Intracellular Transport of the Glycoproteins gE and gI of the Varicella-Zoster Virus

The varicella-zoster virus (VZV) is the etiological agent of two different human pathologies, chickenpox (varicella) and shingles (zoster). This alphaherpesvirus is believed to acquire its lipidic envelope in the trans-Golgi network (TGN). This is consistent with previous data showing that the most abundant VZV envelope glycoprotein gE accumulates at steady-state in this organelle when expressed from cloned cDNA. In the present study, we have investigated the intracellular trafficking of gI, another VZV envelope glycoprotein. In transfected cells, this protein shows a very slow biosynthetic transport to the cell surface where it accumulates. However, upon co-expression of gE, gI experiences a dramatic increase in its exit rate from the endoplasmic reticulum, it accumulates in a sialyltransferase-positive compartment, presumably the TGN, and cycles between this compartment and the cell surface. This differential behavior results from the ability of gE and gI to form a complex in the early stages of the biosynthetic pathway whose intracellular traffic is exclusively determined by the sorting information in the tail of gE. Thus, gI provides the first example of a molecule localized to the TGN by means of its association with another TGN protein. We also show that, during the early stages of VZV infection, both proteins are also found in the TGN of the host cell. This suggests the existence of an intermediate stage during VZV biogenesis in which the envelope glycoproteins, transiently arrested in the TGN, could promote the envelopment of newly synthesized nucleocapsids into this compartment and, therefore, the assembly of infective viruses.

The trans-Golgi network (TGN) 1 is a tubuloreticular compartment located on the trans-most side of the Golgi complex (1). This organelle houses different proteins that are involved in adding post-translational modifications to polypeptides traveling along the secretory pathway. In addition, this organelle constitutes the main sorting station in the secretory pathway (2,3). The TGN can undergo rapid tubularization and mixing with endosomal compartments in the presence of the fungal metabolite brefeldin A (BFA) (4).
The TGN has also been used by certain viruses as a membrane donor for their lipidic envelope (envelopment or budding process), as it happens in the case of the varicella-zoster virus (VZV) (5,6). VZV is a human alphaherpesvirus causing chickenpox (varicella), as a result of the primary infection, and shingles (zoster) upon reactivation of the latent virus (7,8). As it occurs in other alphaherpesviruses, the VZV nucleocapsids are assembled in the nuclei of the infected cells. These nucleocapsids are then released into the periplasmic space by budding through the inner nuclear membrane, thereby acquiring a transient envelope that is lost upon fusion with the outer nuclear membrane. In this way, the nucleocapsids are released in the cytosol, where they acquire a second and definitive envelope. This envelope is derived from the TGN, as initially demonstrated by Gershon et al. (5) by examining VZV-infected cells at the ultrastructural level. Mature viruses accumulate finally in an intracellular endosomal compartment (9).
As happens in other cases of viruses that undergo intracellular assembly, envelopment of VZV in the TGN requires that the corresponding envelope glycoproteins have to be delivered to this compartment during viral infection (10 -12). This implies that sorting signals must exist within these glycoproteins to ensure their correct targeting, making these molecules very useful tools for analyzing the mechanisms involved in TGN localization. We and others have recently shown that the most abundant envelope glycoprotein of VZV (gpI or gE) accumulates in the TGN when expressed from cloned cDNA and that this accumulation results, at least partially, from its ability to be rapidly retrieved form the cell surface (10 -12). The sorting information in the sequence of gE has been mapped to its cytoplasmic tail, and shown to consist of two tyrosine-containing tetrapeptides related to endocytosis motifs (11,12) and a more C-terminal acidic cluster that contains casein-kinase IIphosphorylatable residues (11,12). These signals are similar to those found in other molecules known to be localized in the TGN at steady state, such as TGN38 or furin (13)(14)(15)(16)(17)(18)(19).
In addition to gE, there are at least five additional glycoproteins in the envelope of VZV (gB, gH, gI, gC, and gL, formerly known as gpII, gpIII, gpIV, gpV, and gpVI, respectively) (20), whose sequences are apparently devoid of TGN-sorting information. If VZV indeed acquires its final envelope in the TGN, then mechanisms must exist to ensure that all these molecules reach this compartment in order to promote infective VZV formation. In the present article, we have focused our attention on another type I glycoprotein of the viral envelope, the glycoprotein gI (or gpIV). This molecule has been shown to physically interact with gE in VZV (21), as well as in herpes simplex virus (HSV-1) (22), feline herpesvirus (FHV-1) (23), and pseudorabies virus (PRV) (24), three other members of the alphaherpesvirinae subfamily. Our results indicate that gI, which is found in the cell surface when expressed alone, accumulates in the TGN when expressed together with gE. This accumulation of gI in the TGN also relies on its rapid internalization from the cell surface. Our data indicate that gE and gI precursors can form a stoichiometric complex in the endoplasmic reticulum (ER), which results in an increased maturation rate of gI. We have also found that, in VZV-infected cells, both gE and gI can be found shortly after infection in a perinuclear compartment that most likely corresponds to the TGN.

EXPERIMENTAL PROCEDURES
Materials-Monoclonal antibodies SG1 and SG4, against VZV gE and gI, respectively, were obtained from Viro Research Inc. (Rockford, IL). The polyclonal serum against the cytoplasmic tail of furin was generously provided by Dr. W. Garten (University of Marburg, Marburg, Germany). The SA48 HeLa clone stably expressing VSVG-tagged sialyltransferase (ST-VSVG) was a generous gift of Dr. Tommy Nilsson (EMBL, Heidelberg, Germany). All secondary antibodies against the Fc of mouse or rabbit IgGs coupled to FITC or rhodamine were purchased from Dianova (Hamburg, Germany).
Cloning of the VZV Envelope Glycoproteins gE and gI in Mammalian Expression Vectors-Construction of the gE expression vector has been previously described (12). In order to clone VZV gI, the complete open reading frame was amplified from a lysate of VZV (Dumas strain)infected cells using the Expand High Fidelity kit (Boehringer Mannheim, Mannheim, Germany). The resulting fragment was digested with XbaI and HindIII and was cloned into the same sites of the eukaryotic expression vector pSFFV6 (25), or downstream the T7 promoter in pGEM1.
Construction of gE-KK and -SS Mutants-The gE mutants containing the cytoplasmic tail of the yeast protein Wbp1p with either the C-terminal KKXX or the SSXX signals were constructed by polymerase chain reaction-based amplification using reverse primers in which the corresponding sequences of the wild-type or mutated Wbp1p cytoplasmic tails had been introduced as translational fusions with the sequence of the gE transmembrane domain. The resulting polymerase chain reaction fragments were digested with XbaI and HindIII and cloned into the same sites in the pSFFV6 vector. The sequences of both mutants were verified using the Sanger dideoxy chain termination method.
Antibody Generation-The antibody 1667 against the full-length gE was obtained by cloning a cDNA fragment coding for the mature VZV gE open reading frame with a hexahistidine tag at the C terminus into the NcoI/BamHI sites of the pET15b vector (Novagen, Wiesbaden, Germany). The protein was expressed in BL21 cells and the insoluble fraction (containing most of the recombinant gE) was solubilized in 8 M urea and loaded on a Talon metal-affinity column (CLONTECH, Heidelberg, Germany). After extensive washing, the bound protein was eluted with SDS-loading buffer, and approximately 50 g were loaded on a 7.5% preparative SDS-polyacrylamide gel. The part of the gel containing the recombinant protein was excised, homogenized using a Teflon-glass homogenizer, mixed with either Freund's complete or incomplete adjuvant, and used to immunize rabbits following standard procedures.
To produce the 2679 antibody against the cytoplasmic tail of gI, a fragment comprising amino acids 314 -354 of the gI precursor form was cloned into the pGEX-4T-1 vector (Pharmacia, Freiburg, Germany) as a fusion to glutathione S-transferase. The glutathione S-transferase-gI fusion was expressed in XL-1 Blue cells and purified by affinity chromatography on a glutathione-Sepharose column (Pharmacia, Freiburg, Germany), following the manufacturer's instructions. After elution, the fusion protein was loaded on a preparative 7.5% preparative SDSpolyacrylamide gel. The gel fragment containing the band was excised, homogenized, mixed with Freund's adjuvant, and used to inoculate rabbits following a standard immunization schedule. The serum was affinity-purified by incubation with a nitrocellulose strip onto which the recombinant glutathione S-transferase-gI had been previously bound (26).
Antibody Uptake-For the internalization assays, a continuous uptake was performed in which transfected cells seeded on coverslips were washed with prewarmed ␣-MEM and overlaid with 200 l of complete ␣-MEM in which the antibodies had been diluted as specified in the figure legends. After 1-h incubation, the internalization medium was removed and the cells were immediately fixed and processed for immunofluorescence using fluorescein or rhodamine-coupled secondary antibodies.
VZV Infection-Due to the cell-associated nature of VZV, infections were carried out as described previously (27), by co-culture of VZV (Ellen strain)infected Vero cells with noninfected either Vero or HeLa cells. For the immunofluorescence experiments, infected and noninfected cells were plated on coverslips at a 1:4 ratio and grown in complete ␣-MEM for different times as indicated in the figure legends. The cells were subsequently fixed and processed for immunofluorescence.
Miscellaneous-Published procedures were used for vaccinia T7 infection (12), metabolic labeling of the cells and immunoprecipitation (28), and for calcium-phosphate transient transfection and indirect immunofluorescence (12).

RESULTS
The Subcellular Localization of gI Depends on the Simultaneous Expression of gE-To address the subcellular localization of the VZV-envelope glycoprotein gI (gpIV), we have cloned the complete gI open reading frame in the mammalian expression vector pSFFV6 (25). We have used this construct to perform transient transfection assays in HeLa cells followed by immunofluorescence using anti-gI-specific antibodies. This experiment revealed that, in every transfected cell, gI was exclusively found at the cell surface (Fig. 1b). As a control, we also performed transient transfections with an analogous construct in which the complete gE (gpI) open reading frame had been inserted into the same expression vector (12). In agreement with previous data (12), in cells transfected with the gE expression vector, this protein was exclusively localized in the perinuclear region of the cell (Fig. 1c), in a compartment that has been previously identified as the TGN, based on its co-localization at the light microscopy level with the TGN markers TGN38, furin, and sialyltransferase (12), at the electron microscopic level with galactosyltransferase, 2 and by its sensitivity to BFA and nocodazole, two drugs that affect the morphology of this compartment (12).
We next looked at the localization of gI in HeLa cells that had been simultaneously transfected with gE and gI expression constructs. In these cells, expression of gI was mainly restricted to the perinuclear region of the cell, largely colocalizing with gE, and almost absent from the cell surface (Fig. 1, e and f). We also performed an analogous double-transfection experiment using gI and an unrelated TGN marker (the convertase furin), whose intracellular traffic closely resembles that of gE (13)(14)(15)(16). In this case, whereas expression of furin was restricted to the perinuclear region of the cell (Fig. 1g), gI was exclusively detected at the cell surface ( Fig. 1h), therefore excluding the possibility that the perinuclear localization of gI in gE-expressing cells was simply due to an inability of the cell to properly sort gI at the TGN in the presence of another highly expressed molecule in this compartment. It has been previously suggested that gE and gI might share common antigenic determinants (29,30), which could explain the perinuclear signal attributed to gI in cells expressing gE if the antibodies used in this study would recognize any of these shared epitopes. However, this does not seem to be the case, since in cells expressing gI, no signal was detected with the anti-gE antibody (Fig. 1a) and, conversely, no signal was observed with anti-gI antibodies in cells exclusively transfected with gE (Fig. 1d). To exclude that the strong cell surface gI-staining observed in cells expressing gI alone or gI and furin could mask any labeling of intracellular compartments, we analyzed the single or doubletransfected cells by laser scanning confocal microscopy. As expected, no gI-staining could be detected intracellularly (data not shown). The same distribution was observed when polyclonal antibodies against the cytoplasmic domain of gI were used (data not shown). From all these results, we concluded that the localization of gI in transfected cells can be shifted from the cell-surface to the perinuclear region by the simultaneous co-expression of gE.
Intracellular Distribution and Traffic of gI in the Presence of gE-The colocalization experiments shown above indicate that, when expressed together, gE and gI are localized to the same cellular compartment, but they do not prove that this compartment is indeed the TGN, the organelle where gE accumulates when expressed alone (10 -12). We have previously used the rapid tubularization in response to BFA as a hallmark of the TGN to distinguish it from other membrane-bound compartments clustered in the perinuclear region of the cell (12). When HeLa cells expressing gE and gI were treated for 5 min with 10 g/ml BFA, fixed and decorated with anti-gE and anti-gI antibodies, both molecules were found to colocalize in thin tubules that emanated from the perinuclear region into the cell periphery (Fig. 2, a and b), strongly suggesting that the TGN is the compartment where gE and gI accumulate upon co-expression.
Another property of certain TGN markers (gE, furin, and TGN38) is their ability to constantly cycle between the TGN and the cell surface (16,18,31). Since gI is found mainly in the TGN in the presence of gE, we wanted to investigate whether this also involves cycling of gI between these two compart-ments. To address this question, anti-gE and anti-gI antibody uptake experiments were performed on HeLa cells that had been double-transfected with gE and gI expression constructs. As shown in Fig. 2c and in agreement with our previous findings (12), after 1 h of incubation, the anti-gE antibodies were mainly concentrated in the perinuclear region of the cell, as a result of their internalization bound to the luminal domain of the recycling gE molecules. Interestingly, the anti-gI monoclonal antibody, when incubated with the cells for the same time, was also found in the perinuclear region, colocalizing with the anti-gE antibodies (Fig. 2d). In contrast, when the same experiment was performed on cells transfected exclusively with the pSFFV-gI construct, only cell surface bound anti-gI antibody could be detected (data not shown). These data indicate that gI, when simultaneously co-expressed with gE, cycles between the TGN and the cell surface, and that its accumulation in the TGN most likely depends on its rapid internalization from the cell surface together with gE, suggesting that the distribution of gI when co-expressed with gE is indistinguishable from that observed for gE when expressed alone.
Localization of gI in Cells Expressing an ER Resident Form of gE-The results presented so far suggest that gE and gI are found within a complex in the cell whose traffic and distribution is solely determined by the sorting information in the cytoplasmic tail of gE. To verify this hypothesis, we constructed a modified version of gE in which its cytoplasmic tail had been replaced by that of the yeast protein Wbp1p, a type-I membrane protein that forms part of the ER resident oligosaccharyl-transferase complex (32). The tail of Wbp1p, which contains a consensus KKXX ER retention motif, has been shown to be sufficient to confer ER localization to reporter molecules both in mammalian and yeast cell systems (33,34). When the 2 A. Alconada, S. Röttgers, and B. Hoflack, unpublished observations.

FIG. 2.
Recycling of gI and BFA effect on its intracellular distribution. a and b, HeLa cells that had been double-transfected with the pSFFV-gE and pSFFV-gI plasmids were treated with 10 g/ml BFA for 5 min immediately prior to fixation. The cells were double-labeled with a polyclonal antibody against gE and a monoclonal antibody against gI. gE and gI were detected using FITC-coupled goat anti-rabbit IgGs (a) and TRITC-coupled goat anti-mouse IgGs (b), respectively. c and d, HeLa cells that have been double-transfected with the pSFFV-gE and pSFFV-gI plasmids were incubated for 2 h in ␣-MEM containing the polyclonal 1667 anti-gE (1:200 dilution) and the monoclonal antibody SG4 against gI (1:20 dilution). The cells were immediately fixed and stained using FITC-coupled goat-anti-rabbit IgGs and TRITC-coupled donkey anti-mouse IgGs to detect, respectively, the internalized anti-gE antibody (c) or the SG4 monoclonal antibody (d). Bar ϭ 20 m.
gE-KKXX and the gI expression constructs were simultaneously transfected into HeLa cells and the localization of both molecules was assessed by indirect immunofluorescence, both the gE-KKXX chimera and gI were found in a cytoplasmic reticular compartment showing all the morphological features of the ER (Fig. 3, a and b). As a control, we also constructed a gE-SSXX expression plasmid, in which the two lysines at positions Ϫ3 and Ϫ4 in the KKXX signal have been replaced by serines. This mutation is known to abolish the ER retention capacity of the KKXX motif (33). As expected, in cells coexpressing the gE-SSXX mutant and gI, both molecules were only detected at the cell surface (Fig. 3, c and d). These result confirms our prediction that the intracellular traffic of gI is exclusively determined by the sorting information on the tail of gE, presumably as a reflect of their association in the early secretory pathway.
Effect of gE on the Maturation of gI-We then asked whether the expression of one given protein could influence the maturation of the other. In order to address this question, gE and gI were expressed either alone or simultaneously in HeLa cells with the help of a T7 RNA-polymerase recombinant vaccinia virus. The cells were metabolically labeled with radioactive methionine, chased for increasing periods of time, and lysed, and the lysates were immunoprecipitated with anti-gE-and anti-gI-specific antibodies. To identify the precursor and mature forms of gE and gI, both molecules were immunoprecipitated from cells lysates that were obtained either immediately after the labeling period or after 6 h of chase. The results showed that gE was initially synthesized as a 70-kDa band that was converted during the chase to a 100-kDa polypeptide, and gI was initially found as a 50-kDa band that matured to yield a fuzzy 65-kDa band (Fig. 4a). These values are in agreement with those found by other groups, either in transfected (21,35) or in VZV-infected cells (36), for both the precursor and mature forms of gE and gI. When gE was expressed alone, immunoprecipitation with anti-gE antibodies revealed that maturation of the protein occurred rather rapidly, since as early as 20 min after initiation of the chase, almost 50% of the labeled 70-kDa precursor molecule was converted to the mature 100-kDa form (Fig. 4, b and c). When gI was expressed alone and immunoprecipitated with anti-gI antibodies under analogous conditions, its processing occurred very slowly, requiring more than 1 h to convert only 20% of the precursor to the mature form (Fig. 4, b and c). However, when gE and gI were expressed together, processing of gI was considerably enhanced, because 50% of the mature form could be detected after only 40 min of chase (Fig. 4, b and c). Under the same conditions, no difference was observed in the maturation of gE, when compared with the results obtained when this protein was expressed alone (Fig.  4b). In addition, the anti-gE-and anti-gI-specific antibodies failed to immunoprecipitate any gI and gE, respectively (data not shown).
It is worth mentioning that, when both proteins were expressed together, processing of gI occurred almost with identical kinetics as the processing of gE (Fig. 4c). However, under these conditions, we reproducibly observed a decrease in the amounts of gE and gI that could be immunoprecipitated with their cognate antibodies when compared with the single transfections (Fig. 4b). Since the expression levels of gE and gI were similar in double-transfected as in single-transfected cells, this finding could be explained if gE and gI would form large oligomeric complexes in the ER that would be poorly extracted from the membrane during the preparation of the cell lysate. Furthermore, when both proteins were co-expressed and immunoprecipitations were performed with the anti-gI antibody, bands of similar intensity for the precursor and mature forms of gE were also found in the immune complex (Fig. 4b), thereby suggesting that both molecules form a stoichiometric complex. Our data also indicate that the absence of gE does not lead to an absolute block on the maturation of gI, but rather to a decrease in its processing rate, as shown by the appearance of gI in the cell surface in transfected cells when the cells are observed 48 h, after transfection (see Fig. 1) and by the complete maturation of gI observed after 6 h of chase (data not shown). Thus, gE and gI can associate in the early compartments of the secretory pathway. This association facilitates the exit of the whole complex from the ER and determines the targeting of gI which lacks trafficking signals.
Localization of gE and gI in VZV-infected Cells-We then investigated the localization of gE and gI in cells infected with VZV. This virus is extremely cell-associated when propagated in cultured cells and therefore, infections must be performed by co-culture of noninfected with already infected cells. An additional difficulty was the unavailability of antibodies able to recognize endogenous TGN markers in a susceptible host cell line (Vero cells are a convenient host for VZV). To circumvent this problem, we undertook a double approach. First, we cultured VZV-infected Vero cells with noninfected Vero cells and used these cells to determine by immunofluorescence the localization of gE and gI. Second, VZV-infected Vero cells were co-cultured with a HeLa clone that stably expresses a VSV-G epitope-tagged version of sialyltransferase (37). The results obtained using both approaches are shown in Fig. 5.
In VZV-infected Vero cells, numerous cells could be observed where gE and gI were exclusively found in the perinuclear region of the cell (Fig. 5, a and b). In addition, when these cells were treated for 5 min with 10 g/ml BFA, numerous tubules that could be simultaneously labeled with anti-gE and anti-gI antibodies could be detected (Fig. 5, c and d). In many other cells, the perinuclear labeling was accompanied by a multitude of cytoplasmic vesicles that could also be labeled simultaneously with anti-gE and anti-gI antibodies (Fig. 5, a and b) and that seemed to be resistant to BFA treatment (not shown). In other cells, especially in those showing an evident cytopathic effect, the signal for both glycoproteins was much stronger and present also in the cell surface (not shown). When observed at lower magnification, the strongly expressing cells were normally found in the central region of the infectious plaque, whereas those cells with a restricted perinuclear expression of gE and gI were almost exclusively found in the plaque periphery (not shown), suggesting that these latter cells represent an earlier stage of VZV infection.
In the second approach, VZV-infected Vero cells were cocultured with ST-VSV-G-expressing HeLa cells for 12 h. The cells were fixed and processed for immunofluorescence to simultaneously detect the VSV-G epitope contained in the sialyltransferase and either gE or gI. Despite the very limited number of ST-VSV-G-expressing HeLa cells that could be infected with VZV, whenever a HeLa cell could be labeled with anti-gE (Fig. 5e) or anti-gI (Fig. 5g) antibodies, the signals for both VZV glycoproteins were restricted to the perinuclear region of the cell and were found to colocalize with ST-VSVG (Fig.  5, f and h). These results strongly suggest that, during the early stages of VZV infection, both gE and gI can be found concentrated in a region of the cell that most likely corresponds to the TGN.

DISCUSSION
In the present work, we show that, in cells transfected with gE and gI, two different glycoproteins of the VZV envelope, both molecules localize to the TGN. A similar situation is also observed in VZV-infected cells during the early stages of infection. Our data also indicate that gE and gI form a complex in the early stages of the secretory pathway. This is supported by the following evidences. First, the expression of an ER resident form of gE (the gE-KKXX mutant) leads to the accumulation of gI in this compartment. Second, gE and gI can be coimmunoprecipitated shortly after their synthesis as precursor forms containing nonprocessed oligosaccharides (data not shown) (21). For this reason, it should be expected that the interaction between these two proteins would occur directly through their polypeptide backbones. In this regard, Yao et al. (21) have mapped a cysteine-rich highly conserved region in the ectodomain of gE (amino acids 342-446), whose deletion leads to a substantial decrease in the yield of co-immunoprecipitation with gI. In addition, the gE⅐gI complex seems to contain equimolar amounts of both components. Coimmunoprecipitation experiments show that radiolabeled gE and gI exhibit similar intensity, as it could be expected for two molecules containing similar proportion of sulfur-containing amino acids.  N-[1-(2,3-dioleoyloxy)propyl]-N,N,Ntrimethylammonium salts with either pGEM-gE or pGEM-gI. The cells were labeled with radioactive methionine for 30 min and either immediately lysed (lanes 1 and 3) or after 6 h of chase in the presence of an excess of cold methionine (lanes 2 and 4). After the chase, the cells were lysed, and immunoprecipitations were carried out with either the 1667 (polyclonal anti-gE) (lanes 1 and 2) or the SG4 (monoclonal anti-gI) (lanes 3 and 4) antibodies. The position in the gel of the precursor and mature forms of gE and gI are indicated by arrowheads. b, maturation of gE and gI. Cells were transfected using an analogous procedure with either pGEM-gE (upper panel), pGEM-gI (middle panel), or a combination of the two plasmids (lower panel), labeled for 30 min with radioactive methionine, and chased for 0, 10, 20, 40, 60, and 100 min. After the chase, the cells were lysed and immunoprecipitated with the 1667 antibody (polyclonal anti-gE) (upper panel), or the SG4 (monoclonal anti-gI) (middle and lower panels) using protein A-Sepharose. The resulting immune complexes were analyzed by electrophoresis on 10% SDS-polyacrylamide gels. c, gels derived from three to six experiments as that shown in b were exposed to a storage phosphor screen, and the intensities of the band at each time point for the precursor and mature forms of gE and gI were quantitated. The mean values Ϯ S.E. of the percentage of the mature forms at each time point were plotted for gE expressed alone (squares), gI expressed alone (triangles), and gI when expressed together with gE (circles). Besides, when the expression level of gI is higher than that of gE (for instance, using mammalian expression vectors with different promoters), only a small fraction of the total gI can leave the ER in association with gE (data not shown).
According to all these data, the most likely scenario is that gE, by interacting with gI, facilitates its folding and, consequently, accelerates its release from the ER quality control system. Similar situations have been previously described for the homologous complexes from three other alphaherpesviruses, PRV (24), BHV-1 (38), and FHV-1 (23,39). In cells infected with different alphaherpesviruses, it has been reported that processing of gE is impaired in the absence of gI (23,24,38) and, in at least one of them, gE was found by immunofluorescence to remain in the ER in the absence of gI (23). However, it has been recently shown that, in cells infected with a gI-deleted VZV strain, glycosylation and intracellular delivery of gE can occur normally (40).
After association in the ER, gE, and gI reach the TGN, where they are found at steady state. The localization of this complex results from a rapid recycling between the TGN and the plasma membrane. Since gI is devoid of sorting information, the targeting to the TGN and the recycling of the gE⅐gI complex must solely rely on the sorting signals found in tail of gE (12,41). The association of viral envelope glycoproteins to form hetero-oligomers that are transported according to the sorting information found in just one of the glycoproteins is a well known mechanism to ensure co-segregation of viral glycoproteins. This situation can be found, for example, in the viruses of the bunyavirus family, whose budding occurs in the Golgi complex (42). In this case, the G2 envelope glycoprotein, which is found in the cell surface when expressed alone (43), can be targeted to the Golgi complex by forming an heterodimer with the G1 glycoprotein (44), which contains a Golgi-retention signal (45). A similar example is provided by the rubella virus, whose E1 glycoprotein is targeted to the Golgi via its interaction with E2 (46). The interaction between VZV gE and gI constitutes, to our knowledge, the first example of such a mechanism for localization to the TGN. The fact that complex formation between gE and gI occurs not only in VZV (21,36), but also in their homologues from HSV-1 (22), PRV (47), FHV-1 (23), and BHV-1 (38), suggests that our findings illustrate a common feature for all alphaherpesviruses.
We have observed that the localization of gE and gI to the TGN occurs not only in cells expressing both proteins from cloned cDNAs, but also in VZV-infected cells. This finding provides additional support to the idea that this organelle plays an essential role in VZV assembly (5). However, in VZV-infected cells, the gE⅐gI complex also appears in cytoplasmic vesicles devoid of any TGN markers. These structures could correspond to the previously described intracellular vacuoles that originate as a result of the viral infection and that have been proposed to constitute the major sites of VZV virion accumulation (48). Similar structures were also shown to be accessible to endocytic tracers and to contain gE and mannose 6-phosphate receptor immunoreactivity (9). According to these properties, the gE-and gI-positive structures detected in VZVinfected cells would correspond to endocytic compartments, to which VZV could have access by means of the interaction in the TGN between the mannose 6-phosphate receptors and its envelope glycoproteins (that have been shown to contain mannose 6-phosphate modifications (9). In this way, mature viruses could be packaged into clathrin-coated vesicles and subsequently delivered to endocytic compartments in the same way as are lysosomal enzymes (49). This could agree with morphological studies illustrating the budding of nucleocapsids in the TGN (5).
How do all these data on the intracellular trafficking of the gE⅐gI complex agree with the current knowledge on its function and localization? The available information suggests that both components of the VZV gE⅐gI complex (or of the homologous complexes from HSV-1 and PRV) are dispensable for viral entry, replication, and release of new viruses (50 -52). However, these two proteins seem to be required for direct cell-tocell spread of the corresponding viruses, as suggested by studies performed with alphaherpesviruses bearing deletions in the gE or gI genes (50,(52)(53)(54)(55)(56)(57)(58)(59)(60). The mechanism underlying the process of direct cell-to-cell spread remains largely uncharacterized, although it probably involves direct fusion of cell membranes. In the case of VZV, direct cell-to-cell spread seems to be the only productive way of infection when the virus is propagated in cultured cells, since secreted viruses appear to be non-infective (9,61,62). Earlier studies have reported that both gE glycoproteins and the gE⅐gI complexes from VZV (35,63) and HSV-1 (64) display Fc binding activity, although this activity is not present in every member of the subfamily (38,47). Therefore, the gE⅐gI could interact with plasma membrane receptors that contain domains belonging to the immunoglobulin superfamily and, in this way, promote cell fusion and consequently, contribute to the direct cell-to-cell spread. Since appearance of the gE⅐gI complex at the cell surface seems to be restricted to the late stages of infection, we hypothesize that the gE⅐gI complex could have a dual role during VZV biogenesis. (i) During the early stages of infection, the expression of gE and gI would be kept at low levels. As a consequence, the complex would remain in the TGN. Under these conditions, the gE⅐gI complex, most likely functioning together with the other envelope glycoproteins, could promote recruitment of the nucleocapsids and, therefore, contribute to the formation of mature viruses at the TGN (5). (ii) During the late stages of infection, expression of gE and gI would reach much higher levels, which would then lead to the appearance of the complex at the cell surface. This situation would therefore resemble the results obtained by other groups when using strong promoters to express gE and gI in transient transfections (21,35) and would simply reflect the known mislocalization of TGN molecules to the cell surface upon overexpression (12,15,19,(65)(66)(67). At this stage, the gE⅐gI complex, as a consequence of its appearance on the cell surface, could interact with receptors in neighboring noninfected cells and contribute to cell-to-cell spread.
In summary, we have characterized the intracellular traffic of the complex formed by two of the VZV envelope glycoproteins, gE and gI, and we have provided evidence that this complex traffics in an identical manner both in double-transfected as in VZV-infected cells. This complex is found mainly at steady-state at the TGN, which provides further support to the idea that this organelle plays an essential role during VZV assembly. Future studies will be aimed to characterize the intracellular transport of other VZV envelope glycoproteins as well as to identify putative viral-encoded factors that could control this process.