Membrane insertion and intracellular transport of yeast syntaxin Sso2p in mammalian cells

The compartmental organization of eukaryotic cells requires targeting and retention of the resident proteins in the appropriate compartments. The integral membrane proteins are inserted in the lipid bilayer either co-translationally during their biosynthesis or post-translationally after they have been synthesized in the cytoplasm. The classical signal sequence for translocation of secretory or membrane proteins to the endo-plasmic reticulum (ER) is usually located at the N terminus and is able to bind the signal recognition particle (SRP), which guides the protein to the ER membrane (for a review see Rapoport, 1992). Integral membrane proteins are retained in the membrane either by the signal sequence or by one or more hydrophobic amino acid stretches following it, called stop transfer sequences. These proteins expose either the amino (type I) or the carboxy (type II) terminus to the exoplasmic side of the membrane (for a review see High and Dobberstein, 1992). Membrane proteins destined to different compartments of the secretory pathway such as the Golgi complex or the plasma membrane (PM) are first targeted to the ER and subsequently transported to their final locations.

A number of membrane proteins functioning in the vesicle-mediated transport from the ER to the plasma membrane have been identified during the past few years. Many of these fall into a specific group of type II membrane proteins, termed tail-anchored proteins, which carry a single hydrophobic region at or near the C terminus (see Kutay et al., 1993). Several of the recently identified proteins functioning in vesicular traffic are conserved from yeast to mammals (see Bennett and Scheller, 1993; Ferro-Novick and Jahn, 1994). Furthermore, similar proteins are involved in the different transport steps between the intracellular compartments (see Aalto et al., 1993). Homologues of the synaptobrevin/VAMP proteins of synaptic vesicles (Südhof et al., 1993) have been identified in non-neuronal higher eukaryotic cells as well as on membranes of the yeast secretory pathway (see Ferro-Novick and Jahn, 1994 and references therein). Another set of conserved tail-anchored proteins involved in membrane transport are syntaxins of neuronal (Bennett et al., 1992; Inoue et al., 1992; Yoshida et al., 1992) and non-neuronal cells (Bennett et al., 1993), homologues of which also function in several steps of the yeast secretory process (Sed5p, Hardwick and Pelham, 1992; Pep12p, EMBL database/GenBank M90395; Sso1p, Sso2p, Aalto et al., 1993). The synaptobrevins and the syntaxins have been found in a 20 S complex which is thought to be respon-sible for the docking of synaptic vesicles at the presynaptic plasma membrane (Wilson et al., 1993; Söllner et al., 1993a,b; Calakos et al., 1994). Synaptobrevin/VAMP-like proteins residing on transport vesicles are denoted as v-SNAREs (vesicle-associated soluble NSF attachment protein (SNAP) receptors) and syntaxin-like molecules of the target membrane as t-SNAREs (target membrane-associated SNAP receptors). The presence of synaptobrevin and syntaxin homologues in various types of eukaryotic cells and their functions in the different transport steps suggest that these membrane proteins may in all eukaryotic cells form an important part of the apparatus conferring specificity to membrane recognition and fusion events.

It is evident from the sequence data available that, in addition to the C-terminal anchor, the tail-anchored proteins do not contain other hydrophobic sequences that could guide the protein into the target membrane. In all cases studied, it has been shown that the protein is bound to the membrane via the C-terminal hydrophobic sequence and the bulk of the polypeptide resides on the cytoplasmic side of the membrane (Carmichael et al., 1982; Markland et al., 1986; Chen-Levy and Cleary, 1990; Frangioni et al., 1992; Mitoma and Ito, 1992; Bennett et al., 1993). However, the mechanisms and the sites of membrane insertion of these proteins have remained elusive. Tail-anchored proteins are found on cellular compartments outside the ER (Kutay et al., 1993). On the other hand, e.g. cytochrome b₅ appears to be inserted to ER membranes in an SRP-independent manner (Anderson et al., 1983; Mitoma and Ito, 1992). It is thus possible that these proteins are directly inserted to their target membranes. In the case of SNAREs, this could be a means of circumventing problems encountered upon transport of these regulatory proteins through the secretory pathway.

We recently isolated cDNAs encoding the S. cerevisiae Sso1 and Sso2 proteins homologous to neuronal cell syntaxins (Aalto et al., 1993). These cDNAs suppress mutations in several genes acting in the Golgi-PM step of the yeast secretory pathway, and the proteins are known to localize to the yeast PM (Brennwald et al., 1994). In this study we transiently expressed Sso2p in baby hamster kidney (BHK) cells. The localization and the mode of membrane insertion of this protein were investigated by immunocytochemical and biochemical methods.

Brefeldin A was purchased from Epicenter Technologies. Other reagents, if not otherwise stated, were from Sigma. Restriction enzymes and other standard molecular biology reagents were from New England Biolabs.

Baby hamster kidney cells (BHK-21 CCL10, C-13; American Type Culture Collection CRL 8544) were grown in Glasgow’s modified Eagle’s medium (G-MEM) supplemented with 5% foetal calf serum (FCS; Life Technologies), 2 mM L-glutamine, 10% tryptose phosphate broth, 100 i.u./ml penicillin, and 100 μg/ml streptomycin. Viral infection and low-temperature incubations of BHK-21 cells were performed in bicarbonate-free Eagle’s minimal essential medium with Earle’s salts (MEM), 20 mM HEPES, pH 7.4, 0.2% bovine serum albumin (BSA). Methionine-free Dulbecco’s modified Eagle’s medium (D-MEM) was used in the pulse-labeling.

The SSO2 open reading frame (ORF) was isolated from the original cDNA (YEpSSO2; Aalto et al., 1993) by PCR and tranferred via pSP73 (Promega) to the unique BamHI site of pSFV1 (Liljeström and Garoff, 1991; Life Technologies). A modified SSO2 cDNA encoding a protein that lacks the membrane-spanning sequence and the four very C-terminal, supposedly exoplasmic amino acids (Sso2p-ma, truncated at Arg²⁶⁹; Aalto et al., 1993) was created by PCR and cloned into pSFV1 as above. Recombinant SFVs carrying the SSO2 or SSO2-ma inserts were prepared as described (Liljeström et al., 1991; Olkkonen et al., 1994). The viral stocks were titrated on BHK-21 cells by immunofluorescence microscopy.

The SSO2-ma cDNA was transferred as an EcoRI-SalI fragment from pSP73 to pK212-2 (Dr Kari Kein änen, VTT Biotechnology and Food Research, Helsinki, Finland) and overexpressed in E. coli BL21 DE3 pLysS (Studier et al., 1991). The protein was purified from the cells by preparative SDS-PAGE for immunization of rabbits. The antibodies were affinity-purified with a GST-Sso2p-ma fusion protein (prepared using the Pharmacia pGEX-1λT vector and purified with glutathione Sepharose 4B) coupled to CNBr-activated Sepharose 4B (Pharmacia). For western blotting, cells were solubilized in reducing SDS-PAGE loading buffer and the proteins were resolved in 12.5% Laemmli (1970) gels and blotted onto nitrocellulose (Towbin et al., 1979). Unspecific binding of antibodies was blocked with 5% fat-free milk in TBS (10 mM Tris-HCl, 150 mM NaCl, pH 7.4), the primary antibody incubation was performed in 0.5% milk/TBS, and horseradish peroxidase-conjugated goat anti-rabbit IgG and 4-chloro-1-naphthol (Bio-Rad) were used for detection according to the manufacturer’s instructions.

Cells grown on coverslips or on 6 cm Petri dishes were cooled on ice and washed twice with ice-cold PBS. Recombinant Semliki Forest virus diluted in cold HEPES-buffered MEM, 0.2% BSA was bound to the cells for 30 minutes on ice, after which the cells were (=infection zero time point) transferred to 37°C for 30 minutes to allow virus entry, followed by transfer to prewarmed complete G-MEM and subsequent incubation in 5% CO₂, 37°C. A synchronized burst of Sso2p expression was generated by adding cycloheximide (100 μg/ml) to the infected cells at 2 hours 30 minutes p.i., followed by a 2 hour incubation at 37°C. The cells were then washed 5 times with prewarmed HEPES-buffered MEM, and incubation in this medium was continued for 1-20 minutes, after which the cells were fixed.

BHK-21 cells grown on coverslips were fixed for 30 minutes with 3% paraformaldehyde in 250 mM HEPES, pH 7.4, 0.1 mM CaCl₂, 0.1 mM MgCl₂, free aldehyde groups were quenched with 50 mM NH₄Cl/PBS, and the cells were permeabilized for 30 minutes with 0.05% Triton X-100/PBS at room temperature. Unspecific binding of antibodies was blocked by a 30 minute incubation with 1% BSA/PBS. The primary (anti-Sso2p and anti-mannosidase II monoclonal antibody (mAb) 53FC3, kindly provided by Drs B. Burke and G. Warren) and the secondary (FITC-conjugated goat anti-rabbit or TRITC-conjugated goat anti-mouse F(ab)₂, TAGO Immunologicals) were incubated for 30 minutes at 37°C. The coverslips mounted in Mowiol (Hoechst) were viewed and photographed with a Zeiss Axiophot photomicroscope using Kodak T-Max 400 ASA film.

BHK-21 cells were fixed with 0.06% glutaraldehyde + 3% paraformaldehyde in 250 mM HEPES, pH 7.4, 0.1 mM CaCl₂, 0.1 mM MgCl₂ for 1 hour at RT. After fixation free aldehyde groups were quenched for 15 minutes with 50 mM NH₄Cl/PBS. Permeabilization, antibody incubations, development of the peroxidase reaction and embedding were performed essentially as previously described (Brown and Farquhar, 1989). The samples were viewed with a JEOL 1200 EX microscope operated at an acceleration voltage of 60 kV.

BHK-21 cells grown on 6 cm dishes and infected with the recombinant SFVs for 4 hours were incubated in methionine-free D-MEM for 30 minutes, after which they were labeled for 1 hour using 70 μCi/ml L-[³⁵S]methionine (1000 mCi/mmol, cell labeling grade, Amersham). The cells were washed twice with ice-cold PBS, scraped, collected by centrifugation, resuspended in 10 mM HEPES, pH 7.4, 10 mM KCl, and broken by repeated passages through a 21 G needle. The supernatant obtained after a low-speed centrifugation step (+4°C, 5000 g, 20 minutes) was subjected to ultracentrifugation (+4°C, 100,000 g, 1 hour) to harvest total microsomes. The microsome pellets solubilized in NET buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 50 i.u./ml Trasylol™, Bayer, 0.5 mM 4-amidinophenyl-methansulphonylfluoride, Boehringer Mannheim) and the supernatants (cytosol fractions) adjusted to contain the same buffer were subjected to immunoprecipitation with the anti-Sso2p antiserum and Protein A-Sepharose (Pharmacia) essentially as described (Kuismanen et al., 1992). Some microsome pellets were resuspended in 0.1 M Na₂CO₃, pH 11.3, followed by recentrifugation at 100,000 g and solubilization in NET. The precipitates were analyzed under reducing conditions in 12.5% SDS-polyacrylamide gels, which were fixed with 10% acetic acid and treated for 30 minutes with Amplify™ (Amersham) before drying and exposure of Kodak X-OMAT™ AR films at −70°C. The relative amounts of the proteins were analyzed by scanning the autoradiographs with a laser scanner (LKB).

The SSO2 and SSO2-ma pSP73 constructs were in vitro transcribed and translated using the Promega TNT™ T7 coupled reticulocyte lysate system in the presence or the absence of canine pancreatic microsomal membranes (Promega) according to the manufacturer’s instructions. The β-lactamase mRNA supplied with the microsomes and an interferon-α1 cDNA in pGEM3 (kindly provided by Dr I. Julkunen, National Public Health Institute, Helsinki, Finland) were used for translation of signal sequence-containing control proteins. For analysis of post-translational insertion, translations were performed without microsomes for 90 minutes at 30°C. Thereafter, RNAseA and DNaseI (both at 100 μg/ml) or cycloheximide (1 mM) were added to the reactions for 10 minutes, followed by a 20 minute incubation with microsomes. Insertion of Sso2p to microsomes was assayed by pelleting the membranes through a cushion of 50 mM tri-ethanolamine, 250 mM sucrose, pH 7.5, in the Beckman Airfuge (+4°C, 140,000 g, 10 minutes). The pellets and the supernatants were analyzed by 15% SDS-PAGE and fluorography.

For morphological and biochemical analysis, we generated a rabbit antiserum against Sso2p produced in E. coli. The specific antibodies were affinity-purified and showed, in western blots of SSO2-Semliki Forest virus (SFV)-infected BHK cells or S. cerevisiae cells expressing Sso2p from a multicopy plasmid, a single protein band with the apparent size of 38 kDa (Fig. 1). This is in agreement with the Sso2p molecular mass predicted from the cDNA sequence, 32.5 kDa (GenBank/EMBL database accession number X67730). The endogenous yeast protein of identical size was also detectable with the antibodies (data not shown). The antibody showed no reactivity with uninfected BHK cells. This antibody was used to study the localization of Sso2p expressed in BHK cells using the SFV vector, which provides a convenient tool for relatively well synchronized expression of proteins in various cell types. Expression of the exogenous proteins is detected in the time scale of hours and the expression levels are easily regulated by adjusting the infection time (Liljeström and Garoff, 1991; Olkkonen et al., 1994).

BHK-21 cells were infected with SSO2-SFV, fixed at different time points, and processed for immunofluorescence microscopy. Mock-infected cells displayed only negligible, unspecific labeling (Fig. 2a). Sso2p first appeared in BHK cells at 2.5-3 hours post-infection (p.i.). At 3 hours p.i., the antibody uniformly detected a perinuclear structure with no significant labeling of other cellular membranes (Fig. 2c). The perinu-clearly accumulated Sso2p colocalized with the Golgi stack marker protein mannosidase II (Fig. 2d). At later stages of the infection, strong plasma membrane labeling was observed in addition to the perinuclear Golgi staining (Fig. 2e).

The intracellular distribution of Sso2p was studied in more detail by immunoelectron microscopy. At 5 hours p.i., peroxidase labeling was observed on the plasma membrane and throughout the Golgi stack (Fig. 3a and b, respectively). The labeling was often most prominent on the inner surface of cell processes and pseudopods (Fig. 3c). These results suggested that in SSO2-SFV-infected BHK cells Sso2p localizes to the plasma membrane; the time-dependent change in the labeling pattern indicated that the observed Golgi-associated labeling was due to protein in transit to the PM (see next paragraph). This localization of Sso2p is in agreement with its reported effect on the Golgi to plasma membrane traffic (Aalto et al., 1993) and its plasma membrane localization in yeast (Brennwald et al., 1994). Infected cells frequently displayed additional punctate labeling throughout the cytoplasm (Fig. additional punctate labeling throughout the cytoplasm (Fig. 3a).

Sso2p belongs to a family of proteins carrying a C-terminal membrane anchor. The membrane insertion site of these proteins has remained elusive (Kutay et al., 1993). In order to study the site of insertion of Sso2p and to follow the post-translational transport events of the protein, BHK-21 cells were infected with SSO2-SFV. At 3 hours p.i. cells were shifted for 2 hours to 20°C, a treatment that blocks the exocytic transport in the trans-Golgi network (Matlin and Simons, 1983; Saraste and Kuismanen, 1984; Kuismanen and Saraste, 1989). During this incubation, Sso2p accumulated in the Golgi, which became strongly stained as confirmed by double-labeling with Golgi mannosidase II (man II) (Fig. 4a,b). To study whether the protein in the Golgi was in transit to the plasma membrane, the cells were after 2 hours at 20°C shifted to 37°C in the presence of cycloheximide (50 μg/ml). Cells fixed at 15-30 minutes after the temperature shift showed brightly stained vesicular-appearing structures throughout the cytoplasm and staining of the PM (Fig. 4c). After 60 (data not shown) or 120 minutes chase in the presence of cycloheximide (Fig. 4e), the Golgi labeling of Sso2p had disappeared and the protein accumulated at the plasma membrane. The lack of Golgi labeling after the cyclo-heximide chase indicated that the protein was not significantly recycled from the PM to the Golgi.

The above experiments suggested that Sso2p is not inserted directly to the PM but is transported there via the biosynthetic membrane transport pathway. In the standard expression experiments we were, however, unable to see labeling of the endoplasmic reticulum. To study whether the association of Sso2p with cellular membranes could occur in the ER, SSO2-SFV-infected BHK cells were treated with the fungal antibiotic brefeldin A (BFA) (Takatsuki and Tamura, 1985). BFA causes extensive retrograde movement of Golgi membranes to the ER and blocks the ER exit of transported proteins (Lippincott-Schwartz et al., 1989, 1990; Doms et al., 1989). BFA (5 μg/ml) was added to the infected cells at 2 hours p.i. when Sso2p expression was not yet detected (Fig. 5a). When the cells were incubated in the presence of BFA at 37°C, the newly synthesized Sso2p accumulated in a reticular ER-like structure (Fig. 5c). Immuno-EM of the BFA-treated cells showed labeling on the outer nuclear membrane and associated ER cisternae (Fig. 3d), confirming the identity of the stained compartment. No staining of the plasma membrane was observed. It was thus obvious that, under these conditions, Sso2p was inserted to ER membranes and could not exit this compartment. When the BFA was removed, the protein progressed along the secretory pathway. Perinuclear Golgi structures labeled by anti-Sso2p emerged at 10 minutes after BFA washout (the 20 minute time point with prominent Golgi staining is shown in Fig. 5e), and immunostaining of the PM appeared at 30-60 minutes (not shown).

To study the membrane insertion site using a different experimental set-up, we devised a method allowing highly synchronous protein expression from the SFV vector. The cells were infected with SSO2-SFV for 2 hours 30 minutes to allow synthesis of the SFV RNA polymerase, followed by a 2 hour protein synthesis block with cycloheximide (100 μg/ml). During this time SFV-encoded messages accumulated in the cells (the situation after the cycloheximide block is shown in Fig. 5b), and upon wash-out of the drug, synchronous expression of Sso2p ensued. In specimens fixed at 5-10 minutes after cycloheximide wash-out (Fig. 5d), ER-like staining was seen throughout the cells. At these early time points, no Golgi immunofluorescence was observed. Double-labeling with man II showed that the Golgi remained intact during the entire course of the experiments (not shown). During longer chase times, Sso2p progressed to the Golgi complex with kinetics similar to that seen after the BFA washout: Golgi staining appeared at 15-20 minutes post cycloheximide wash-out (Fig. 5f). If the cells were at 5-10 minutes after the cycloheximide wash-out transferred for 30-60 minutes to 15°C (see Saraste and Kuismanen, 1984), Sso2p appeared in ER-like structures and in peripheral punctate pre-Golgi elements, but no juxtanuclear Golgi staining was seen (not shown).

To assess the role of the putative C-terminal membrane anchor of Sso2p, we constructed a truncated SSO2 cDNA encoding a protein (Sso2p-ma, predicted molecular mass 29.5 kDa) lacking the anchor sequence (the 26 C-terminal amino acids). The truncated and the intact proteins were expressed in BHK cells and their distributions between membrane and cytosolic fractions were studied. BHK-21 cells infected with the recombinant SFVs were pulse-labeled with [³⁵S]methionine, homogenized, and fractioned to obtain total microsomes and cytosol. Aliquots of the microsomes were washed with 0.1 M sodium carbonate, pH 11.3 to strip the membranes of any peripheral proteins. These samples were subjected to immunoprecipitation with the anti-Sso2p antibody. More than 90% of Sso2p was found in the total microsomal fraction (Fig. 6). Of this, over 90% was resistant to the carbonate wash (Fig. 6, hatched bar), and only approximately 7% was removed from the membranes by the treatment. In contrast, only 13% of the Sso2-ma was found in the microsomes and 87% was cytosolic. The carbonate wash removed Sso2-ma completely from the membranes and the protein was recovered in the supernatant. Taken together, these results show that Sso2p is an integral membrane protein retained in the membrane by its C-terminal anchor sequence.

Sso2p belongs to a class of proteins which are devoid of N-terminal signal sequence. It has been suggested that these proteins are inserted to membranes in a post-translational manner, since their single hydrophobic segment is too close to the C terminus to be exposed for SRP binding before translation terminates (Kutay et al., 1993). We therefore investigated the membrane insertion mechanism of Sso2p in vitro. The protein was translated in the presence or the absence of canine pancreatic microsomal membranes. Sso2p-ma and two signal sequence-containing proteins (interferon-α1 and β-lactamase) were used as controls in these experiments. Since signal sequence cleavage, glycosylation, or protection against protease digestion could not be used to detect Sso2p insertion to membranes, we assayed this by collecting the microsomes from the translation mixtures by centrifugation through a sucrose cushion. In the presence of microsomes, 20-40% of Sso2p associated with the membrane pellet (Fig. 7, lanes 3-4), whereas 95-100% of interferon-α1 and β-lactamase displayed signal sequence cleavage (Fig. 7, lanes 14 and 18, respectively) and were found in the microsome pellet. When detergent (1% Triton X-100) was added in the reaction mixtures to solubilize the microsomes, only trace amounts of Sso2p were found in the pellet (Fig. 7, lanes 5-6), indicating that the protein was truly microsome-associated. Sso2p-ma lacking the C-terminal anchor sequence showed no association with the microsomes (Fig. 7, lanes 11-12).

To study whether Sso2p can be inserted to the membranes post-translationally, canine microsomes were added to in vitro translation mixtures treated with nucleases or with cycloheximide to inhibit any further translation. In interferon-α1 or β-lactamase translation mixtures treated in this way, no signal sequence cleavage was observed (Fig. 7, lanes 15-16 and 19-20, respectively). In contrast, Sso2p was capable of post-translational membrane insertion, at an efficiency similar to that seen in the co-translational experimental set-up (Fig. 7, lanes 7-10).

Recently, a number of tail-anchored proteins have been shown to be essential components of the machinery regulating the membrane traffic along the secretory pathway both in yeast and in higher eukaryotic cells (for reviews see Bennett and Scheller, 1993; Ferro-Novick and Jahn, 1994). In this study we have investigated the intracellular distribution and membrane association of Sso2p (Aalto et al., 1993), a S. cerevisiae counterpart of the mammalian syntaxins. The protein was transiently expressed in BHK-21 cells using the SFV vector system. This, together with in vitro translation/membrane insertion experiments, allowed us to perform detailed immunocytochemical and biochemical analysis of the biosynthesis and the localization of Sso2p.

The SFV expression vector provides a relatively synchronous means of exogenous protein expression in the time scale of hours. This allowed us to follow the intracellular transport of Sso2p. At early infection time points the protein appeared in the Golgi complex. Later, strong labeling of the plasma membrane emerged. This indicated that Sso2p utilizes the secretory pathway to reach its target membrane. The 20°C temperature block experiments showed this to be the case. In the standard experiments, we were, however, unable to see Sso2p labeling in the endoplasmic reticulum. We considered rapid clearance from the ER as a plausible explanation for this, the alternative being direct insertion to Golgi membranes. The question was approached by two experimental strategies: The protein was expressed in the presence of brefeldin A added at an early time point before any expression of Sso2p could be detected in the cells. On the other hand, a synchronous burst of expression was generated by accumulating mRNA with a cycloheximide block, followed by drug wash-out. In both situations, the newly synthesized Sso2p associated with the endoplasmic reticulum and could subsequently be chased to the later compartments of the secretory pathway. These data show that Sso2p is transported to the PM along the secretory pathway and suggest that it is inserted to the cellular membrane system in the ER. This is supported by the fact that the in vitro translated Sso2p is inserted to dog pancreatic microsomes highly enriched in rough ER membranes. Additionally, caffeine treatment combined with the 20°C temperature block (Jäntti and Kuismanen, 1993) inhibited the accumulation of Sso2p in the Golgi (data not shown), further supporting the above conclusion. We therefore suggest that Sso2p is transported to the PM via the secretory pathway, instead of direct insertion to its target membrane. SNARE-type proteins in transit to their sites of action might be expected to cause problems in the maintenance of the cellular membrane organization, unless their activity is strictly controlled. How the function of SNAREs is regulated is an intriguing subject of future studies. This is where other compartment-specific regulatory proteins, e.g. small GTPases of the rab subfamily (Novick and Brennwald, 1993; Zerial and Stenmark, 1993), may play an important role.

For the tail-anchored proteins studied in detail, the C-terminal hydrophobic sequence has been shown to be responsible for the membrane anchoring (see Kutay et al., 1993 and references therein). To establish the role of the hydrophobic C-terminal segment of Sso2p, we generated a truncated variant of the protein (Sso2p-ma), whose association with membranes was studied in vivo and in vitro. The results presented here confirm that Sso2p is an integral membrane protein and that the hydrophobic amino acid strech is necessary for the membrane association. Interestingly, some of the truncated protein was, by immunofluorescence, observed associated with the ruffling edge of the BHK cells (data not shown). This may indicate that the Sso2p cytoplasmic domain alone is capable of interaction with some components of the mammalian cell PM or structures beneath. A similar observation was made by Bennett et al. (1993) who reported that a truncated isoform of syntaxin 2 lacking most of the C-terminal hydrophobic stretch still showed PM association.

The membrane insertion mechanism of tail-anchored proteins has largely remained obscure. The C-terminal hydrophobic segments of many of these proteins are unlikely to be capable of association with the SRP (Rapoport, 1992; Kutay et al., 1993). Thus, the proteins presumably follow another, perhaps post-translational, route of membrane insertion (Anderson et al., 1983; Kutay et al., 1993). So far, one SNARE-type protein, Aplysia californica synaptobrevin (this protein is not strictly tail-anchored, having a lumenal domain of 75 aa), has been reported to insert to membranes post-translationally (Yamasaki et al., 1994). To assess the translocation mechanism of Sso2p, we performed a series of in vitro translation/microsome insertion experiments using cotranslational and post-translational experimental arrangements. Under co-translational conditions, Sso2p associated with the canine pancreatic microsomes with an efficiency clearly lower than that of the control proteins carrying an N-terminal signal sequence. When microsomes were added to translation mixtures after inhibition of further protein synthesis, translocation of the control proteins was abolished, whereas Sso2p associated with the microsomes as efficiently as in the co-translational situation. This indicates that Sso2p predominantly uses a post-translational ER insertion mechanism. Furthermore, the cytosolic punctate labeling observed in the electron micrographs may reflect translation of the protein on cytosolic ribosomes. Whether the insertion process involves SRP and/or the Sec61 translocation complex in the ER membrane (see Rapoport, 1992; Sanders and Schekman, 1992; Gilmore, 1993), or perhaps some other, yet unidentified protein translocation machinery, remains to be determined.

Proteins following a post-translational membrane insertion mode could potentially be directly targeted to their sites of action in the cell. This is the first report describing passage of a tail-anchored SNARE protein through the secretory pathway to reach its cellular location. It will be interesting to see whether this is the case with other tail-anchored proteins with a post-ER cellular localization. Seija Puomilahti, Riitta Lampinen, Mervi Lindman and Tuula Kaleva are thanked for excellent technical assistance. Dr Ilkka Julkunen is acknowledged for kindly providing the IFN-α1 cDNA, Drs Brian Burke and Graham Warren for the 53FC3 mAb, and Dr Kari Kein änen for the pK212-2 vector. Dr Kai Simons is acknowledged for critical reading of the manuscript, Dr Marino Zerial for fruitful discussions, and Dr Patrick Brennwald for sharing his unpublished data. The electron microscopy was performed at the Department of Electron Microscopy, University of Helsinki. This study was supported by the Academy of Finland and by TEKES.