Determination of the transmembrane topology of yeast Sec61p, an essential component of the endoplasmic reticulum translocation complex.

Sec61p is a highly conserved integral membrane protein that plays a role in the formation of a protein-conducting channel required for the translocation of polypeptides into, and across, the membrane of the endoplasmic reticulum. As a major step toward elucidating the structure of the endoplasmic reticulum translocation apparatus, we have determined the transmembrane topology of Sec61p using a combination of C-terminal reporter-domain fusions and the in situ digestion of specifically inserted factor Xa protease cleavage sites. Our data indicate the presence of 10 transmembrane domains, including several with surprisingly limited hydrophobicity. Furthermore, we provide evidence for complex intramolecular interactions in which these weakly hydrophobic domains require C-terminal sequences for their correct topogenesis. The incorporation of sequences with limited hydrophobicity into the bilayer may play a vital role in the formation of an aqueous membrane channel required for the translocation of hydrophilic polypeptide chains.

Sec61p is a highly conserved integral membrane protein that plays a role in the formation of a proteinconducting channel required for the translocation of polypeptides into, and across, the membrane of the endoplasmic reticulum. As a major step toward elucidating the structure of the endoplasmic reticulum translocation apparatus, we have determined the transmembrane topology of Sec61p using a combination of C-terminal reporter-domain fusions and the in situ digestion of specifically inserted factor Xa protease cleavage sites. Our data indicate the presence of 10 transmembrane domains, including several with surprisingly limited hydrophobicity. Furthermore, we provide evidence for complex intramolecular interactions in which these weakly hydrophobic domains require C-terminal sequences for their correct topogenesis. The incorporation of sequences with limited hydrophobicity into the bilayer may play a vital role in the formation of an aqueous membrane channel required for the translocation of hydrophilic polypeptide chains.
Protein translocation across the membrane of the endoplasmic reticulum (ER) 1 is a decisive step in the biosynthesis of many classes of proteins in eukaryotes. The process of signal sequence recognition and targeting of proteins to the ER has been substantially elucidated (1), but the mechanism by which they subsequently cross the membrane remains less well understood. Increasingly, genetic, biochemical, and biophysical approaches have been applied in probing the translocation process.
The suggestion that the passage of nascent polypeptides across the membrane occurs through a proteinaceous, aqueous channel (2) is supported by a growing body of evidence. Electrophysiological studies indicate the presence of large aqueous channels in rough ER membranes (3), and fluorophores incorporated into a translocating polypeptide report an aqueous environment when trapped within the membrane (4,5). Considerable support for the proposal has been provided by the identification of proteins located in proximity to translocating nascent polypeptides by use of cross-linking techniques (6).
Genetic studies in Saccharomyces cerevisiae have led to the identification of a number of genes involved in the translocation process (7)(8)(9)(10)(11)(12)(13). These include SEC61 that is an essential gene encoding the 53-kDa integral membrane protein, Sec61p (13). Sec61p appears to exist as part of a stable complex together with Sss1p and Sbh1p (14,15). Like Sec61p, Sss1p is essential for viability and for protein translocation in vivo (9). In contrast, Sbh1p is encoded by a nonessential gene, and its role in translocation is uncertain (16). Both the Sec61 complex and a second membrane protein complex (comprising Sec62p, Sec63p, Sec71p, and Sec72p) are required for the efficient posttranslational translocation of prepro-␣-factor into reconstituted proteoliposomes (15). Cross-linking studies indicate that Sec61p is in close proximity to prepro-␣-factor at different stages of its translocation through the bilayer (17,18), suggesting that Sec61p may play a direct role in the formation of a protein-conducting channel in the ER membrane. A mammalian homologue of Sec61p, Sec61␣ (19), also contacts nascent polypeptides during their membrane transfer (19 -21) and exists in a homologous Sec61 complex together with Sec61␤ (Sbh1p) and Sec61␥ (Sss1p) (22).
The determination of transmembrane topology is essential for any understanding of the structure-function relationship of Sec61p in the translocation process. Hydropathic analysis of the Sec61p sequence reveals a number of regions comprising predominantly hydrophobic residues, which may correspond to transmembrane domains. The establishment of transmembrane topology at the ER membrane is thought to depend upon topogenic sequences directing sequential translocation and membrane integration events (23,24). Support for this comes largely from the generation of predicted topologies by expression of protein chimeras encoding signal, signal-anchor, and stop-transfer sequences (25)(26)(27).
In this report we have analyzed the orientation of the yeast Sec61p in the ER membrane. We used a C-terminal reporter technique whereby N-terminal portions of Sec61p were fused to a topological reporter protein. This provided evidence for seven transmembrane domains. As a result of anomalies arising from this approach, the topology was probed using insertion fusions and expression of Sec61p as complementary polypeptide fragments. This provided evidence for a further three transmembrane domains, resulting in a model in which Sec61p spans the ER bilayer 10 times. Several of the transmembrane domains are relatively weakly hydrophobic and may be crucial in the formation of an aqueous protein-conducting channel. This combined approach to topology determination also strongly suggests that some transmembrane domains are dependent on C-terminal sequences for their correct topogenesis.

EXPERIMENTAL PROCEDURES
Materials and DNA Manipulations-DNA restriction and modification enzymes, Endo H, and in vitro mutagenesis reagents were pur-chased from Boehringer Mannheim. Sequenase TM version 2.0 and sequencing reagents were from U. S. Biochemical Corp. Lytic enzyme, factor Xa, and proteinase K were from ICN, Promega, and Sigma, respectively. All other chemicals and reagents were purchased from Boehringer Mannheim, Sigma, and British Drug House at analytical grade. Routine DNA manipulations and sequencing were carried out according to standard protocols (28,29).
Construction of a Yeast Strain with Conditional Regulation of SEC61-In order to generate a yeast strain with conditional expression of Sec61p, SEC61 was placed under the control of the GAL1 promoter. A 2.4-kbp HindIII fragment containing SEC61 from pCS43 (13) was cloned into pRS316 (32), followed by restriction of the single BamHI site (located in the polylinker), filling in with Klenow and dNTPs, and subsequent religation. A unique BamHI site was inserted immediately upstream of the SEC61 initiation codon by site-directed mutagenesis (33). The resulting plasmid was linearized with BamHI, filled in as above, and ligated with an 819-bp BamHI-EcoRI fragment containing the GAL1-GAL10 promoter region end-filled as above to generate pBW62 containing the GAL1 promoter inserted immediately upstream of SEC61. The diploid CSY110 (sec61::HIS3/SEC61) was transformed with pBW62. Sporulated transformants were subjected to tetrad dissection onto YPGAL using standard procedures (34). Spores were screened for a His ϩ Ura ϩ phenotype indicating the null sec61::HIS3 allele complemented by pBW62. Conditional expression from the GAL1 promoter (repression by glucose, induction by galactose) was confirmed by the complete inability of such spores to grow upon transfer to glucose supplemented medium. A resulting haploid strain, BWY47, was used to assess the function of sec61 mutants introduced on a plasmid (LEU2) with selection on galactose medium. Growth of transformants after transfer to glucose medium (gal shut-off) demonstrated the ability of mutants to provide Sec61p function. Counterselection of pBW62 (URA3) in BWY47-derived strains was carried out on 5-FOA containing medium (35), enabling Sec61p fusions encoded on a second plasmid to be shuffled into the null mutant background.
Antibodies-The antiserum raised against a C-terminal epitope of Sec61p has been previously described (13). Antisera were raised in rabbits against a 15-residue N-terminal epitope of Sec61p corresponding to residues 12-26 (PFESFLPEVIAPERK) and tested for reactivity against Sec61p by immunoblotting. Sera recognized a protein migrating at 38 kDa on 10% SDS-PAGE gels, which was amplified by Sec61p overexpression. The antisera directed against both the Sec61p epitopes were used at a titer of 1/4000 in immunoblots. Anti-Suc2p antiserum was raised against a purified oligohistidine-Suc2p fusion protein as follows. A 2.2-kbp HindIII/PvuII, end-filled fragment containing the coding sequence of the mature portion of yeast invertase (Suc2p) from pCS29 (see below) was ligated into XhoI-linearized and end-filled pET-16b (Novagen). The oligohistidine fusion was expressed from pET-16b (Novagen) under the control of the T7 promoter (36). The fusion protein was purified by nickel affinity chromatography using materials and protocols from Qiagen. Antisera raised against this fusion protein recognized a protein migrating at 60 kDa by 10% SDS-PAGE gels and a heterogeneous species of 80 -85 kDa in yeast cell extracts grown in low (0.1%) glucose medium consistent with the constitutive and secreted forms of Suc2p, respectively. The antiserum was used at a dilution of 1/5000 in immunoblots. Polyclonal antisera against Kar2p and Sec63p were used at 1/3000 and 1/4000 in immunoblots, respectively.
Construction of Random Sec61p-Suc2p Fusion Proteins-Plasmid pCS29 containing a signal truncated form of the SUC2 gene was derived from pSEY304 (37) by excision of the polylinker by EcoRI/HindIII digestion and ligation in its place with the EcoRI/HindIII linker fragment from pUC118 (38). The unique SalI site was filled in and selfligated to bring the unique BamHI site into the same frame as the pBW11-BamHI mutants (see below). Subsequently, plasmid pCS30 expressing SEC61 and SUC2 in the same direction was made by cloning an end-filled 2.4-kbp HindIII fragment encoding SEC61 into SmaI-linearized pCS29. Plasmid pCS30 was jointly restricted with BamHI and SphI and subjected to unidirectional exonuclease III digestion according to Henikoff (39), end-filled, and ligated. Sequence analysis of isolated fusion candidates identified 8 in-frame fusions at residues Gln 156 , Gln 192 , Val 247 , Phe 256 , Gln 261 , Gln 394 , Arg 412 , and Glu 460 with Suc2p starting at the third residue of the mature protein.
Introduction of BamHI Sites into SEC61-Plasmid pBW11 was made by cloning a 2.4-kbp HindIII fragment encoding SEC61 (13) into Hin-dIII-linearized pRS315 (32). The unique BamHI site (located in the polylinker) was removed by BamHI digestion, end-filling, and religation. Plasmid pAC21 was made by cloning the same fragment into pRS425 (40) and removal of the unique BamHI site. Single-stranded templates prepared from pBW11 and pAC21 were used to introduce BamHI sites by substitution with appropriate oligonucleotides (33 The six pAC21-BamHI plasmids were digested jointly with BamHI and SmaI to remove the C-terminal region of SEC61, and a 2.2-kbp BamHI/ PvuII fragment from pCS29 was cloned in its place. BamHI sites at nucleotides 349, 637, 826, 1000, 1336, and 1423 were used to make Suc2p fusions to amino acid residues Ile 116 , Gly 213 , Gly 276 , Gly 334 , Gly 445 , and Gly 475 , respectively. In each case the Sec61p was fused to the third residue of the mature portion of Suc2p. Construction of Sec61-fXa Fusion Proteins-The complementary oligonucleotides GATCCATCGAGGGTAGAG and GATCCTCTACCCTC-GATG were annealed to generate an oligomer encoding the factor Xa tetrapeptide recognition motif (IEGR) on BamHI ends. This was then cloned into pBW11-BamHI plasmids with sites at nucleotides 211, 316, and 1399. DNA sequence analysis of potential recombinants was used to determine the number and orientation of inserts. The following constructs were used in this report: L 70 GSSTGDGSIEGRGSSTGDG-SIEGRGSN 73 ; P 105 GSIEGRGSK 108 ; and G 466 GSIEGRGSIEGRG ST 468 ). These are referred to as HS1x2, HS2x3, and HS10xC fusions, respectively. Plasmids expressing these fusions were introduced into BWY47 and subsequently "shuffled" by plating onto media containing 5-FOA producing strains expressing the fusion as the only form of Sec61p.
Construction of a C 5 Sec61p Fragment-The C-terminal Sec61p fragment C 5 consisting of HS6-10 was constructed by digestion of pBW11-40 jointly with BamHI and XhoI removing all but the extreme N-terminal coding region, followed by ligation with a 1.3-kbp BamHI/XhoI fragment from pBW11-688. The Sec61p-C 5 fragment consists of a fusion of residues NH 2 M1 3 F13-GS-L2323 M480COOH (264 amino acids). A 1.8-kbp HindIII fragment encoding Sec61p-C 5 was cloned into HindIIIlinearized pRS424 (42) to provide multicopy, TRP1-selectable expression of C 5 .
Preparation of Yeast Whole Cell Extracts-Yeast cultures were grown overnight to an A 600 of 0.2-0.5, and 10 A 600 units were collected. The cells were washed with 10 mM azide and resuspended in 3% SDS, 62 mM Tris-HCl, pH 6.8, 1 mM PMSF at 1 A 600 units/20 l. Glass beads (0.5 mm diameter) were added to the meniscus, and vortexing was carried out at top speed in a mini-beadbeater TM (Biospec Products, Bartlesville, OK) for 50 s. The samples were incubated at 50°C for 5 min, and the vortexing procedure was repeated. After a further incubation at 50°C for 10 min, the samples were boiled 5 min.
Preparation of Yeast Microsomes-200 ml of yeast cultures were grown in YPD to an A 600 of 2-4. Cells were collected and spheroplasts were prepared according to Sanders  Immunoblot Analysis of Proteins-After migration on SDS-PAGE gels, proteins were transferred electrophoretically to nitrocellulose filters (Hybond-C, Amersham Corp.). Preincubation, antibody incubations, and washes were performed in TBS (10 mM Tris-HCl, pH 7.6, 2 mM KCl, 150 mM NaCl) plus 0.1% Nonidet P-40 and 1% fat-free dry milk. Immunodetection was carried out using enhanced chemiluminescence (Amersham Corp.) with a peroxidase-conjugated anti-rabbit IgG (Sigma).

RESULTS
Hydropathic Profile of the Sec61p Sequence-The polypeptide chain of Sec61p is 480 amino acids in length, and its hydropathic profile is shown in Fig. 1A. The 10 most hydrophobic segments (HS) have been indicated. Several of these segments exhibit the extreme mean hydropathies typical of transmembrane domains (e.g. HS1, -3, -4, and -6) (13, 41). However, others are much less hydrophobic, and their membrane disposition is uncertain. The shortest hydrophobic segment being HS5 representing a stretch of 12 residues in which only 11 are nonpolar (see Fig. 1A).
Construction of Sec61p-Suc2p Fusion Proteins-We have analyzed the transmembrane topology of yeast Sec61p using a C-terminal reporter approach originally developed in Escherichia coli for the study of cytoplasmic membrane proteins (42) and adapted in yeast for the study of ER and plasma membrane protein topology (43)(44)(45). Suc2p was chosen as a topological reporter, because it does not possess any membrane topogenic preferences and becomes rapidly modified by asparaginelinked glycosylation at multiple sites upon translocation to the lumen of the ER, resulting in a 20 -26-kDa increase in molecular mass (46). Suc2p fusions have also been successfully applied in determining the topology of both Sec62p and Sec63p (30,47).
A total of 22 Sec61p-Suc2p fusions were constructed either randomly by exonuclease III deletion or by the use of specific BamHI restriction sites introduced at various points throughout the SEC61 coding sequence (see "Experimental Procedures"). In all cases the resulting fusion proteins were expressed from the native SEC61 initiation codon under the control of its own promoter. The precise positions of the fusions are shown in Fig. 1B.
Analysis of Fusion Protein Topology-Plasmids encoding the Sec61p-Suc2p fusion proteins were transformed into suitable yeast strains, and whole cell extracts were prepared from transformants grown under selective conditions. Extracts were incubated with Endo H to remove asparagine-linked oligosaccharide. Control samples were mock-digested in the absence of Endo H. Reaction products were resolved by SDS-PAGE and then immunoblotted with antisera raised against either Suc2p or Sec61p (Fig. 2).
In the case of a fusion in which Suc2p is fused to Sec61p after the glutamine residue at position 261 (Gln 261 ), a single immunoreactive species was observed in mock-digested extracts with a relative molecular mass of 96 kDa ( Fig. 2A, lane 9). This band is absent in control extracts prepared from cells carrying a vector plasmid (Fig. 2D, lanes 1 and 2). Upon Endo H digestion the immunoreactive band exhibited a higher gel mobility with a mass of 75 kDa ( Fig. 2A, lane 10). This gel mobility shift is indicative of the removal of 21 kDa of asparagine-linked oligosaccharides. From these results we concluded that the Gln 261 -Suc2p fusion is a glycoprotein, from which we infer that the Suc2p domain is oriented within the ER lumen. In contrast, the gel mobility of a Suc2p fusion at residue arginine 412 (Arg 412 ) is unaffected by Endo H treatment ( Fig. 2A, lanes 11 and 12), indicating that it is unglycosylated. This result is consistent with the Suc2p domain in this case being oriented toward the cytoplasmic compartment. A more complex result is seen for fusion Gln 156 ( Fig. 2A, lanes 1 and 2) where two forms of the fusion protein are observed in mock extracts, the larger of which shifts upon Endo H treatment, suggesting that this fusion exists as a mixture of glycosylated and unglycosylated forms at steady state. We interpret this result as indicating that fusion Gln 156 exists as a mixture of topological forms.
The data for all other fusion constructs shown in Fig. 2 were interpreted by the same criteria, taking into consideration the following notable points. First that extracts prepared from CSY142 transformants contain an immunoreactive band of 60 kDa also seen in a vector control extract (Fig. 2C, lanes 1 and  2) corresponding to the cytosolic form of Suc2p. Second, the observed relative molecular masses for the deglycosylated/unglycosylated forms of the Sec61p-Suc2p fusions fell in the range of 65 kDa (for fusion Leu 70 ), to 96 kDa (for fusion Gly 475 ). These compare with the predicted molecular masses of 68 to 112 kDa for Leu 70 to Gly 475 . The apparent anomaly for fusions carrying larger portions of Sec61p sequence would be expected given that Sec61p itself migrates aberrantly in SDS-PAGE (13). In all cases where an Endo H-induced shift in gel mobility was observed this shift was of the order of 20 -24 kDa consistent with a fully glycosylated Suc2p domain (46).
The most N-terminal fusion for which we have data is Leu 70 , which is extensively glycosylated (Fig. 2C, lanes 3 and 4). The simplest interpretation of this result is that the first 70 amino acids of Sec61p contain sufficient information to translocate the Suc2p domain to the ER lumen. The hydropathy analysis indicates that the first major hydrophobic segment (HS1) is formed by residues 33-55 (Fig. 1A). This result indicates that the C-terminal end of HS1 is located toward the ER lumen. The Leu 70 fusion can be detected with an antiserum raised against residues 12-26 of Sec61p (not shown), from which we conclude that HS1 is not cleaved during biogenesis. It therefore follows that HS1 spans the bilayer with its N-terminal end oriented toward the cytoplasm.
The next Sec61p-Suc2p fusions created at Sec61p residues Pro 105 and Ile 116 are both downstream of HS2. If HS2 spans the bilayer then one might have expected the C terminus of each fusion to be located in the cytosol and to therefore be unglycosylated. However, both Pro 105 and Ile 116 were found to be expressed exclusively as glycoproteins (Fig. 2D, lanes 5 and 6;  Fig. 2B, lanes 1 and 2) indicating that HS2 does not span the bilayer in the context of these fusion proteins. Yet more surprising was the finding that fusion Ala 141 is also expressed exclusively as a glycoprotein (Fig. 2D, lanes 7 and 8). The simplest interpretation of these results would be to conclude that neither HS2 nor HS3 is capable of spanning the membrane. However, this would seem unlikely given the extremely hydrophobic nature of HS3 (Fig. 1A). Residue Ala 141 is in a very short hydrophilic region between HS3 and HS4 (Fig. 1B), and this proximity to the end of HS3 may interfere with its membrane insertion. An alternative explanation for this anomaly would arise were HS2 to require the presence of HS3 in order to assemble into the bilayer. Under these circumstances HS2 would be excluded from the membrane in fusions Pro 105 and Ile 116 , but both HS2 and HS3 would span the bilayer in fusion Ala 141 . Evidence for such an interaction will be presented in a later section.
The glycosylation of Ala 141 is consistent with the hydrophilic loop between HS3 and HS4 being located on the lumenal face of the ER membrane. A fusion at Ser 179 is then unglycosylated (Fig. 2D, lanes 9 and 10) consistent with HS4 spanning the membrane with its C terminus toward the cytoplasmic compartment. A fusion within HS4, at Gln 156 , is expressed as a mixture of glycosylated and unglycosylated forms ( Fig. 2A,  lanes 1 and 2) suggesting that the truncated HS4 domain can function, albeit inefficiently, as a stop-transfer domain during fusion protein biogenesis.
Fusions Gln 192 and Gly 213 occur at, or near, the beginning of HS5 and, like Ser 179 , are unglycosylated ( Fig. 2A, lanes 3 and  4; Fig. 2B, lanes 3 and 4). However, fusion Lys 229 that lies clearly within a hydrophilic sequence between HS5 and HS6 (Fig. 1B) is also unglycosylated (Fig. 2D, lanes 11 and 12), suggesting that the relatively short HS5 does not span the bilayer. Fusions Phe 256 and Gln 261 are exclusively glycosylated ( Fig. 2A, lanes 7-10) consistent with HS6 being the next transmembrane sequence. This simple interpretation is made more complex by the finding that, while Gln 261 is glycosylated, the vast majority of Tyr 265 (Fig. 2C, lanes 5 and 6) and all detectable Gly 276 (Fig. 2B, lanes 5 and 6) are found to be unglycosylated. These results suggest the existence of two transmembrane spanning domains between residues Lys 229 and Gly 276 and appear to exclude HS5 as a candidate for a transmembrane domain. However, data provided in a following section indicate that the interpretation of the Sec61p-Suc2p fusion data in this region is complicated by an interaction between HS5 and downstream sequences.
The remaining fusion data appear to be consistent with HS7, -8, -9, and -10 having the ability to span the ER membrane. Fusion Gly 276 places the sequence between HS6 and HS7 within the cytoplasm. Fusions Gln 334 and Ser 351 were both very substantially glycosylated (Fig. 2B, lanes 7 and 8; Fig. 2C, lanes 7 and 8) consistent with HS7 acting as a transmembrane domain. The subsequent absence of glycosylation in fusions Gln 394 (Fig. 2D, lanes 3 and 4) and Arg 412 ( Fig. 2A, lanes 11 and  12) is consistent with HS8 spanning the bilayer with its C terminus on the cytosolic face of the membrane. A fusion immediately following HS9 at residue Ser 436 is not glycosylated (Fig. 2C, lanes 9 and 10), whereas a small proportion of fusions Gly 445 and Glu 460 is glycosylated (Fig. 2B, lanes 9 and 10; Fig.  2A, lanes13 and 14 ). Finally fusion Gly 475 is again unglycosyl-

FIG. 2. Analysis of the glycosylation status of Sec61p-Suc2p fusions expressed in yeast.
Cells were grown in minimal medium with appropriate supplements and selection for the plasmid expressing the fusion. Protein extracts prepared by glass bead lysis of harvested cells were subjected to Endo H digestion. Two A 600 equivalent units were diluted 15-fold into 50 mM sodium phosphate, pH 6.0, containing 0.5 mM PMSF and 2 units of Endo H followed by incubation at 37°C for 4 h. Controls were mock-digested. After trichloroacetic acid precipitation, 1.0 A 600 equivalent units from each reaction was resolved by 10% SDS-PAGE and immunoblotted with Suc2p antiserum unless stated otherwise. Fusions are identified above their mock (Ϫ) and Endo H (ϩ)-treated lanes. A, fusions expressed in strain YT455 and detected using anti-Suc2p antiserum. B, fusions expressed in strain CSY142 immunoblotted with antiserum directed against the N terminus of Sec61p. C, all fusions in this panel were expressed in CSY142 as well as the control vector alone (pAC21 ; CON, lanes 1 and 2). D, all fusions in this panel were expressed in CSY142, with the exception of Gln 394 that was expressed in YT455. The control (CON, lanes 1 and 2) represents samples derived from YT455 containing control vector (pCS30) alone. Please note that an unglycosylated band of 115 kDa was intermittently detected in CSY142 extracts (see C, lanes 3, 4, 9,and 10, and D lanes 5 and 6). This band does not cross-react with the N-terminal antiserum and should not be confused with a glycoprotein of similar size detected in A, lane 13, which is detectable with both antisera (not shown) and that therefore clearly corresponds to a glycoform of fusion Glu 460 . ated (Fig. 2B, lanes 11 and 12). These data suggest the presence of two transmembrane sequences in this region. We therefore interpret these data to suggest that HS9 is capable of spanning the bilayer with an end point between Ser 436 and Gly 445 and that HS10 also spans the membrane with an end point between Glu 460 and Gly 475 . Both HS9 and HS10 are only moderately hydrophobic and are linked by a very short hydrophilic loop. It is therefore possible that HS9 and HS10 may interact with one another during normal topogenesis. This might explain the very limited glycosylation of fusions Gly 445 and Glu 460 . The Sec61p-Suc2p fusion data are summarized in Fig. 1B.
Hydrophobic Segment 2 Requires C-terminal Sequences in Order to Adopt a Transmembrane Topology-The C-terminal reporter data described above suggests the absence of functional transmembrane domains between residues Leu 70 3 Ala 141 , despite the fact that this region contains two strikingly hydrophobic segments, HS2 and HS3 (Fig. 1A). The C-terminal fusion approach might be expected to be problematic should any given domain be dependent upon downstream sequences for its correct membrane assembly. Such a consideration led us to develop an alternative strategy that was not dependent upon radical C-terminal deletions. We have employed an insertion approach similar to reported methods (48 -50) using the recognition motif for a highly specific protease, factor Xa, with the intention of using in situ cleavage of these Sec61p derivatives in microsomes as an assay for the topology of the cleavage site. This protease recognizes the tetrapeptide motif IEGR, cleaving C-terminal to the arginine residue (51). We inserted this recognition motif into hydrophilic segments 1/2 (at Leu 70 ) and 2/3 (at Pro 105 ) in order to examine the potential of hydrophobic stretches HS2 and HS3 to span the bilayer in the context of the intact Sec61p molecule. An insertion containing two tandem fXa sites at Leu 70 (HS1x2) and a single site at Pro 105 (HS2x3) were subsequently used (see "Experimental Procedures").
Any insertion into a multi-spanning integral membrane protein might interfere with either the targeting or assembly of the protein, thus invalidating any topological assessments. In order to guard against this possibility, we have examined the ability of these different Sec61p-fXa derivatives to functionally complement the normally lethal sec61::HIS3 null mutation. Single copy plasmids encoding Sec61p-fXa fusions HS1x2 or HS2x3 were capable of complementing the sec61::HIS mutation resulting in cells with normal growth rates in both cases (Fig.  3). The ability of these fusions to provide the essential function of Sec61p represents the best possible evidence that these insertion mutants attain their native conformation and are correctly assembled in the ER membrane. All subsequent experiments with these fusions were performed with strains after passage on 5-FOA medium to remove the GAL-SEC61 plasmid, thus shuffling the fusions into the null mutant background (see "Experimental Procedures").
Initially, the ability of fXa to cleave these various Sec61p-fXa derivatives was tested in detergent-solubilized membrane preparations. Immunoblot analysis of cleavage products revealed that both HS1x2 and HS2x3 proteins could be efficiently cleaved after membrane solubilization with Nonidet P-40 (Fig.  4A). Before proceeding to cleave these derivatives in situ, we first tested the membrane orientation and integrity of our microsome preparations by examining the protease sensitivity of the ER membrane protein, Sec63p (10,30), and the ER lumenal protein, Kar2p (53).
Yeast microsomes were exposed to proteinase K in the presence or absence of 0.2% Nonidet P-40 (sufficient to permeabilize membranes), prepared for SDS-PAGE, and analyzed by immunoblotting with either Sec63p or Kar2p antiserum. Using membranes derived from cells expressing the HS1x2 derivative, we found that the 73-kDa protein Sec63p was completely susceptible to proteinase K resulting in a major 68-kDa protected form (Fig. 4B) consistent with reported data (30). Crucially only trace quantities of a 43-kDa species produced in the presence of detergent (30) were observed in the absence of detergent, indicating the presence of a very low level of broken membranes (Fig. 4B). The integrity and stability of these membranes was further confirmed by the complete protection of the ER lumenal protein Kar2p from proteinase K digestion unless membranes were first permeabilized with 0.2% Nonidet P-40 (Fig. 4C). These results demonstrate that our microsomal membranes were predominantly comprised of sealed membrane vesicles oriented with their cytoplasmic side facing out.
Having established suitable conditions for topological cleavage, in situ cleavage of Sec61p-fXa proteins was examined. Microsomes prepared from strains expressing either the HS1x2 or HS2x3 fusion proteins were incubated with fXa in the presence or absence of 0.2% Nonidet P-40. The reaction products were analyzed by SDS-PAGE and immunoblotted with antisera directed against the C terminus of Sec61p (Fig. 4D). Native Sec61p does not contain any fXa recognition sites and was not cleaved under the conditions described (not shown). An fXa cleavage product was generated from HS1x2 membranes only under conditions of permeabilization (Fig. 4D, lane 3) strongly suggesting that the fXa site and hence segment 1/2 are located in the ER lumen, consistent with the Suc2p fusion data. With HS2x3 microsomes, an fXa cleavage product was generated in the absence of permeabilization (Fig. 4D, lane 5), and the appearance of this product was not affected by 0.2% Nonidet P-40 (Fig. 4D, lane 6). These data are only consistent with a cytosolic location of the fXa site and hence hydrophilic segment 2/3, from which it follows that HS2 must span the bilayer in the intact protein. This result conflicts with the observation that Suc2p fusions within loop 2/3 (at residues P105 and I116) report a lumenal topology for Suc2p indicating that HS2 does FIG. 3. Functional analysis of Sec61p-fXa fusion proteins expressed in yeast. The ability of fXa fusion proteins to provide Sec61p function was examined by introduction of low copy plasmids expressing these fusions into BWY47. This haploid strain contains a chromosomal null mutation (sec61::HIS3) that is rescued by the presence of a low copy number URA3-GAL1/SEC61 plasmid allowing growth on galactose but not glucose containing medium. Transformants selected on galactose containing medium were plated to minimal medium containing appropriate supplements with 2% galactose or 2% glucose but lacking leucine and uracil. All strains grew on galactose medium, whereas the vector control (pRS315) was unable to grow on glucose containing medium since it was unable to provide Sec61p function. The growth on glucose of strains expressing fXa fusion proteins HS1x2, HS2x3, and HS10xC comparable with the wild type SEC61 control (plasmid pBW11) demonstrated their ability to provide Sec61p function. not span the bilayer in the context of these fusion proteins.
Analysis of the C-terminal Membrane Location by fXa Insertion-Factor Xa recognition site insertion was also used to examine the membrane location of the Sec61p C terminus. A BamHI site created at the codons representing residues Gly 467 / Phe 468 was used as the site of insertion for two tandem fXa motifs to generate Sec61p mutant HS10xC. A low copy plasmid encoding this fusion was able to complement the sec61::HIS3 null mutation (Fig. 3) confirming that the insertion of fXa motifs into the C terminus does not significantly perturb the structure and function of Sec61p. Microsomes were prepared from cells expressing HS10xC and tested for factor Xa cleavage both in situ and after membrane solubilization. The solubilized fusion protein proved to be an efficient substrate for fXa protease yielding a fragment detected by immunoblotting with antisera raised against the N terminus of Sec61p migrating slightly faster than the intact fusion protein (Fig. 4A, lane 6). Cleavage in situ was detected in the absence and presence of detergent (Fig. 4D, lanes 8 and 9). These data are consistent with the location of the C terminus on the cytosolic face of the membrane as determined with the Suc2p fusion made to residue Gly 475 .

Hydrophobic Segment 5 Can Span the Bilayer in the Presence of C-terminal Sequences That Are Capable of Acting in
Trans-We sought to investigate the inconsistencies obtained with the Suc2p fusion analysis around hydrophobic regions HS5 and HS6. The topology of the HS5 and HS6 region was investigated by fXa site insertion at residues Ser 179 and Lys 229 , but despite retaining function the Sec61p-fXa proteins could not be cleaved in situ (not shown). Interestingly, while neither a 243-residue polypeptide encoding HS1-5 (N 5 ) nor a 264residue polypeptide encoding HS6 -10 (C 5 ) could complement sec61::HIS3, the co-expression of these two fragments provided functional complementation of sec61::HIS3. 2 This result clearly indicates that N 5 must assemble into its correct, functional topology at least in the presence of C 5 . We therefore re-examined the topology of fusion Lys 229 (N 5 -Suc2p) in the presence of the C 5 domain.
A low copy LEU2 plasmid encoding the Lys 229 fusion was co-transformed together with a multicopy TRP1 plasmid encoding the C 5 fragment into the sec61::HIS3 strain, BWY47. Remarkably, transformants grew upon gal shut-off, although with a severe growth defect (not shown), indicating that at least some of the Lys 229 -Suc2p/C 5 fragments could interact to provide Sec61p function. Membranes prepared from a strain expressing only the Lys 229 -Suc2p/C 5 form of Sec61p were analyzed by immunoblotting using antisera directed against Suc2p and the C terminus of Sec61p. The C 5 fragment and three higher molecular weight species were detected (Fig. 5, lane 1). Of the three fusion protein bands, two migrated as a doublet similar in molecular weight to the unglycosylated form of the fusion protein detected previously (Fig. 2D, lane 11). Endo H digestion revealed that the third, and largest, fusion protein species was glycosylated (Fig. 5, lane 2) indicating that HS5 can become transmembrane in the presence of its cognate C 5 fragment. These data suggest that HS5, a weakly hydrophobic membrane anchor, requires downstream sequences for its stable topogenesis and that these sequences can be provided in trans. It is not known if the effect of this co-expression is to promote translocation of HS5 and the Suc2p moiety in the Lys 229 fusion or the stabilization of the glycoform that may be particularly unstable when individually expressed and hence not detected previously (Fig. 2D, lane 11).  5 and 6). B, membrane orientation and integrity in fXa buffer assayed by protease protection of Sec63p. Microsomes prepared from the strain expressing HS1x2 were resuspended in fXa buffer containing 250 mM sorbitol and 300 g of proteinase K/ml. Digestion was carried out on ice with (lanes 5-8) or without (lanes 1-4) 0.2% Nonidet P-40, and the reaction was terminated by addition of trichloroacetic acid to 20%. 0.1 A 280 microsome equivalent units were resolved by 10% SDS-PAGE and immunoblotted with Sec63p antiserum. C, analysis of membrane integrity in fXa buffer assayed by proteinase K protection of Kar2p. Reactions were performed on HS1x2 microsomes as described above with (lanes 4 -6) or without (lanes 1-3) 1-6) or the N terminus of Sec61p (lanes 7-9).

DISCUSSION
We have used a combination of approaches to determine the topology of yeast Sec61p in the ER membrane. Our data suggest the presence of 10 transmembrane domains as shown in diagrammatic form in Fig. 6. Given the high level of amino acid sequence identity and the similarity of hydropathy profiles, it would seem likely that Sec61␣ (19) also has a similar structure. The 10 transmembrane structures and orientation of membrane anchors of Sec61p are similar to the topological model proposed for the SecY subunit of the bacterial cytoplasmic membrane translocase (54). SecY has been widely implicated in protein translocation across the cytoplasmic membrane of E. coli (55) and appears to be distantly related to Sec61p (56). The hydrophobic segments in Sec61p vary enormously in their length and relative mean hydropathies (Fig. 1A). Our data indicate that several domains of quite limited hydrophobicity are incorporated into the ER bilayer where they might play an important role in the formation of an aqueous protein-conducting channel.
Our data indicate that HS1 spans the bilayer with its N terminus oriented toward the cytoplasmic compartment. While, by inference, this would place the N terminus of Sec61p in the cytoplasm, we cannot exclude the possibility that sequences N-terminal to HS1 might be buried within or might even span the bilayer. In particular, residues 3-21 have the potential to form a highly amphipathic ␣-helical structure. Intriguingly, the introduction of an fXa tetrapeptide motif (IEGR) within this region resulted in the loss of Sec61p function, whereas insertion of an inverted motif (STLD) retained function. 3 Significantly, the fXa motif is predicted to disrupt the amphipathic helix, whereas the inverted motif could conform to an amphipathic structure. These findings would be consistent with the predicted amphipathic helix playing an important role in Sec61p biogenesis and/or function. A structure such as this may be partially embedded in the plane of the bilayer (as shown in Fig. 6) or might actually span the bilayer with its hydrophilic surface lining the interior of the translocation pore. Such a domain might play an important role in the gating of the translocation pore, either during the initiation of translocation or perhaps to permit the lateral diffusion of hydrophobic domains into the lipid bilayer during the assembly of integral membrane proteins. The potential importance of this putative amphipathic structure is underlined by its conservation both in Sec61␣ (19) and in a Schizosaccharomyces pombe homologue. 4 Our data indicate that both HS2 and HS5 require C-terminal sequences for their topogenesis. In the case of HS2, the Cterminal fusion data clearly report that neither HS2 nor HS3 span the bilayer. However, in the intact Sec61p (HS2x3) the presence of downstream sequences facilitates the stable membrane insertion of HS2 resulting in the observed cytoplasmic orientation of the HS2x3 loop. If the "stabilizing" sequence were itself a transmembrane domain then one would predict that the inclusion of this domain should lead to the coordinated assembly of HS2. This would create an apparent anomaly in our Suc2p-fusion data in which the addition of one hydrophobic segment would lead to the formation of two transmembrane domains, such that the topology of the Suc2p reporter would be unaltered. Exactly such an anomaly appears to occur in fusion Ala 141 , where the addition of the extremely hydrophobic HS3 domain has no apparent effect on the topology of the fusion protein. While our data present no direct evidence that HS3 actually spans the ER membrane, the simplest interpretation of our data would be to conclude that HS3 functions as a signal-anchor sequence and that it serves to promote, or stabilize, the membrane insertion of HS2. Our data do not indicate whether the proposed interaction between HS2 and HS3 occurs after their sequential insertion in the bilayer or whether it is required to mediate their pair-wise insertion in a manner similar to that described for domains of human P-glycoprotein (57,58), a plasma membrane ATPase (59), and lactose permease (50). The nature of the interaction between HS2 and HS3 is unclear. HS2 contains a conserved glutamate residue, Glu 79 , which may reduce the ability of this domain to act as a stop transfer sequence in the context of fusions Pro 105 and Ile 116 . However, HS3 does not contain a cognate positively charged residue that would allow the formation of a salt bridge, as has been suggested in the stabilization of the M9-M10 domains of lactose permease (60).
Our results also indicate that HS5 requires downstream sequences in order to function as a signal anchor sequence. In this case the Suc2p moiety in fusion Lys 229 could be translocated to the ER lumen only when the C-terminal portion of Sec61p was present in trans. This effect is clearly specific to the cognate C-terminal fragment since intact Sec61p does not promote the glycosylation of Lys 229 when present in trans (in strain CSY142; see Fig. 2). This, coupled to the observed complementation of a sec61 null mutant, leads us to conclude that the Lys 229 fusion interacts directly with the C-terminal fragment of Sec61p to form a functional ER translocase. Our current data do not indicate whether the assembly of the glycosylated form of Lys 229 occurs co-or post-translationally. Should it prove to be post-translational, then this particular phenomenon may be uniquely dependent upon the role of Sec61p as a protein translocase. In other words, the insertion of Lys 229 with a cytoplasmically oriented C terminus may occur by default. Subsequent interaction with the C-terminal fragment would then facilitate the stable integration of HS5 into the bilayer. Under these circumstances, there might be a window of opportunity, prior to the folding of the Suc2p domain, where the assembly of a functional translocation channel might facilitate the "autocatalytic" translocation of the Suc2p domain into the ER lumen giving rise to a glycosylated form. This hypothesis will require further study.
Clearly, sequences C-terminal to HS5 are required for its topogenesis. Our Suc2p fusion data around this region are complex but appear to indicate that the stabilizing sequences occur in, or around, HS6. Fusion Lys 229 is unglycosylated indicating a failure of HS5 to form. Fusions Phe 256 and Gln 261 are then completely glycosylated, whereas Tyr 265 is again fully unglycosylated (Fig. 1B). These data might indicate that transmembrane domain 5 extends significantly beyond residue Lys 229 despite the fact that residues 225-238 are predominantly charged. There are several alternative explanations for these findings, but given the partial glycosylation of Val 247 and the assembly of Lys 229 in the presence of the C 5 fragment, we believe that the most likely explanation to be as follows. First, that sequences between Lys 229 and Phe 256 serve to stabilize the formation of transmembrane domain 5. Second, that the Cterminal boundary of transmembrane domain 6 extends beyond Gln 261 , such that the additional residues present in fusion Tyr 265 are required for it to span the bilayer. HS5 may contain at least one charged residue dependent upon its exact position, but the lack of any charged residues in HS6 again appears to rule out the formation of a salt bridge between these domains. It is worth noting that HS5 corresponds to a stretch of 12 residues of which 11 are nonpolar (residues Gly 213 -Val 224 ), flanked on both sides by several polar residues. Our data do not indicate the precise end points of any given transmembrane domain, only that particular hydrophobic segments have membrane spanning potential. For example, HS5 may span the membrane in ␣-helical conformation thus incorporating numerous polar residues in the bilayer. Alternatively, HS5 would be sufficiently long to span the bilayer in ␤-sheet conformation. Either scenario would be consistent with the observation that HS5 requires some stabilizing interaction(s) within the bilayer.
The proposed topology of Sec61p (Fig. 6) provides a basis for the functional dissection of this essential protein. Deletion of individual domains would determine whether the essential function of Sec61p can be retained upon loss of significant portions of the protein, and a systematic isolation and characterization of point mutations defective in Sec61p function would identify the critical residues and domains. Using the deduced topology the position of mutations with respect to the ER membrane would be known, thus aiding the detailed molecular description of the function(s) of Sec61p in ER translocation.