffer, separated on SDS-PAGE, and the dry gel was exposed to film for at least 24 h. Radioactive bands were quantitated by a densitometer. Prestained molecular weight markers were used to estimate the molecular weight of the different hybrid proteins.
Enzymatic Assays-The assay we used to determine the subcellular localization of invertase is a modification of the method described by Johnson et al. (33) that distinguishes between extracellular and intracellular invertase activity. S. cerevisiae DGY505 cells harboring fusion plasmids were grown to mid-log phase in SC medium lacking tryptophan. Cells (5 OD 600 units) were harvested, washed once in 10 mM sodium azide, once in 50 mM sodium acetate, and resuspended in 0.5 ml of the same buffer. One volume of cells was untreated, the remaining sample was frozen on dry ice and thawed at room temperature. The untreated cells were assayed according to Goldstein and Lampen (34) in 50 mM sodium acetate and the freeze-thawed cells were assayed either in a 50 mM sodium acetate buffer, or 50 mM sodium acetate with 1% Triton X-100. The value determined by assaying untreated whole cells corresponds to the periplasmic activity (a), that of freeze-thawed cells in the absence of detergent corresponds to the combined periplasmic plus cytosolic activity (b), and that of freeze-thawed cells in the presence of Triton corresponds to the total invertase activity (c), including periplasmic, cytosolic, and lumenal invertase. Cytoplasmic activity (b-a) was expressed as a percentage of the total (c) and is presented graphically in Fig. 6. It should be noted that the freeze-thaw step, carried out in the absence of detergent permitted minimal rupture of organelles and allowed us to distinguish cytosolic invertase from lumenal invertase that had not yet reached the periplasm. Controls produced the expected results, including wild-type SUC2 (1% cytosolic, 95% periplasmic, 4% luminal), a suc2 mutant deleted for its signal sequence suc2⌬sp (91% cytosolic, 7% periplasmic, 1% luminal), and suc2-S1, a mutant reported by Schauer et al. (35) to be secreted very slowly (26% cytosolic, 40% periplasmic, 34% lumenal).
For AP assays, overnight cultures of E. coli UT5600 cells harboring the STE6-phoA gene fusions, or the vector as negative control, were diluted 1:50 into 2 ml of LB containing 100 g/ml ampicillin and grown for 3 h at 30°C. The cells were then induced with isopropyl-1-thio-␤-Dgalactopyranoside (0.5 mM final concentration) for 2 h at 30°C. After appropriate treatment (1, 32) cleavage of p-nitrophenyl phosphate was measured at OD 420 as described (1,32). Alkaline phosphatase activity was calculated as in Brickman and Beckwith (36).

Construction of STE6-SUC2 and STE6-phoA Gene Fusions-
To determine the topology of the N-terminal half of Ste6 expressed in yeast and to compare it to the topology of the N-terminal half of Ste6 expressed in E. coli, a series of fusion constructs were created (Fig. 1A) in which varying portions of the N-terminal half of Ste6 are joined to the mature coding sequence of the yeast SUC2 gene (STE6-SUC2) or to the bacterial phoA gene (STE6-phoA). The hybrid proteins encoded by these fusions are referred to as Ste6-Inv and Ste6-AP, respectively. Of importance for this comparative study is the fact that the Ste6 portion of each Ste6-Inv hybrid is identical to that of the corresponding Ste6-AP hybrid. The specific fusion joint in each construct is indicated in Fig. 1B, in relationship to the predicted TMs and the connecting loops (designated L1-6). It should be noted that L6 corresponds to the N-terminal nucleotide binding domain of Ste6. A detailed description of the construction of the plasmids is provided under "Experimental

Procedures."
The Orientation of the N-terminal Transmembrane Segment of Ste6 -We initially wished to determine the orientation of the first hydrophobic segment of Ste6 in the membrane and to assess its possible role as a "start transfer" signal. We examined two Ste6-Inv hybrids, Ile 20 and Arg 21 , whose junctions lie in a region preceding putative TM1 of Ste6 and two hybrids, Ala 60 and Val 72 , whose junctions lie in the first putative extracellular loop (L1) of Ste6 by Western blot analysis ( Fig. 2A). Samples were separated by SDS-PAGE with or without treatment with Endo H, which removes glycosyl groups from translocated polypeptide products. Wild-type Inv exhibits a dramatic shift in mobility after Endo H treatment, indicative of glycosylation ( Fig. 2A, lanes 7 and 8), while the mobility of Ile 20 and Arg 21 is unaffected by Endo H treatment ( Fig. 2A, lanes 3, 4,  and 9, 10). Thus, as expected, the Ile 20 and Arg 21 hybrids are not glycosylated, presumably because they are unable to cross the ER membrane. Indeed, we find that these hybrids are not membrane-associated (not shown). In contrast, hybrids Ala 60 and Val 72 , which lie in the first predicted extracellular loop (L1) of Ste6 exhibit a high M r Endo H-sensitive species (Fig. 2A,  lanes 5, 6, 1, and 2), indicating that the Inv moiety portion of these hybrids is glycosylated, and thus must be translocated into the lumen of the ER. Surprisingly, Endo H-treated Ala 60 and Val 72 migrate similar to Endo H-treated invertase (see below). In addition, we find that hybrids Ala 60 and Val 72 and all subsequent hybrids examined in this study are membraneassociated (data not shown). Overall, these results suggest that the N-terminal region of Ste6 is cytosolic, that L1 of Ste6 is lumenal (i.e. on the extracellular face of the membrane). Thus, TM1 of Ste6 possesses an N in -C out orientation, and acts as a start transfer sequence.
An important advantage of using Inv as a reporter is its enzymatic and biological activities. The native homodimeric periplasmic enzyme converts extracellular sucrose into fructose and glucose, thus enabling cells to utilize sucrose as a carbon source for growth. We expect that outwardly exposed Inv, detected as a glycosylated species in certain Ste6-Inv hybrids, will function in the yeast periplasm in a manner analo- gous to the native secreted form of Inv and permit cells to grow on sucrose. In agreement with the glycosylation results above, and as shown in Fig. 5, cells expressing the L1 hybrids Ala 60 and Val 72 are able to grow on 0.5% sucrose, while the hybrids preceding TM1, Ile 20 and Arg 21 , do not support growth on sucrose.
Ste6 Does Not Contain a Cleavable Signal Sequence-In the experiment described above we noted an anomolous migration pattern for the Endo H-treated deglycosylated forms of hybrids Ala 60 and Val 72 . These hybrids should contain an additional 60 and 72 amino acid residues, respectively, as compared to the deglycosylated form of mature Inv, yet the mobility of all three species appears to be identical ( Fig. 2A, compare lanes 1, 5, and  7). One possible explanation for this result is that Ste6 contains a cleaved N-terminal signal sequence and that after cleavage, the final molecular weight of the hybrids is actually similar to mature wild-type Inv. A second explanation is that anomalous migration arises due to a technical artifact, namely that certain glycosyl groups in mature Inv (which has 9 -10 N-linked oligosaccharide chains per subunit) are not accessible to Endo H (37). According to this explanation, Endo H-treated Inv would still contain some glycosyl moieties and thus would be expected to migrate more slowly than true, non-glycosylated mature Inv. To distinguish these possibilities, cells expressing wild-type Inv or the various hybrids were treated in vivo with tunicamycin, a treatment which completely blocks glycosylation. Following tunicamycin treatment, cells were labeled with [ 35 S]methionine, proteins were immunoprecipitated, and separated by SDS-PAGE. As shown in Fig. 2B, after tunicamycin treatment wild-type Inv now exhibits its expected mobility, which is notably faster than that of hybrid Val 72 (Fig. 2B, lanes 3 and 1,  respectively), and similar to hybrid Arg 21 (Fig. 2B, lanes 3 and  5) thus obviating the need to propose a cleaved Ste6 signal sequence. Based on the equivalent migration pattern of mature Inv and the Val 78 hybrid, after Endo H treatment, Kolling and Hollenberg (23) concluded that Ste6 contains a cleavable leader peptide; they also noted the presence of a potential signal sequence cleavage site between Gly 62 and Ser 63 , according to an algorithm developed by von Heijne (38). However, our results clearly do not support this interpretation. Instead, our studies indicate that the results of Kolling and Hollenberg (23) can be explained as an artifact of incomplete deglycosylation of Inv by Endo H.
To verify that Ste6 does not contain a cleavable signal sequence, we compared the metabolic stability of full-length Ste6 tagged with an HA epitope, either at its N terminus (between amino acids 7 and 8) or at its C terminus. Cells bearing these HA-tagged STE6 constructs were pulse-labeled with [ 35 S]Met and Cys, radioactivity was chased for varying lengths of time, labeled proteins were immunoprecipitated with anti-HA antibodies, and separated by SDS-PAGE. Signal sequence cleavage is a rapid process that occurs either during or immediately after the completion of protein synthesis. Thus, if a cleavable N-terminal signal peptide were present in Ste6, we would expect significantly more rapid disappearance of the radioactive label from the N-terminally tagged molecule due to its removal by signal peptidase than from the C-terminally tagged molecule. However, as shown in Fig. 3, there is no difference in the stability of the tagged molecules. Instead, both tagged versions of Ste6 exhibit the same half-life (ϳ20 min); this moderately unstable pattern is typical of Ste6, and reflects its degradation in the vacuole (27). Overall, in contrast to a previous report (23), our analysis shows that Ste6 does not appear to contain a cleaved signal sequence.

Membrane Topology of the N-terminal Half of Ste6 Expressed in S. cerevisiae, as Assessed by Glycosylation of Inv-
To study the membrane topology of the N-terminal half of Ste6 in yeast, we examined a series of Ste6-Inv hybrids in which Inv is joined to amino acids in the hydrophilic loops (L1-L6) of Ste6 predicted from hydropathy analysis (Fig. 1B). To determine the orientation of Inv with respect to the ER membrane, we examined glycosylation, as detected by gel mobility of the metabolically labeled hybrid proteins. Labeling was carried out both in the presence or absence of tunicamycin, and the results of this set of experiments are shown in Fig. 4A (the results shown in Fig. 4B are described below). We first examined hybrids in L1 (Ala 60 and Val 72 ; Val 72 is not shown), L3 (Thr 168 and Ser 169 ) (Ser 169 is not shown), and L5 (Ile 290 ) which are predicted from hydropathy analysis to lie on the extracellular face of the membrane. Representative results, shown in Fig. 4A, indicate that all of these hybrids undergo glycosylation that is abolished by tunicamycin treatment, indicating that loops 1, 3, and 5 are on the lumenal side of the membrane, matching the hydropathy prediction. FIG. 2. A, Western blots of Ste6-Inv hybrid proteins expressed in S. cerevisiae. Extracts prepared from S. cerevisiae cells containing plasmids that direct the expression of native Inv (pRS/SUC2) or each of the following hybrids Ile 20 , Arg 21 , Ala 60 , or Val 72 were split. One-half of the extract was treated with Endo H, the other half was not, as indicated. 50 g of protein from Endo H treated or untreated samples were subjected to SDS-PAGE (7.5%), electroblotted, and the nitrocellulose was incubated with anti-Inv serum. After incubation with goat antirabbit antibodies conjugated to horseradish peroxidase, followed by a short incubation with fluorescent substrate, the nitrocellulose was exposed to film for approximately 5 min. B, immunoprecipitation of metabolically labeled hybrids from S. cerevisiae treated with tunicamycin. Cells harboring the vector alone (pRS/PGK) or expressing native Inv or various hybrids were incubated with tunicamycin and labeled with [ 35 S]methionine. Extracts were immunoprecipitated with anti-Inv antibodies and the immunoprecipitated material was subjected to SDS-PAGE (7.5%) followed by autoradiography.  Ste6 (pSM322, lanes 1-5), Ste6 tagged at its C terminus (pSM498, lanes 6 -10), or Ste6 tagged at its N terminus (pSM1018, lanes [11][12][13][14][15] were metabolically labeled for 5 min with 75 Ci of [ 35 S]Express (Amersham) and the label was chased with cold methionine and cysteine for the time indicated above each lane. Proteins were subjected to immunoprecipitation with anti-HA antibodies (12CA5), separated on 8% SDS-PAGE, and exposed to film. We next analyzed hybrids predicted to be cytosolic. Hybrids in L4 (Ala 213 and Arg 236 ) and L6 (Ile 314 and Leu 350 ) exhibit an identical pattern of migration on SDS-PAGE with or without tunicamycin treatment, indicating that these hybrid proteins do not undergo glycosylation (Fig. 4A). Likewise, as shown above using Endo H ( Fig. 2A), the migration of hybrid Ile 20 and Arg 21 in the N-terminal region of Ste6 is also indifferent to prior treatment of cells with tunicamycin (data not shown). Thus, as predicted from hydropathy analysis, the N terminus of Ste6, L4, and L6 appear to lie on the cytosolic face of the membrane. A unique pattern of migration was observed for the L2 hybrid, Ser 152 , in which a glycosylated species as well as a distinct nonglycosylated species was detected. This "mixed phenotype" result is discussed in detail below.
Membrane Topology of the N-terminal Half of Ste6 Expressed in S. cerevisiae, as Assessed by Growth on Sucrose and Localization of Inv Activity-Hybrids that are glycosylated should promote growth on sucrose as the sole carbon source if they are targeted to the plasma membrane and their Inv moiety is exposed to the periplasm. This was tested by spotting cells harboring various fusions on minimal agar plates supplemented with 0.5% freshly made sucrose. Growth was monitored after 24 h (Fig. 5). Cells expressing native Inv, or the glycosylated hybrid proteins with junctions in L1 (Ala 60 , Val 72 ), L3 (Thr 168 ), and L5 (Ile 290 ) exhibit robust growth on sucrose as expected. Not surprisingly, cells harboring the vector alone or expressing the non-glycosylated hybrid proteins with junctions in cytosolic domains, i.e. the N terminus, Ile 20 , Arg 21 , L4, Ala 213 and Arg 236 , and L6, Ile 314 or Leu 350 , do not grow under these conditions. On the other hand, expression of the L2 hybrid Ser 152 , which exhibits the mixed glycosylation phenotype described above, enables yeast to grow somewhat on sucrose, but to a much lesser extent than is the case for the L1 (Ala 60 ), L3 (Thr 168 ), and L5 (Ile 290 ) hybrids.
The localization of the Inv moiety of the hybrids was further analyzed by assaying enzyme activity, under conditions that allowed us to distinguish cytosolic invertase from translocated (periplasmic plus intracellular lumenal) invertase. The percentage of cytosolic invertase activity for each Ste6-Inv hybrid protein is shown in Fig. 6. For the control suc2⌬sp, which contains no Ste6, about 90% of the Inv activity is cytosolic (Fig.  6, fusion location 0), while for wild-type SUC2, Ͻ5% of the invertase activity is cytosolic (data not shown). A high level of cytosolic invertase activity (50 -90%, except for hybrid Ile 20 with 42%) is detected for hybrids in the N terminus of Ste6 (Arg 21 ), in L4 (Ala 213 , Arg 236 ) and L6 (Leu 350 , Cys 642 ), reaffirming the conclusions from the glycosylation and growth assays that the N terminus, L4 and L6 are cytosolic. Interestingly, the hybrids in L2 (Ala 114 , Ser 152 ) which show a mixed phenotype in other assays also exhibit unexpectedly low cytoplasmic invertase activity, and are discussed in more detail below. As anticipated, for the presumed extracellular hybrids in loops L1 (Ala 60 , Val 72 ), L3 (Thr 168 , Ser 169 ), and L5 (Ile 290 ), the cytosolic activity is low (20 -30%), consistent with the prediction that L1, L3, and L5 are lumenal. It should be noted, however, that the cytoplasmic activity in these fusions is not as low as for WT Inv (Ͻ5%). This may be due to incomplete translocation of Inv in these hybrids or, alternatively, to some inherent leakiness in the assay; for instance, a low level of organelle rupture during processing of our samples could contribute to apparent cytoplasmic activity. Nevertheless, the relatively low activity of the L1, L3, and L5 hybrids confirms the overall view that these loops of Ste6 are lumenal.
Hybrids in L2 of Ste6 Exhibit Anomalous Behavior-In general, the enzymatic activities of nearly all of the hybrids are in agreement with the glycosylation patterns and with the sucrose growth experiments, and confirm the predictions based on hydropathy analysis (Fig. 1B). An exceptional entity is hybrid Ser 152 (in predicted L2) which displays a less clearcut result. By hydropathy, Ser 152 is predicted to lie in a cytosolic loop, L2. However, by the Inv activity assay, glycosylation analysis, and sucrose growth experiments Ser 152 exhibits an ambiguous phenotype. One possible explanation for the mixed topology of hybrid Ser 152 is that TM2, which precedes it, is not stably anchored in the membrane in the context of a C-terminally truncated Ste6, causing downstream sequences, including Inv, to exist on both sides of the membrane. To determine whether sequences within L2 might be responsible for desta- FIG. 4. Immunoprecipitation of Ste6-Inv hybrids from S. cerevisiae either treated with tunicamycin or untreated. Cells harboring the vector alone (pRS/PGK) or expressing each of the indicated hybrids were split and one aliquot of each sample was incubated with tunicamycin prior to labeling with [ 35 S]methionine. Extracts were immunoprecipitated with anti-Inv antibodies and the immunoprecipitated material was subjected to SDS-PAGE (7.5%) followed by autoradiography.

FIG. 5. Growth of cells harboring
STE6-SUC2 fusions on sucrose. S. cerevisiae cells harboring the indicated gene constructs were grown in minimal broth containing glucose. Mid-log cultures were harvested, washed with buffer as described to remove traces of carbon source, and suspensions were spotted on agar plates containing 0.5% sucrose as the carbon source. Photographs were taken after incubation for 24 h at 30°C. bilizing TM2, we constructed another hybrid, Ala 114 , in which the fusion junction is located on the N-terminal side of L2. An additional hybrid was also constructed, which serves as a control, and contains Inv joined to Val 200 located at the N-terminal side of the adjacent cytoplasmic loop (L4). Cells expressing these new hybrids were either treated or not treated with tunicamycin, labeled, and analyzed after immunoprecipitation and SDS-PAGE. The control, hybrid Val 200 (Fig. 4B, lanes 3  and 4), exhibits the same mobility with or without tunicamycin treatment, confirming its cytosolic orientation as predicted. However, just as with hybrid Ser 152 , only a fraction of the molecules of hybrid Ala 114 undergo glycosylation (Fig. 4B, lanes  1 and 2), indicating that a significant portion of the invertase molecules (the non-glycosylated species) are retained in the cytoplasm. These results further suggest that the mixed properties of hybrids Ser 152 and Ala 114 are most likely caused by the instablility of TM2 and not by signals located in L2. Therefore, we suggest that L2 is cytoplasmic, and that this region of the protein is conformationally unstable. Notably, a similar observation suggesting two alternative membrane topologies has been described for Mdr (17).
Membrane Topology of the N-terminal Half of Ste6 Expressed in E. coli-Recently, the membrane topology of the mouse Mdr1 protein has been studied using Mdr-AP hybrids in E. coli (1,2). Although the information obtained from that study is in general agreement with conclusions reached regarding the topology of Mdr1 in eukaryotic expression systems (17,18), a direct comparison between the outcome of topology studies of a single eukaryotic membrane protein in the prokaryotic and eukaryotic systems, using similar molecular tools, has not been undertaken. For this reason, we felt it would be extremely valuable to directly compare the topogenic behavior of the eukaryotic polytopic membrane protein Ste6 in both the eukaryotic and the prokaryotic systems in vivo.
To study the topology of Ste6 in E. coli, we generated a series of STE6-phoA fusions corresponding to a subset of the STE6-SUC2 fusions analyzed above (Fig. 1A). To enable direct comparative analysis, the AP is joined to Ste6 in these fusions at exactly the same Ste6 junctions as in the Ste6-Inv fusions. In E. coli, AP becomes active only when translocated to the periplasmic space; non-translocated cytosolic AP is inactive. Therefore, the location of the reporter with respect to the cytoplasmic membrane can be determined by assaying the specific activity of AP in the hybrids. To calculate the specific activity of various Ste6-AP hybrids, we first examined their level of expression. One problem that we encountered with respect to the expression of Ste6 in E. coli is proteolysis, 2 which is significant even in E. coli UT5600, a strain devoid of the outer membrane protease OmpT. This problem is apparent in the immunoprecipitation experiments with cells expressing various Ste6-AP hybrids; in addition to the full-length hybrid protein, we also observed rapidly migrating species corresponding to breakdown products (Fig. 7). Nevertheless, all the rapidly migrating bands shown in the autoradiogram must contain AP epitopes, since we used anti-AP antibodies for immunoprecipitation. Previous studies on the properties of Mdr-AP proteolytic fragments suggest that the AP is still connected to its immediate N-terminal topogenic determinant (22) and thus may be included for the calculation of the specific activity (1). The AP activity of the Ste6-AP hybrids was determined as described under "Experimental Procedures:" the band intensity (fulllength and proteolytic products) divided by the number of methionines was used as an estimate of the level of expression and this in turn was used to normalize the AP activity (Table I). The final estimation is based on the average of three independent immunoprecipitation and activity experiments. In Fig. 7, a representative immunoprecipitation experiment is shown. The results, summarized in Table I (42), appear to exhibit similar membrane topologies, with their Nterminal halves containing 6 transmembrane segments. In addition, extensive hydropathy comparison predicts that the MSDs of Ste6 share features in common with MSDs of Mdr and CFTR, namely, 6 TMs and intervening loops of conserved length (43). Indeed, the gene fusion analysis presented in this paper in both the eukaryotic and the prokaryotic expression systems strongly supports the proposed topology in the Nterminal half of Ste6.
Use of Inv as a Reporter Protein-The methodology of gene fusions provides an attractive approach to deduce topology of a membrane protein in vivo. In this report we compare the use of AP and Inv fusions to study Ste6 topology in vivo, in both a prokaryotic and eukaryotic organism, E. coli and S. cerevisiae, respectively. Previous work had established the use of another reporter, the histidinol dehydrogenase (HD) protein encoded by HIS4C, as a topologically sensitive monitor for gene-fusion analysis of the topology of ER membrane proteins in S. cerevisiae (44) and of a plasma membrane protein, arginine permease (45). Here we have chosen to examine the use of Inv as a topology detector for the plasma membrane protein Ste6. Invertase has been utilized in a previous topology study for examining the ER membrane resident protein Sec62p (46). The advantages of using Inv to study topological elements in plasma membrane proteins are 3-fold: the protein becomes heavily glycosylated upon translocation into the lumen of the ER which can be detected as a mobility shift by SDS-PAGE, the secreted form of Inv enables cells to utilize sucrose as a sole carbon source for growth of cells, and finally, its enzymatic activity can be assayed in vivo by growth of cells on indicator agar plates or by an enzyme activity assay. Importantly, while the non-secreted form of Inv is also fully active, it is unable to promote growth on sucrose. On the one hand, the use of Inv in this study has proved to be informative, lending support for the proposal that the N-terminal half of Ste6 contains 6 TMs. On the other hand, our study suggests that caution should be taken when utilizing Inv in gene-fusion technology, as exemplified by the mixed phenotypes observed with hybrids in which Inv is joined to residues in the rather large cytoplasmic loop (L2) between TM2 and TM3. It is not yet clear whether the reason for the complex phenotype of these hybrids is due to a particular structure within the Ste6 protein or to the specific properties of Inv. It is possible that for proper assembly of L2 between TM2 and TM3, C-terminal domains of Ste6 are essen-tial. Alternatively, the mixed topology we observe with the fusions could represent a mixed or alternating topology actually adopted by Ste6 in S. cerevisiae. However, we cannot exclude the possibility that the mature portion of Inv harbors signals that under certain conditions become sufficient to promote translocation across the ER membrane.
Use of the Heterologous E. coli System for Topological Analysis-When AP was used as a reporter for Ste6 in the heterologous E. coli expression system, no discrepancy was observed. The results clearly support a secondary structure model in which Ste6 contains 6 TMs in its N-terminal half. Therefore, we conclude that in this case, the E. coli-phoA-fusion system, unlike the S. cerevisiae-SUC2 fusion system, provides information that is less controversial. Interestingly, and in line with this conclusion, previous gene-fusion studies with Mdr show that results obtained in the E. coli system could clarify uncertainties observed in analogous gene-fusion studies in eukaryotic expression systems (2). What, then, is the possible explanation for the clear phenotypes obtained with gene fusions in E. coli, in contrast to the mixed phenotypes obtained in eukaryotic systems. Schematically, polytopic membrane proteins are composed of two types of transmembrane segments, signal anchor and stop transfer sequences (reviewed in Ref. 47). The topology of these proteins is determined by the orientation of the signal anchor and the stop transfer stretches. Flanking charged residues are probably major determinants of the transmembrane orientation of these sequences. More specifically, orientation preferences are imparted by positively charged amino acids, termed by von Heijne (48) as the positive-inside rule. However, predictive methods which are based on positive charge distribution in addition to hydrophobicity profile predictions seem to perform most reliably on bacterial membrane proteins (49). Electrochemical potential across the bacterial membrane may in part provide a mechanistic basis for the positive-inside rule in bacteria (50). Therefore, the substantial difference between the ER membrane (of which the membrane potential is substantially lower than that in the eukaryotic plasma membrane which has a ⌬ H ϩ of approximately Ϫ60 mV) and the inner membrane of E. coli (⌬ H ϩ of about Ϫ150 mV) may explain the clear-cut results that we obtain with Ste6-AP hybrids in E. coli. The hydrophilic stretch of amino acid residues N-terminal to the fusion joint in hybrid Ala 114 , which exhibits a mixed phenotype in yeast, contains only one net positive charge (the sum of 3 Arg residues and 2 Glu residues). Therefore, it is possible that this weak positive charge performs as a better stop transfer signal in E. coli than in S. cerevisiae.