Hepatitis C Virus Glycoprotein Complex Localization in the Endoplasmic Reticulum Involves a Determinant for Retention and Not Retrieval*

The hepatitis C virus (HCV) genome encodes two envelope glycoproteins (E1 and E2). These glycoproteins interact to form a noncovalent heterodimeric complex which in the cell accumulates in endoplasmic reticulum (ER)-like structures. The transmembrane domain of E2, at least, is involved in HCV glycoprotein complex localization in this compartment. In principle, ER localization of a protein can be the consequence of actual retention in this organelle or of retrieval from the Golgi. To determine which of these two mechanisms is responsible for HCV glycoprotein complex accumulation in the ER, the precise localization of these proteins was studied by immunofluorescence, and the processing of their glycans was analyzed. Immunolocalization of HCV glycoproteins after nocodazole treatment suggested an ER retention. In addition, HCV glycoprotein glycans were not modified by Golgi enzymes, indicating that the ER localization of these proteins is not because of their retrieval from the cis Golgi. Retention of HCV glycoprotein complexes in the ER without retrieval suggests that this compartment plays an important role for the acquisition of the envelope of HCV particles. A true retention in the ER was also observed for E2 expressed in the absence of E1 or for a chimeric protein containing the ectodomain of CD4 in fusion with the transmembrane domain of E2. These data indicate that, in HCV glycoprotein complex, the transmembrane domain of E2, at least, is responsible for true retention in the ER, without recycling through the Golgi.

Resident proteins in cellular organelles contain some information in their primary structure for determining their subcellular localization. Keeping these proteins in a particular compartment can be achieved either by a strict retention in this compartment or by retrieval. Although most integral membrane proteins of the Golgi seem to be genuinely retained (1,2), endoplasmic reticulum (ER) 1 residence largely depends on re-trieval mechanisms (3,4). Many ER, luminal, and type I transmembrane proteins contain carboxyl-terminal sequences of the prototypes KDEL and KKXX, respectively (5,6). These sequences act as retrieval signals, returning proteins that have left the compartment in which they reside (7).
Hepatitis C virus (HCV) is a positive-strand RNA virus that belongs to the Flaviviridae family (8). Its genome contains a long open reading frame of 9,030 to 9,099 nucleotides that is translated into a single polyprotein of 3,010 to 3,033 amino acids (9). Cleavages of this polyprotein are co-and post-translational and generate at least 10 polypeptides including 2 glycoproteins, E1 and E2 (10). These glycoproteins are heavily modified by N-linked glycosylation and are believed to be type I transmembrane glycoproteins with an amino-terminal ectodomain and a carboxyl-terminal hydrophobic anchor. For E2, carboxyl-terminal deletions removing its hydrophobic region result in secretion of the ectodomain. This is in accordance with other data proposing that the hydrophobic anchor domain begins at amino acid 718 (position on the polyprotein) (11,12). The situation appears to be more confusing for E1, because a truncated form ending at amino acid 340 is secreted only if it contains an internal deletion between amino acids 262 and 290, suggesting that a second membrane anchor might exist (13). However, truncated forms ending at amino acid 311 or 334 and containing this internal sequence can also be secreted (11,14). Because E1 and its truncated forms do not fold properly in the absence of E2 (11), the lack of secretion of some truncated forms of E1 can be because of retention in the ER by interaction with chaperones. E1, like E2, is probably anchored by its carboxyl-terminal hydrophobic sequence, and the amino-terminal limit of the potential transmembrane domain of E1 remains to be established. Studies using transient expression systems have shown that E2 interacts with E1 to form a noncovalent heterodimer (15). However, the efficiency of HCV glycoprotein assembly is low, and a large portion of them form heterogenous disulfide-linked aggregates (16,17).
Prolonged interactions of HCV glycoprotein complexes with ER chaperones has been observed (18). It has initially been suggested that association of HCV oligomers with calnexin could be responsible for their continued retention in an intracellular compartment (17). However, more recent data indicate that E1E2 complexes in their native form do not interact any longer with calnexin and are still retained in a pre-medial Golgi compartment (15). This suggests that HCV glycoprotein com-plexes contain a retention signal for localization in an intracellular compartment. In addition, we have shown that a signal for ER localization of E2 maps to its transmembrane domain (19). The purpose of this work was to identify more precisely the subcellular localization of HCV glycoprotein complex and to study the mechanism involved. Immunolocalization of HCV glycoproteins suggested an ER retention. In addition, HCV glycoprotein glycans were not modified by Golgi enzymes, indicating that the ER localization of these proteins is not because of their retrieval from the cis Golgi.
Cell Culture-The HepG2 and CV-1 cell lines were obtained from the American Type Culture Collection (ATCC), Manassas, VA. The UHCV-11.4 cell line derives from the U-2 OS human osteosarcoma cell line (ATCC HTB-96) and has been described (23). This cell line contains a tetracycline-regulated system that allows tightly controlled expression of the full-length HCV polyprotein. Cell monolayers were grown in Dulbecco's modified essential medium supplemented with 10% fetal bovine serum. The UHCV-11.4 cell line was also supplemented with 500 g/ml G418, 1 g/ml puromycin, and 1 g/ml tetracycline. To allow expression of HCV polyprotein, UHCV-11.4 cells were rinsed twice with Dulbecco's modified essential medium and incubated for an additional 24 h in the same medium deprived of tetracycline as described (23).
Growth of Viruses-Stocks of vaccinia virus recombinants expressing HCV proteins were grown and titrated on CV-1 monolayers. Vaccinia virus recombinants vTF7-3 (a vaccinia virus recombinant expressing the T7 DNA-dependent RNA polymerase) (24), vE1E2p7 (a vaccinia virus recombinant expressing HCV glycoproteins E1 and E2, and the p7 polypeptide), vE2p7 (a vaccinia virus recombinant expressing HCV glycoprotein E2 and the p7 polypeptide) (25), and vCD4-E2 718 (a vaccinia virus recombinant expressing the ectodomain of CD4 fused to the transmembrane domain of E2) (19) were used in this work.
Immunofluorescence Microscopy-Subconfluent HepG2 or UHCV-11.4 cells were grown on coverslips. For HCV glycoprotein expression in HepG2 cells, subconfluent monolayers were coinfected by vTF7-3 and vaccinia virus recombinants expressing HCV glycoproteins. For HCV protein expression in UHCV-11.4 cells, subconfluent monolayers were rinsed twice with medium deprived of tetracycline and incubated in the same medium. At 6 h post-infection (HepG2 cells) or after 24 h without tetracycline (UHCV-11.4 cells), cells were fixed for 10 min with paraformaldehyde (4% in phosphate-buffered saline). Cells were permeabilized or not for 30 min at room temperature with TBS (50 mM Tris-Cl (pH 7.5), 150 mM NaCl) containing 0.1% Triton X-100. Immunofluorescence was carried out as described (15).
For in vivo labeling of glycan moieties, HepG2 cells were infected with the appropriate vaccinia virus recombinants and pulse-labeled for 30 min with 3.7 ϫ 10 6 Bq/ml [2-3 H]mannose in ␣-minimum essential medium containing 0.5 mM glucose and 10% dialyzed fetal bovine serum. After 4 h of chase, cells were lysed in TBS, 0.5% Igepal CA-630, and the lysates were used for immunoprecipitation.
Analysis of Oligosaccharide Material-Immunoprecipitated [2-3 H] mannose-labeled proteins were digested overnight at room temperature with 0.2 mg of L-1-tosylamido-2-phenylethyl chloromethyl ketonetreated trypsin in 0.1 M ammonium bicarbonate (pH 7.9). Trypsintreated proteins were boiled for 10 min to inactivate the trypsin, and the peptides were dried and dissolved in 20 mM sodium phosphate (pH 7.5) containing 50 mM EDTA and 0.2 mg/ml NaN 3 , in 50% glycerol. The peptides were incubated overnight at 37°C in the presence of PNGase F (0.5 units). For some samples, endo H (10 milliunits) digestion was performed in 50 mM sodium citrate buffer (pH 5.5). Size analysis of the glycan moieties was achieved by HPLC on an amino-derivatized column ASAHIPAK NH 2 P-50 (250 ϫ 4.6 mm) (Asahi, Kawasaki-ku, Japan) with a solvent system of acetonitrile/water from 70:30 (v/v) to 50:50 (v/v) at a flow rate of 1 ml/min over 80 min. Oligomannosides were identified as described previously (28) by their retention time. Separation of labeled oligosaccharides was monitored by continuous flow detection of radioactivity with Flo-one ␤ detector (Packard).
Affinity Chromatography of Labeled N-Glycans-The lectin column (concanavalin A-Sepharose, 5 ϫ 0.5 cm) was equilibrated at room temperature in 5 mM sodium acetate buffer (pH 5.2) containing 0.1 M NaCl, 1 mM MnCl 2 , 1 mM CaCl 2 , and 1 mM MgCl 2 . Glycan fractions (resulting from PNGase F digestion) were applied to the column, which was then eluted with the equilibration buffer (buffer a). Weakly retained glycans were eluted with 10 mM methyl ␣-D-glucoside in the equilibration buffer (buffer b), and strongly retained glycans were eluted with 100 mM ␣-D-mannoside (buffer c).

Immunolocalization of HCV Glycoprotein Complex-Recent
data from our group indicate that native HCV glycoprotein complexes are localized in ER-like structures and are not transported beyond the cis or medial Golgi (15). However these data do not show whether HCV glycoproteins are retained in the ER or transported beyond the ER and cycle between early compartments of the secretory pathway. As a first approach to answer this question, cells expressing HCV glycoproteins were analyzed by immunofluorescence studies. Because of the absence of an efficient system for cell culture replication of HCV, we used a cell line that can express the full-length HCV polyprotein in a tightly regulated manner (UHCV-11.4) (23). This also has the advantage of avoiding morphological modifications of some subcellular compartments because of a viral vector such as vaccinia virus. Before identifying the precise subcellular localization of HCV glycoproteins by immunofluorescence, we first analyzed the formation of native E1E2 complexes in this cell line. Using the vaccinia/T7 expression system, we have previously shown that the assembly of E1E2 heterodimers is slow and inefficient (15). When expressed in UHCV-11.4 cells, HCV glycoproteins showed the same tendency to aggregate (data not shown), suggesting that accumulation of misfolded HCV glycoproteins is not because of high levels of expression or to the vaccinia virus vector. This tendency to aggregate is therefore probably an intrinsic property of these glycoproteins. In addition, the formation of native E1E2 complexes in UHCV-11.4 cell line was equally slow, as shown by the kinetics of recognition of HCV glycoproteins with a conformation-sensitive E2reactive mAb (H53), which specifically reacts with the noncovalent E1E2 heterodimer ( Fig. 1) (15,19). mAb H53 was used throughout this work to localize E2 or E1E2 complexes. The use of this mAb allowed us to study the subcellular localization of properly folded E2 or E1E2 complexes without the background because of the presence of a large portion of misfolded proteins that can be retained in the ER by interacting with chaperones.
To identify the HCV glycoprotein-containing organelle(s), we employed double-label immunofluorescence microscopy using antibodies to known ER, intermediate compartment, and Golgi antigens. As previously observed with the vaccinia/T7 expression system, no fluorescence was detected on the surface of nonpermeabilized cells expressing HCV polyprotein when revealed with mAb H53 (data not shown). When the cells were permeabilized, HCV glycoproteins showed mainly an ER-like distribution of fluorescence (Fig. 2), and some spots with a higher intensity of fluorescence were also detected. Treatment with cycloheximide for 4 or 8 h did not induce any change in the distribution of fluorescence (data not shown). The pattern observed after labeling with mAb H53 was similar to that revealed by an antibody against PDI (an ER resident protein) and different from those shown by antibodies directed against mannosidase II (a marker of the Golgi apparatus) or Rab1 (a marker of the ER-to-Golgi intermediate compartment (ERGIC) (Fig. 2). Similar results were observed when HCV glycoproteins were expressed in HepG2 cells by using the vaccinia/T7 expres-sion system (data not shown). It has to be noted that the brighter spots detected with mAb H53 colocalized with PDI and did not colocalize with markers of the endocytic pathway (data not shown). Because intracellular native HCV glycoproteins are rather stable (15), these spots likely correspond to accumulation of HCV glycoproteins in some areas of the ER. It is likely that expression of HCV polyprotein induces proliferation of ER membranes as observed in flavivirus infection (10).
Although no clear colocalization of HCV glycoproteins was observed with markers of the Golgi apparatus or the ERGIC, our data do not exclude transport to the cis Golgi followed by recycling to the ER as observed for luminal proteins containing a carboxyl-terminal sequence of the prototype KDEL (3,29,30). As a first approach to answer this question, cells expressing the HCV polyprotein (UHCV-11.4) were treated with nocodazole. This agent disrupts microtubules, leading to a desintegration of the Golgi and interruption of traffic between the Golgi, the ERGIC, and the ER (31,32). The fate of HCV glycoproteins was compared with that of PDI, an ER resident protein that contains a KDEL signal for retrieval. After treatment for 5 h with 20 M nocodazole, much of PDI concentrated in large spots (Fig.  3). Such concentrated spots were not observed with mAb H53. However, there was a redistribution of the bright spots observed in the absence of treatment. An ERGIC marker, ER-GIC-53 (22), showed the expected perinuclear staining without treatment, which changed to more punctuated staining after treatment with nocodazole. A similar desintegration of structure was observed for the Golgi (Fig. 3).
Analysis of the Processing of HCV Glycoprotein Glycans-Although some differences were observed by immunofluorescence between HCV glycoproteins and PDI after nocodazole treatment, these data did not allow to clearly conclude that HCV glycoproteins are strictly retained in the ER. Another cells were fixed with paraformaldehyde, permeabilized with Triton X-100, and labeled with anti-E2 mAb H53 (secondary donkey anti-mouse IgG-Cy2) and antibodies to PDI, Rab1, or mannosidase II (Man II) (secondary donkey anti-rabbit IgG-Rhodamine). Magnification 625ϫ. approach to tackle this problem was to analyze the modifications that their glycans have potentially acquired in the compartment into which they have transited. It has been previously shown that HCV glycoproteins are endo H-sensitive when analyzed in pulse-chase experiments with chase times of up to 8 h (15). Similar results were observed for HCV glycoproteins expressed in the UHCV-11.4 cell line (Fig. 4). The lack of complex-type glycosylation excludes transit through the medial-but not the cis Golgi. In the cis Golgi, HCV glycoproteins would be exposed to Golgi ␣-mannosidase I, which would process its sugar chains to Man 5 GlcNAc 2 (33). Molecules containing Man 5 GlcNAc 2 should accumulate after several cycles through the cis Golgi and back to the ER, and they should be sensitive to endo D (34). As shown in Fig. 4, HCV glycoproteins remained endo D-resistant after 4 h of chase, suggesting that these molecules do not cycle between the ER and the cis Golgi. However, one last intermediate in sugar chain processing in the Golgi that would be sensitive to endo H and resistant to endo D could be Man 5 GlcNAc 3 , formed by the action of GlcNActransferase I in the medial Golgi.
To better characterize their potential processing, HCV glycoprotein glycans were removed by PNGaseF treatment and analyzed by affinity chromatography and HPLC. For this approach, HCV glycoproteins were labeled with [2-3 H]mannose and immunoprecipitated with mAb H53 before PNGaseF treatment and characterization of labeled glycans. However, because the level of HCV glycoprotein expression in UHCV-11.4 cells was too low for such analysis, we used the vaccinia/T7 system to obtain higher expression levels of HCV glycoproteins. As a preliminary analysis, we first compared the sensitivity of HCV glycoproteins to endo H, endo D, and jack bean ␣-mannosidase to confirm that the processing of their glycans is similar in both expression systems. HCV glycoproteins were sensitive to endo H treatment and resistant to endo D treatment in both systems (Fig. 4). In addition, HCV glycoproteins showed some sensitivity to jack bean ␣-mannosidase treatment in both ex- pression systems, and their shift in electrophoretic mobility was similar using both expression systems. Affinity chromatography analysis of HCV glycoprotein glycans removed by PNGase F treatment showed that 100% of the radioactive glycans bound strongly to concanavalin A and eluted in a buffer containing 100 mM ␣-D-mannoside (buffer c) (Fig. 5A), indicating that HCV glycoprotein oligosaccharide moieties are of the oligomannoside type only. In addition, HPLC analysis of the glycans released by PNGase F treatment demonstrated the presence of three species: Man 9 , Man 8 , and Man 7 GlcNAc 2 , respectively (Fig. 5B). These results were confirmed by the fact that 100% of the radioactivity bound to E1E2 complexes could be released by endo H. As expected, HPLC analysis of the oligosaccharides released by endo H revealed the same species liberated as Man 9 , Man 8 , and Man 7 GlcNAc 1 (Fig. 5C). That the oligosaccharide precursor, which is transferred onto nascent proteins, is the Glc 3 Man 9 GlcNAc 2 reveals the sequential actions of ER glucosidases I and II and at least the action of ER mannosidase-yielding Man 8 species. The presence of Man 7 is probably because of trimming of mannose residues, occurring after prolonged residence in the ER (27). Indeed, both Man9 mannosidase (35) and soluble ER mannosidase (36) have been demonstrated to be able to trim mannosidic linkage down to Man 6 GlcNAc 2 species. The nature of the glycans observed in this work suggested two possibilities: either HCV glycoproteins were retained in the ER and could not reach the Golgi vesicles where additional processing takes place or the N-glycosylation sites could not be modified even in the presence of mannosidases and glycosyltransferases. To test the second hypothesis, a redistribution of Golgi enzymes into the ER was induced by BFA treatment (37). Labeled HCV-glycoprotein oligosaccharides obtained from BFA-treated cells were analyzed by affinity chromatography on a concanavalin A column (Fig. 6A). Twenty-four percent of the radioactivity was eluted in the equilibration buffer (buffer a), demonstrating that some glycans have been processed into complex types. Seventy-six percent of the radioactive oligosaccharides remained of oligomannoside type and were eluted in a buffer containing 100 mM ␣-D-mannoside (buffer c). In addition, HPLC analysis of this fraction revealed the presence of smaller species (Man (7-3) GlcNAc 2 ) (Fig. 6B) when compared with the ones observed in the absence of BFA (Fig. 5B). Together these data indicate that the absence of processing of HCV glycoprotein glycans is not because of their inaccessibility to Golgi mannosidases and glycosyltransferases.
The Transmembrane Domain of E2 Is a Determinant for ER Retention and Not Retrieval-Previous data from our group have shown that E2 glycoprotein expressed in the absence of E1 is retained in the ER or a pre-medial Golgi compartment and that the signal for E2 subcellular localization maps to its transmembrane domain (19). We were therefore interested to determine whether E2 expressed in the absence of E1 would be retained in the ER, like E1E2 complex, without any recycling from the cis Golgi. When analyzed by double immunofluorescence, expression of E2 alone showed a pattern similar to that revealed by an antibody against PDI and different from those shown by antibodies directed against mannosidase II or Rab1 (data not shown). As shown in Fig. 4, E2 expressed in the absence of E1 was endo H-sensitive and endo D-resistant. It also showed some sensitivity to jack bean ␣-mannosidase treatment, similar to what was observed when E1 and E2 were coexpressed. HPLC analysis of E2 glycans revealed the same species (Man (9 -7) GlcNAc 2 ) as observed for E1E2 complexes (Fig. 7A). In addition, the glycans of a chimeric protein containing the ectodomain of CD4 (a protein normally expressed to the cell surface) fused to the transmembrane domain of E2 showed a similar HPLC profile (Fig. 7B). Together these data indicate that the transmembrane domain of E2 is responsible for genuine retention in the ER, without recycling through the Golgi.

DISCUSSION
Immunolocalization of HCV glycoproteins and analysis of their glycans showed that these proteins are strictly retained in the ER. In addition, this retention could be mediated by at least the transmembrane domain of E2. A growing number of ER retention signals have been identified. The best characterized of these motifs are KDEL-COOH, KKXX-COOH and NH 2 -RR, all of which seem to function by retrieval from post-ER compartments (38 -40). For KDEL-bearing proteins, deletion of the retrieval motif leads to their secretion. However, these proteins are secreted at different rates and in general very slowly (41). The KDEL motif, as originally proposed (5), acts then as a salvage mechanism to return proteins that have left the ER. Resident ER proteins would be largely excluded from vesicular export because their high local concentration favors intermolecular interactions or because of the presence of retention signals (42). Recently, it has been proposed that export from the ER may occur through a selective mechanism (43). Such a selective export may involve a di-acidic signal (DXE) on the cytoplasmic tail of transmembrane proteins. HCV glycoproteins do not contain a cytoplasmic tail and therefore do not possess a positive DXE-like export signal. However, in the absence of this type of signal, we would expect to detect some slow release of HCV glycoproteins out of the ER. Strict retention in the ER has recently been described for unassembled IgM intermediates and for the uncleaved precursor of the human asialoglycoprotein receptor H2a subunit (26,44). Retention of these proteins is because of quality control, which contrary to what is usually observed (45,46), does not involve retrieval in these cases. For HCV glycoproteins, it has been shown that ER retention is not because of quality control. Indeed, HCV glycoprotein complexes in their native form do not interact with ER chaperones (15,18). In addition, replacement of the transmembrane domain of E2 by the transmembrane domain and cytosolic tail of CD4 leads to its export to the cell surface (19). ER localization of some cytosolic proteins can be mediated by a hydrophobic sequence at their carboxyl termini (47,48). For these proteins, it has not been shown whether a mechanism for retrieval or strict retention is involved. Similarly, the transmembrane domain of Golgi proteins and part of their flanking regions contain sufficient information for Golgi retention (7). For these proteins, subcellular localization is not because of retrieval from other compartments but to strict retention. Although the mechanism by which Golgi retention occurs is still unclear, it has been suggested that membrane thickness could play a role (49). Such lipid-based mechanism has also been proposed for proteins that are localized in the ER by their transmembrane domain (47), and ER retention of HCV glycoprotein E2 could fit in this model.
A strict ER retention of HCV glycoprotein complexes suggests that budding of HCV particles occurs in the ER. Enveloped viruses acquire their envelope by budding through one of several host cellular membranes. In the case of viruses such as Semliki Forest virus, vesicular stomatitis virus or influenza virus, the viral membrane glycoproteins are synthesized and transported to the plasma membrane in a manner indistinguishable from cellular proteins. It is the accumulation of these proteins at the cell surface that is responsible for viral budding (50). Some other viruses, however, bud at internal membranes, such as those of the ER (e.g. rotaviruses), the ERGIC (coronaviruses), or the Golgi complex (Bunyaviridae) (50,51). Virus particles are then released from the infected cells either after cell lysis (e.g. rotaviruses) or after transport of virus-containing vesicles to the cell surface and fusion of these vesicles with the plasma membrane (e.g. coronaviruses and bunyaviruses). For viruses that bud intracellulary, there needs to be an accumulation of the viral membrane glycoproteins, which form the spikes, in the appropriate compartment (51). A strategy that most of these viruses have developed is to endow the spike proteins with signals for compartment-specific targeting and retention, similarly to normal compartment-specific cellular proteins (52)(53)(54)(55)(56)(57). For the flaviviruses, virions appear in intracellular vesicles (probably modified ER) and are released from cells via the exocytosis pathway (reviewed in Refs. 50 and 51). In the case of HCV, such studies have been hampered by the absence of an efficient cell culture system for its replication. In addition, efficient HCV particle formation has not been ob-served in transient expression assays, suggesting that essential viral or host factors are missing or blocked. The process of viral envelope formation is not well understood. In most cases, the viral nucleocapsid plays an important role (58), and the icosahedral capsid probably acts as a scaffold responsible for the curvature of the envelope. In some cases, viral membrane proteins are secreted in the absence of the nucleocapsid. This is the case for most flaviviruses (10), which, like HCV, belong to the Flaviviridae family. However such virus-like particles are not observed for HCV.
In conclusion, we show that the transmembrane domain of E2 is responsible for genuine retention in the ER without recycling through the Golgi. This indicates that besides the classical retrieval mechanisms described for proteins with a KDEL or a KKXX signal, retention of native proteins without retrieval can also occur in the ER.