Targeting of a Short Peptide Derived from the Cytoplasmic Tail of the G1 Membrane Glycoprotein of Uukuniemi Virus (Bunyaviridae) to the Golgi Complex

Abstract

Members of the Bunyaviridae family acquire an envelope by budding through the lipid bilayer of the Golgi complex. The budding compartment is thought to be determined by the accumulation of the two heterodimeric membrane glycoproteins G1 and G2 in the Golgi. We recently mapped the retention signal for Golgi localization in one Bunyaviridae member (Uukuniemi virus) to the cytoplasmic tail of G1. We now show that a myc-tagged 81-residue G1 tail peptide expressed in BHK21 cells is efficiently targeted to the Golgi complex and retained there during a 3-h chase. Green-fluorescence protein tagged at either end with this peptide or with a C-terminally truncated 60-residue G1 tail peptide was also efficiently targeted to the Golgi. The 81-residue peptide colocalized with mannosidase II (a medial Golgi marker) and partially with p58 (an intermediate compartment marker) and TGN38 (a trans-Golgi marker). In addition, the 81-residue tail peptide induced the formation of brefeldin A-resistant vacuoles that did not costain with markers for other membrane compartments. Removal of the first 10 N-terminal residues had no effect on the Golgi localization but abolished the vacuolar staining. The shortest peptide still able to become targeted to the Golgi encompassed residues 10 to 40. Subcellular fractionation showed that the 81-residue tail peptide was associated with microsomal membranes. Removal of the two palmitylation sites from the tail peptide did not affect Golgi localization and had only a minor effect on the association with microsomal membranes. Taken together, the results provide strong evidence that Golgi retention of the heterodimeric G1-G2 spike protein complex of Uukuniemi virus is mediated by a short region in the cytoplasmic tail of the G1 glycoprotein.

Compartmentalization of the cellular membrane and soluble proteins in eukaryotic cells is thought to be governed by structural determinants (topogenic signals) (5). To date, several sequence motifs, or domains, that are essential for localizing proteins to their proper subcellular compartments have been identified. These include sorting, recycling, and true retention signals (35, 41, 44). Such signals are both necessary and sufficient for directing proteins to their correct localization, since their removal results in mislocalization, and their transfer to other proteins will direct the reporter protein to the correct compartment. The identification of topogenic signals in proteins is important for understanding how compartmentalization of a eukaryotic cell is achieved.

Enveloped viruses acquire their lipoprotein coat by budding through a cellular membrane into which virus-encoded membrane (spike) glycoproteins have been inserted. For most viruses, budding occurs at the plasma membrane. In these cases, the spike proteins are transported along the exocytic pathway from the endoplasmic reticulum (ER) to the plasma membrane. The members of the other category of enveloped viruses bud at internal membranes, including the ER (rotaviruses and flaviviruses), intermediate compartment (ERGIC) (coronaviruses and poxviruses), Golgi complex (coronaviruses, rubellaviruses, and the Bunyaviridae), and inner nuclear membrane (herpesviruses) (17, 46). Selection of the budding site is thought to be determined largely by accumulation of the spike proteins in the budding compartment. Viral spikes are often composed of more than one protein subunit forming either heterodimers or homo-oligomers, which are assembled in the ER and then transported to the budding compartment (11). The accumulation of a spike protein complex seems to be determined by a compartment-specific retention signal residing in one of the protein subunits. To date, these signals have been poorly defined.

We are analyzing the mechanisms underlying the budding in the Golgi complex of members of the Bunyaviridae family of viruses (12, 47) by using Uukuniemi virus, a phlebovirus, as a model. The spikes of this virus are composed of two type I membrane glycoproteins, G1 (M_r 70,000; 479 residues) and G2 (M_r 65,000; 495 residues), that are cotranslationally cleaved from an M_r-110,000 precursor (p110) in the ER (1, 23, 50, 57). Processing is likely to be carried out by the luminal signal peptidase, which cleaves downstream of the internal signal sequence mediating the translocation of G2. This leaves the G2 signal sequence covalently attached to the C terminus of G1 (1). Both G1 and G2 have four sites for N-linked glycosylation and 26 cysteine residues in their ectodomain. G1 and G2 fold in the ER with quite different kinetics before forming heterodimers (45). Following proper folding and heterodimerization, the G1-G2 complex is transported to the Golgi complex where further transport is arrested (2, 15, 24, 37, 45).

G1 and G2 coexpressed from the same or different cDNAs accumulate and colocalize in the Golgi complex. G2 expressed in the absence of G1 is retained in the ER, while G1 expressed alone is competent to exit the ER, albeit inefficiently, and to become targeted to the Golgi (37, 49). Expression of mutant forms of G1, as well as chimeric proteins, led to the mapping of a Golgi retention signal to the cytoplasmic tail of G1 (2). A region encompassing the membrane-proximal half of the 98-amino-acid cytoplasmic tail was found to be both sufficient and necessary for targeting the reporter molecules CD4 (a plasma membrane protein) and a membrane-anchored form of lysozyme (a secretory protein) to the Golgi complex. Neither the putative transmembrane domain (TMD) nor palmitylation of the G1 tail seemed to contribute to Golgi retention. G2 is assumed to become targeted to the Golgi via its association with G1.

The finding that a Golgi localization signal could be mapped to the cytoplasmic tail of G1 was surprising in the light of the conclusions drawn from similar mapping studies of other Golgi-retained viral (18, 32, 58) and cellular proteins, notably the glycosyltransferases (41). In most cases, the crucial region responsible for Golgi localization has been mapped to the TMD, although a contributing role of either some flanking sequences or the cytoplasmic tail has been suggested (39, 42, 43). To further elucidate the role of the cytoplasmic tail of G1 in directing G1-G2 to the Golgi, we have expressed the 81-residue tail located between the G1 TMD and the G2 signal sequence, as well as a range of deletion variants as “soluble” peptides. The minimum peptide still able to become localized to the Golgi was found to be 30 residues long. Furthermore, the 81-residue peptide and a 60-residue truncated version could efficiently direct the green-fluorescence protein (GFP), a soluble cytosolic protein, to the Golgi complex. Thus, we describe here the identification of a peptide sequence that can target both a plasma membrane and a cytosolic protein to the Golgi complex.

MATERIALS AND METHODS

Chemicals.

Enzymes used in the cloning procedures were purchased from Amersham, Boehringer Mannheim, New England Biolabs, or Promega. Lipofectin, cell culture media, fetal bovine serum, HEPES, l-glutamine, penicillin, streptomycin, and tryptose phosphate broth were obtained from Life Technologies, Gibco-BRL; [³⁵S]methionine, [9,10(n)-³H]palmitic acid, and [³⁵S]pro-mix were from Amersham; brefeldin A (BFA), cycloheximide (CHX), and Triton X-100 were from Sigma; Pansorbin was from Calbiochem; protein A- and G-Sepharose were from Pharmacia Biotech; Trasylol (aprotinin) was from Bayer; and the cloning vectors pEGFP-C1 and pEGFP-N1 were from Clontech. Oligonucleotides were synthesized with an Applied Biosystems model 392 synthesizer (Perkin-Elmer); the Sequenase kit, version 2.0, was from United States Biochemicals. En³Hance was from New England Nuclear, Du Pont.

Cells.

HeLa cells were grown on plastic dishes or coverslips in minimal essential medium supplemented with 5% fetal bovine serum, 2 mM l-glutamine, 100 IU of penicillin/ml, and 100 μg of streptomycin/ml. BHK21 cells were grown in the same medium, additionally supplemented with 5% tryptose phosphate broth. Normal rat kidney (NRK) cells were grown in Dulbecco’s minimal essential medium supplemented with 10% fetal bovine serum, 2 mM l-glutamine, 100 IU of penicillin/ml, and 100 μg of streptomycin/ml.

Antisera.

A monoclonal antibody (6G9) (45) was used to detect the Uukuniemi virus G1 protein. The monoclonal antibody (9E10) (13) directed against a 10-amino-acid c-myc peptide sequence was used to detect myc-tagged constructs. CD4 was detected with a monoclonal antibody purchased from Boehringer Mannheim. Antibodies against organelle-specific markers were as follows. For the Golgi complex, we used monoclonal antibody CTR433 (provided by M. Bornens) (22) or a polyclonal rabbit antiserum against mannosidase II (provided by K. Moremen and M. Farquhar) (38). For the intermediate compartment/ERGIC, we used a polyclonal rabbit antiserum against a peptide sequence from the luminal domain of p58 (provided by U. Lahtinen) (27); for the trans-Golgi network, we used a polyclonal rabbit antiserum against TGN 38/41 (provided by K. Howell) (21); for endosomes, we used a polyclonal rabbit antiserum against the transferrin receptor (provided by T. Ebel); for lysosomes, we used a polyclonal rabbit antiserum against lamp-1 (provided by S. Carlsson) (9); for the ER, a polyclonal rabbit antiserum was produced in our laboratory against an 18-residue peptide sequence (EEDEILNRSPRNRKPRRE) corresponding to residues 555 to 573 at the C terminus of calnexin (10). A polyclonal rabbit antiserum against the nonstructural protein nsP3 of the Semliki Forest virus (SFV) (provided by L. Kaariainen) (26) was used to identify SFV-induced type I cytopathic vacuoles.

Construction of recombinant cDNA.

Construction of the cDNAs encoding G1 (37) and CD4-C81 (2) has previously been reported. Chimeric proteins were constructed by standard PCR technology, and all cDNA regions cloned from PCR products were completely sequenced by the dideoxy-chain termination method. The primers used and the details of the PCR and cloning strategies are available from the authors on request.

Expression of cDNA constructs.

The cDNA constructs were expressed by using the SFV system (28). All cDNAs were cloned into pSFV1 (provided by P. Liljeström). Linearized plasmids were transcribed in vitro by using SP6 RNA polymerase, and then the capped mRNAs were electroporated into trypsinized and phosphate-buffered saline (PBS)-washed BHK21 or NRK cells. HeLa cells were transfected with mRNA by using Lipofectin as recommended by the manufacturer. Transfected cells were diluted in cell culture medium, seeded onto coverslips or plastic dishes, and incubated at 37°C.

Metabolic labeling and immunoprecipitation.

Electroporated and seeded BHK21 cells were incubated for 5.5 h, starved for 45 min in methionine-free or both methionine- and cysteine-free minimal essential medium, and labeled for 20 min with 0.1 mCi of [³⁵S]methionine per ml or with 0.14 mCi of [³⁵S]pro-mix (containing [³⁵S]methionine and [³⁵S]cysteine) per ml. For labeling with palmitate, electroporated BHK21 cells were incubated at 37°C for 3 h and then labeled for 5 h with 0.5 mCi of [9,10(n)-³H]palmitic acid per ml. The metabolically labeled cells were solubilized with 1% Triton X-100 buffer containing 0.4 M NaCl, 50 mM Tris-HCl (pH 8.0), 5 mM EDTA, 0.02% NaN₃, and 100 IU of aprotinin. The cell lysates were centrifuged for 5 min at 16,000 × g, and the supernatants were preabsorbed with nonimmune ascites and 10% Pansorbin for 1 h at 4°C. After centrifugation, the supernatants were incubated for 4 h on ice with a monoclonal antibody against the c-myc epitope or with a polyclonal antiserum against CD4. Protein A- and protein G-Sepharose were added in equal amounts, and the samples were incubated for another 1 h at 4°C. The beads were collected by centrifugation, washed three times with 0.2% Nonidet P-40–25 mM Tris-HCl (pH 7.5)–0.15 mM NaCl–2 mM EDTA, and finally washed with 25 mM Tris-HCl (pH 7.5). Reducing electrophoresis sample buffer was added, and the samples were boiled for 3 min, cooled, and alkylated with iodoacetamide at a final concentration of 63 mM to prevent re-formation of disulfide bonds after reduction. The samples were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (5 to 15% polyacrylamide gradient or 15% polyacrylamide for linear electrophoresis) as described by Maizel (34) followed by fluorography with En³Hance.

Indirect and confocal immunofluorescence microscopy.

Cells grown on coverslips were transiently transfected as described above. After a 6-h incubation, CHX was added to a final concentration of 0.18 mM to stop further protein synthesis. The cells were incubated for an additional 3 h and, depending on the antibodies used, either fixed with 3% paraformaldehyde and permeabilized with 0.1% Triton X-100 as described previously (2) or fixed and permeabilized with methanol for 2 min at room temperature. Immunofluorescence staining was carried out as described previously (2). Monoclonal and rabbit polyclonal primary antibodies were visualized with tetramethyl rhodamine isothiocyanate (TRITC)-conjugated anti-mouse immunoglobulin G or fluorescein isothiocyanate (FITC)-conjugated anti-rabbit immunoglobulin G secondary antibodies, respectively. To visualize GFP constructs, cells were chased for 4 h in the presence of CHX and the coverslips were mounted with PBS instead of the standard mounting solution. Immunofluorescence micrographs were obtained with an Axiophot fluorescence microscope (Zeiss).

For confocal laser-scanning immunofluorescence microscopy, sample preparation and immunostaining were the same as for indirect immunofluorescence microscopy. The samples were analyzed with a Bio-Rad MRC-600 confocal microscope (Bio-Rad, Cambridge, Mass.), with an ILT model 5470K laser (Ion Laser Technology, Salt Lake City, Utah) as the source for the crypton-argon ion laser beam. FITC-stained samples were imaged by excitation at 488 nm and with a 505- to 540-nm bandpass emission filter, and TRITC-stained samples were imaged by excitation at 568 nm and with a 598- to 621-nm bandpass emission filter.

BFA treatment.

BHK21 cells transfected with the desired mRNAs were incubated for 6 h without CHX and then for 2 h with CHX. They were then treated with 0.018 mM BFA for 1 h in prewarmed BHK medium containing 0.18 mM CHX. The BFA was washed away by incubating the cells for 1 h at 37°C in fresh prewarmed BHK medium containing 0.18 mM CHX. The cells were then fixed and immunostained as described above.

Permeabilization of cells with SLO.

Electroporated and CHX-treated BHK21 cells were washed three times with cold PBS and once with cold streptolysin O (SLO) buffer (25 mM HEPES [pH 7.4], 115 mM potassium acetate, 2.5 mM MgCl₂). The cells were then incubated for 5 min on ice with 0.25 μg of SLO (provided by S. Bhakdi [4]) per ml in SLO buffer containing 1 mM dithiothreitol and washed four times with cold SLO buffer. Prewarmed SLO buffer was added, and the cells were incubated for 30 min at 37°C, chilled on ice, and washed three times with cold SLO buffer and twice with cold PBS. The cells were fixed and immunostained as described above, except that the permeabilization step with 0.1% Triton X-100 was omitted.

Membrane fractionation.

Electroporated and ³⁵S-labeled BHK21 cells were incubated for 6.5 h with or without a following 1-h chase period. The cells were then washed twice with ice-cold PBS, twice with 250 mM sucrose, and once with 50 mM sucrose and finally scraped off the dish in 50 mM sucrose and homogenized with 15 strokes in a tight-fitting Dounce homogenizer. The homogenate was adjusted to a final concentration of 280 mM sucrose, subjected to five additional strokes, and centrifuged in a microcentrifuge at 380 × g for 15 min; the postnuclear supernatant was then collected. The postnuclear supernatant was centrifuged in a Beckman TL-100 ultracentrifuge at 100,000 × g for 1 h. The pellets and the supernatants were subjected to immunoprecipitation and analysis by SDS-PAGE as described above. All steps were performed at 4°C with prechilled solutions and equipment.

RESULTS

The cytoplasmic tail of G1 is necessary and sufficient for targeting CD4 to the Golgi complex.

The overall structure of the Uukuniemi virus membrane protein G1 is shown in Fig. 1A. Full-length G1 expressed in BHK21 cells by using the SFV system colocalized with mannosidase II (Fig. 2a and b), a medial-Golgi marker. No surface expression was observed after a 3-h chase in the presence of CHX (Fig. 2c). When the G1 cytoplasmic tail, lacking the 17-residue internal signal sequence of the downstream G2, was fused to the ectodomain and TMD of CD4 in place of its own tail (Fig. 1A, CD4-C81), the chimeric protein likewise colocalized with mannosidase II (Fig. 2d and e). Again, no surface immunofluorescence was evident (Fig. 2f). We have designated the amino acid immediately downstream of the proposed TMD of G1 (1, 50) residue 1 and the residue just upstream of the G2 signal sequence residue 81. From the above results and those described previously (2), we thus conclude that residues 1 to 81 of the G1 tail are both sufficient and necessary for targeting a plasma membrane protein to the Golgi complex. In the experiments described below, we therefore focused our attention on the properties of the tail peptide from residues 1 to 81.

FIG. 1.

Schematic representation of the different cDNA constructs used. (A) CD4-C81 is a chimeric protein in which the cytoplasmic tail of CD4 has been replaced by the cytoplasmic tail of G1 lacking the 17-residue internal signal sequence (ss) of G2. G1-tail and G1-tail-myc represent the cytoplasmic tail of G1 expressed as an 81-residue peptide, with or without a 10-residue c-myc tag. The G1 tail-myc was progressively deleted from the N terminus by 10 residues at a time, whereas the peptide from residues 10 to 81 was progressively deleted from the C terminus likewise by 10 residues at a time. (B) Residues 1 to 81 of the G1 tail were fused to either the C or N terminus of GFP, while residues 1 to 60 were fused only to the N terminus. When they were fused to the C terminus, a c-myc epitope tag was added to the C-terminal end. (C) Amino acid sequence of the 81-residue cytoplasmic tail of G1. The two cysteines at positions 25 and 28, known to become palmitylated in the intact G1 protein, are in bold and underlined.

FIG. 2.

Colocalization of G1 and CD4-C81 with mannosidase II by immunofluorescence microscopy. G1 and CD4-C81 were expressed in BHK21 cells by using the SFV system. At 6 h posttransfection, the cells were treated for 3 h with CHX and then either permeabilized with Triton X-100 (a, b, d, and e), or left untreated for detection of surface staining (c and f). The cells were indirectly stained with a monoclonal antibody against G1 (a and c) or a monoclonal antibody against CD4 (d and f). Panels b and e show the cells in panels a and d double stained with a polyclonal antiserum against the Golgi marker protein mannosidase II (man II).

The cytoplasmic tail of G1 is able to direct GFP to the Golgi complex.

To analyze whether the G1 tail could direct a soluble cytoplasmic protein to the Golgi, we fused residues 1 to 81 to the N- or C-terminal end of GFP. In the latter case, a c-myc epitope tag was added to the C-terminal end of the G1 tail (Fig. 1B). When wild-type GFP was expressed by using the SFV system, it localized diffusely throughout the cytoplasm and also entered the nucleus (Fig. 3a). As shown in Fig. 3b to e, both GFP-G1 tail chimeras showed a strong accumulation in a juxtanuclear region costaining with the Golgi marker CTR433 (21). A chimera containing only residues 1 to 60 of the tail fused to the N terminus of GFP likewise efficiently localized to the Golgi (Fig. 3f and g). A weak punctate staining dispersed throughout the cytoplasm was observed for all three chimeras (Fig. 3b, d, and f). These studies clearly showed that the G1 tail is able to localize also a soluble cytosolic protein to the Golgi.

FIG. 3.

Intracellular localization of GFP and GFP-G1 tail chimeras by immunofluorescence microscopy. GFP and the three GFP-G1 tail chimeras shown in Fig. 1B were expressed in BHK21 cells by using the SFV system. At 6 h posttransfection, the cells were treated for 4 h with CHX and permeabilized with Triton X-100. GFP was visualized by virtue of its autofluorescence (a, b, d, and f). The same cells were indirectly stained with a monoclonal antibody against the Golgi marker CTR-433 (c, e, and g).

Targeting of the G1 tail peptide to the Golgi complex and cytoplasmic vacuoles.

We next analyzed the fate of the G1 tail encompassing residues 1 to 81 without or with a C-terminal c-myc epitope tag (Fig. 1A, G1-tail or G1-tail-myc, respectively) expressed with the SFV system. As shown in the confocal images in Fig. 4A, D, and G, the G1 tail-myc localized to a juxtanuclear region, as well as vacuolar structures surrounding primarily the nucleus, but was also present more peripherally. The juxtanuclear staining colocalized with mannosidase II (Fig. 4B and C), and partially also with p58, a marker for the intermediate compartment/cis-Golgi (51). In contrast, the vacuoles were negative for both mannosidase II and p58. Since the nonstructural proteins of SFV induce the formation of virus-specific cytopathic vacuoles type I (CPVI) (14), we also double stained the cells with an antiserum against nsP3, one of the four nonstructural proteins. The G1 tail-positive vacuoles did not costain with the anti-nsP3 antiserum (Fig. 4G to I). The presence of the G1 tail-specific vacuoles is also evident in images presented in Fig. 5, 6, and 8. In summary, we conclude that the G1 tail is targeted to the Golgi complex and, in addition, to vacuoles which are distinct from those induced by the SFV expression vector.

FIG. 4.

Colocalization of G1 tail-myc with marker proteins by confocal laser-scanning immunofluorescence microscopy. The myc-tagged G1 tail peptide was expressed in BHK21 cells by using the SFV system. (A, B, D, E, G, and H) At 6 h posttransfection, the cells were treated for 3 h with CHX before being double stained with a monoclonal antibody against the myc tag (A, D, and G) or with a polyclonal antiserum against the Golgi marker mannosidase II (man II), (B) against the ERGIC marker p58 (E), or against the SFV nonstructural protein nsP3, known to induce and associate with type I cytopathic vacuoles. (C, F, and I) Merged images of the double-stained cell.

FIG. 5.

Colocalization of G1 tail-myc with marker proteins by immunofluorescence microscopy. The myc-tagged G1 tail peptide was expressed in BHK21 (A and B), NRK (C and D), or HeLa (E to H) cells by using the SFV system. At 6 h posttransfection, the cells were treated for 3 h with CHX before being double stained with a monoclonal antibody against the myc-tag (A, C, E, and G) or with a polyclonal antiserum against the ER marker calnexin (CN) (B), against the TGN marker TGN38 (D), or against the endosomal marker transferrin receptor (TfR) (F) or the lysosomal marker lamp-1 (H).

FIG. 6.

Effect of BFA on the distribution of the G1 tail-myc peptide. G1 tail-myc was expressed in BHK21 cells by using the SFV system. At 6 h posttransfection, the cells were treated for 2 h with CHX (A and B) before being subjected to BFA treatment for 1 h in the presence of CHX (C and D). The BFA was washed out for 1 h in the presence of CHX (E and F). Permeabilized cells were double stained with a monoclonal antibody (Mab) against the myc tag (A, C, and E) and a polyclonal antiserum against mannosidase II (man II Pab) (B, D, and F).

FIG. 8.

Intracellular localization of G1 tail-myc and mutated and truncated G1 tail peptides by immunofluorescence microscopy. In panel C (C25,28A), the two cysteines at position 25 and 28 were mutated to alanines to prevent palmitylation. The G1 tail constructs were expressed in BHK21 cells by using the SFV system. At 6 h posttransfection, the cells were treated for 3 h with CHX before being double stained with a monoclonal antibody against the myc tag (A, C, E, G, I, and K) or a polyclonal antiserum against mannosidase II (man II) (B, D, F, H, J, and L).

The G1 tail vacuoles do not colocalize with markers for cellular membrane compartments and are resistant to BFA treatment.

In further attempts to determine the identity of the G1 tail-positive vacuoles, we double stained cells with a series of markers for different cellular membrane compartments. As shown in Fig. 5, the vacuoles did not colocalize with calnexin (a marker for the ER) (Fig. 5A and B), TGN38 (trans-Golgi network) (Fig. 5C and D), transferrin receptor (endosomes) (Fig. 5E and F), or lamp-1 (lysosomes) (Fig. 5G and H). Due to the specificity of the marker antisera used, we had to use NRK cells expressing the G1 tail-myc construct for double staining with anti-TGN38 antiserum (Fig. 5C and D) and HeLa cells for staining with anti-transferrin receptor and anti-lamp-1 antisera (Fig. 5E to H). The vesicular structures were smaller and not as prominent in HeLa cells (Fig. 5E and G) as compared in BHK21 and NRK cells (Fig. 5A and C). As evident from Fig. 5C and D, G1 tail-myc localization also largely overlapped with TGN38 in the juxtanuclear region.

BFA is known to relocate Golgi complex proteins to the ER (29). To study the effect of BFA on the distribution of G1 tail-myc, cells expressing the tail peptide were first chased for 2 h with CHX and then treated with BFA for 1 h in the presence of CHX. As shown in Fig. 6D, mannosidase II was completely relocated to a reticular ER-like staining pattern in BFA-treated cells. The Golgi-localized portion of G1 tail-myc was likewise dispersed to a diffuse reticular staining, including the nuclear envelope. In contrast, the staining of the G1 tail-positive vacuoles was not affected by BFA (Fig. 6C). Following a 1-h washout period, the mannosidase II and G1 tail staining patterns returned to those observed in untreated control cells (compare Fig. 5E and F to Fig. 5A and B).

Thus, we conclude that a portion of the expressed G1 tail-myc localizes to vacuoles that are resistant to BFA treatment and whose identity at present remains unclear. The possible nature of these vacuoles is discussed below.

A 30-residue tail peptide can still be targeted to the Golgi complex.

With the aim of defining the Golgi-targeting signal in the G1 tail in greater detail, we analyzed the intracellular localization of truncated tail peptides. As schematically depicted in Fig. 1A, we progressively deleted 10 residues from both the N- and C-terminal ends. All peptides were C-terminally tagged with the myc epitope. For clarity, the amino acid sequence of the tail from residues 1 to 81 is depicted in Fig. 1C. Mutant peptides were expressed with the SFV system in BHK21 cells and metabolically labeled, and the immunoprecipitated peptides were analyzed by SDS-PAGE (Fig. 7). With the exception of the peptide from residues 30 to 81, which migrated as single bands (lane 4), the other truncated peptides were recovered as three bands. On longer exposure, the peptide from residues 20 to 81 also migrated as a triplet (not visible in Fig. 7, lane 3). The reason for this heterogeneity is unclear, although it is likely that the bands represent peptides with no, one, or two palmitic acid chains. As shown previously, there are two cysteines (residues 25 and 28 [Fig. 1C]) used as sites for palmitylation (1, 2). As shown in lane 8, the peptide from residues 1 to 81 could be readily labeled with [³H]palmitic acid. The palmitylation sites have been deleted in the peptide from residues 30 to 81 but not in the others. The variable intensity of the bands may reflect the efficiency by which the peptides serve as targets for palmitylation. We have recently shown that mutating the palmitylation sites in the CD4-C81 chimera has no effect on Golgi targeting (2). Similarly, mutating the two cysteine residues in the peptide from residues 1 to 81 to alanines has no effect on the ability of the peptide to be targeted to the Golgi complex or to become associated with the vacuoles (Fig. 8C and D).

FIG. 7.

Analysis of the G1 tail-myc and the N- and C-terminally truncated G1 tail peptides by SDS-PAGE. The peptide constructs described in Fig. 1A were expressed in BHK21 cells. At 5.5 h posttransfection, the cells were starved in methionine-free medium for 45 min and then labeled with 0.14 mCi of [³⁵S]pro-mix per ml for 20 min (lanes 1 to 7). At 3 h posttransfection, the G1 tail-myc was labeled in parallel for 5 h with 0.5 mCi of [9,10(n)-³H]palmitic acid per ml (lane 8). The cell lysates were immunoprecipitated with a monoclonal antibody against the myc tag and analyzed by SDS-PAGE (15% polyacrylamide) followed by fluorography. The positions of molecular weight (mw) markers are shown to the left in thousands.

The peptide from residues 10 to 81 was efficiently targeted to the Golgi, but interestingly enough, the typical vacuoles seen in cells expressing this peptide were absent (Fig. 8E and F). This result has been obtained consistently. Thus, it appears that association with, or induction of, the vacuoles is a function of the first 10 residues of the G1 tail. Small amounts of the peptide from residues 20 to 81 (Fig. 7, lane 3) were reproducibly recovered by immunoprecipitation. Whether this was due to an inefficient expression or rapid degradation is unclear. Although the peptides from residues 20 to 81 and 30 to 81 were detected in immunoprecipitates (lanes 3 and 4), they have not been localized by immunofluorescence. The lack of staining of the two peptides could be due to rapid degradation, washing out during preparation, or lack of exposure of the myc epitope.

Since previous results with CD4-tail chimeras indicated that residues 50 to 81 were not important for Golgi targeting (2), we deleted the tail peptide from the C terminus starting from 10 to 60 down to 10 to 40 (Fig. 1A and C). All three peptides were readily targeted to the Golgi (Fig. 8G to L). To a variable extent, all three peptides also revealed a scattered punctate pattern of staining, whose identity was not further investigated. We have also expressed two additional peptides encompassing residues 10 to 35 and 15 to 40. These peptides were partly localized to the Golgi, but they were also found associated with other cellular membranes and scattered in the cytoplasm, suggesting that the Golgi localization signal had been partly abrogated (data not shown).

In conclusion, we have thus far identified the minimal domain of the G1 tail peptide that can still be targeted to the Golgi complex to a 30-residue region spanning residues 10 to 40.

The G1 tail peptide is exposed on the cytosolic face of Golgi membranes and is membrane associated.

To analyze whether the G1 tail-myc peptide Golgi was exposed on the cytosolic face of Golgi membranes, we permeabilized cells expressing the peptide either with Triton X-100, which permeabilizes all cellular membranes, or with SLO under conditions where only the plasma membrane is permeabilized. The tail peptide was detected by immunofluorescence with the aid of the myc antibody. To check that internal membranes were not permeabilized by SLO, we used an antiserum directed against the luminal part of p58. In SLO-treated cells, the tail peptide was readily detected in the Golgi region while p58 was not stained, confirming that the Golgi membranes were not permeabilized. In contrast, both p58 and the tail peptide were readily detected in Triton X-100-permeabilized cells (data not shown). Thus, the tail peptide is exposed on the cytoplasmic side of Golgi membranes.

To analyze whether the G1 tail-myc peptide was membrane associated, lysates from metabolically labeled transfected cells were subjected to subcellular fractionation. Almost all of the labeled peptide cosedimented with the membrane fraction (Fig. 9A, lane 5), with very little recovered from the supernatant (lane 6). Essentially the same result was obtained after a 1-h chase period, except that a fainter upper band seen without a chase was more prominent following the chase (lane 7). As discussed above, the two bands are likely to represent species containing different numbers of palmitic acid side chains, since the upper band was absent from cells expressing G1 tail-myc from which the two cysteine residues had been replaced by alanines (Fig. 9B, lanes 3 to 8). The palmitate-negative G1 tail was still membrane associated, although a small fraction was recovered from the supernatant (lanes 5 to 8).

FIG. 9.

Association of G1 tail-myc and G1 tail-myc C25,28A with membranes. The G1 tail-myc (A) and G1 tail-myc C25,28A (B) peptides were expressed in BHK21 cells by using the SFV system. Mock-transfected cells served as controls. At 5.5 h posttransfection, the cells were starved in methionine-free medium for 45 min and labeled with [³⁵S]pro-mix for 20 min. Immediately after the labeling (lanes 1, 2, and 4 through 6) or following a chase for 1 h (lanes 7 and 8), the cells were homogenized and subjected to subcellular fractionation as described in Materials and Methods. The cell lysate (L), postnuclear supernatant (PNS), membranes (P), and supernatant (S) were immunoprecipitated with a monoclonal antibody against the myc tag, and the precipitates were analyzed by SDS-PAGE (15% polyacrylamide) followed by fluorography. Lanes 1 and 2 show mock-transfected cells used as controls. Lane 3 shows the total amount of expressed peptides in the cell lysate.

DISCUSSION

The data presented here show that a short peptide corresponding to a portion of the cytoplasmic tail of a viral membrane glycoprotein is efficiently targeted to the Golgi complex. The peptide was also able to direct the soluble cytoplasmic protein GFP to the Golgi. The fact that the tail can target the plasma membrane protein CD4, as well as a membrane-anchored form of the secretory protein lysozyme, to the Golgi (2) further strengthens the conclusion that the G1 tail of Uukuniemi virus is both necessary and sufficient for conferring Golgi localization. The minimum size of the peptide able to become targeted to the Golgi was found to be 30 residues.

The mechanisms by which proteins are targeted to and maintained in specific subcellular membrane compartments are poorly understood. Two main principles have emerged: true retention and retrieval (recycling) (41, 44, 56). Both mechanisms are thought to involve specific structural motifs, or signals, located in the protein. Well-known examples of retrieval signals include the KDEL motif at the C terminus of luminal ER proteins (44), the dibasic motif at the C or N terminus of ER membrane proteins (56), and the tyrosine-based recycling signal in the cytoplasmic tail of, e.g., TGN38 (48) and furin (6, 52). The last two proteins localize mainly to the trans-Golgi network but recycle between the plasma membrane and the trans-Golgi network via endosomes. All retrieval signals in transmembrane proteins identified to date appear to reside in the cytoplasmic tail.

Very few true retention signals have so far been identified; the best characterized is the TMD of resident Golgi glycosyltransferases, all of which are type II integral membrane proteins (41). According to a hypothesis originally put forward by Bretscher and Munro (7), it is the short length of the TMD of Golgi glycosyltransferases (about 17 residues) compared to that of plasma membrane proteins (about 23 residues) (39–41) rather than the sequence of the hydrophobic amino acids itself that is the basis for Golgi retention. According to this lipid-based retention hypothesis, a short TMD would result in the segregation of resident Golgi proteins into cholesterol-poor domains while plasma membrane proteins would be incorporated into cholesterol-rich domains and transport vesicles (41). According to another model, the so-called kin recognition model, N-acetylglucosaminyltransferase I and mannosidase II are retained in the medial-Golgi by interacting laterally with each other and forming aggregates too large to be incorporated into vesicles destined for the plasma membrane (42). The luminal stalk region of the two glycosyltransferases has been found to play an important role in oligomerization (39, 42, 43).

Membrane proteins of intracellularly maturing enveloped viruses accumulate in the budding compartment (17, 46). The signals responsible for the compartmentalization of such viral proteins have so far been mapped only to specific domains. Thus, the first TMD of the M glycoprotein of the infectious bronchitis coronavirus has been implicated in Golgi retention (33, 58). On the other hand, the cytoplasmic tail of the mouse hepatitis virus, another coronavirus, was found to be important in localizing the M protein to the trans-Golgi and trans-Golgi network (3, 31, 32). For both coronaviruses, the formation of spike protein aggregates and hence the exclusion from transport vesicles was suggested to be important for retention. The TMD of the E2 membrane protein of rubella virus was found to be sufficient to localize the vesicular stomatitis virus G protein to the Golgi (18), the presumed budding compartment. In the case of Punta Toro virus, another member of the Bunyaviridae family that also buds into the Golgi, both the TMD and the cytoplasmic domain of G1 were found to be important for Golgi localization (36). For herpes simplex virus, which buds through the inner nuclear membrane, the TMD of one of the glycoproteins, gB, was sufficient to target vesicular stomatitis virus G to the inner nuclear membrane (16). A similar role for a TMD was demonstrated for the targeting of the lamin B receptor to the inner nuclear membrane (54). It therefore seems that with only a few exceptions, TMDs play an important role in retaining integral membrane proteins in their correct compartments.

Our present and previous (2) results do not support a role for the TMD in retaining G1 of Uukuniemi virus in the Golgi complex. Central to this conclusion is the question of how the TMD of G1 is defined. A 19-residue hydrophobic amino acid sequence flanked by charged residues has previously been proposed to represent the G1 TMD (1, 2, 49). The N-terminal, membrane-proximal part (residues 1 to about 25) of the cytoplasmic tail is rather hydrophobic (Fig. 1C) but contains several basic residues, making it unlikely that it is inserted into the lipid bilayer. A close interaction with Golgi lipids of this portion of the tail, which seems to be a part of the retention signal, cannot, however, be excluded. Preliminary subcellular fractionation showed that the G1 tail was membrane associated (Fig. 9). Experiments carried out with SLO-permeabilized cells indicated that the G1 tail was accessible to antibodies from the cytosolic side of the membranes. It is of interest that palmitylations at cysteine residues 25 and 28 were not required for Golgi retention of either the CD4-C81 chimera (2) or the G1 tail peptide (residues 1 to 81) (Fig. 8).

What, then, could the mechanism be by which the cytoplasmic tail retains G1 in the Golgi complex? As suggested for other Golgi proteins, G1 (and hence the G1-G2 heterodimers) could form aggregates too large to be incorporated into transport vesicles. We have so far not been able to detect such aggregates. Neither have we found indications for oligomerization of the overexpressed G1 tail (unpublished data). One possibility is that the tail contains a signal that would recycle G1 from distal to more proximal Golgi cisternae. Alternatively, G1 could recycle between the plasma membrane and the Golgi similarly to, e.g., TGN38 (48) or furin (6, 52). Finally, the tail could interact with resident integral or peripheral Golgi proteins or with a submembranous Golgi matrix (19, 53). This latter model envisages a more static retention mechanism as opposed to a dynamic model, which may, according to more recent results, operate for glycosyltransferases (8, 29).

Many peripheral proteins have been found to be associated with Golgi membranes. One such example is glutamic acid decarboxylase (GAD65), in which the first N-terminal residues were found to harbor the Golgi-targeting signal. Palmitylation, normally present within this region, was not required for correct targeting (55). Using an approach very similar to the one used by us, Liu et al. (30) could show that the first 35 N-terminal residues of endothelial nitric oxide synthase were sufficient to target GFP to the Golgi, the principal location of endothelial nitric oxide synthase. In this case, myristylation of the N terminus was necessary for correct targeting. It is possible that peripheral proteins, such as the ones mentioned above, and the G1 membrane protein are targeted to the Golgi membrane by similar mechanisms.

In addition to colocalizing with the Golgi marker mannosidase II and partly also with p58 (intermediate compartment/ERGIC marker) and TGN38 (trans-Golgi marker), the 81-residue tail peptide was found associated with vacuoles in the vicinity of the Golgi but also scattered around the cell. These vacuoles did not stain with antibodies against markers for the Golgi, intermediate compartment, endosomes, lysosomes, or the SFV-induced cytopathic vacuoles type I. Since they were also resistant to treatment with BFA, it is unlikely that they represent Golgi membranes. Whether they represent structures present in untransfected cells or whether they are induced by the tail peptide is likewise unclear. The vacuoles were not seen with shorter forms of the tail. Thus, their formation appeared to be dependent on the first 10 membrane-proximal residues. Similar vacuoles have not been seen with the CD4-C81 chimera (Fig. 2) (2), or with G1 expressed in the absence of G2 (Fig. 2) (2, 37). However, during virus infection, the Golgi complex undergoes an extensive vacuolization (24), which seems to be a function of the G1 and G2 glycoproteins accumulating in the Golgi (15). Whether these two types of vacuoles are related to each other is at present not clear.

During infection, the helical ribonucleoproteins, consisting of the three single-stranded genomic RNA segments and the associated nucleoprotein, accumulate in the Golgi and presumably interact with the G1 cytoplasmic tail to trigger the budding of virus particles into the Golgi lumen (20, 24, 25). Thus, the G1 tail has two separate functions, i.e., mediating Golgi retention and serving as the receptor for nucleocapsid interaction and budding. Since we have mapped the Golgi retention signal to residues 10 to 40 of the tail, we postulate that the nucleocapsids may interact with the region downstream of residue 40 (Fig. 1C).