INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

The alphaherpesvirus glycoproteins E and I (gE and gI, respectively) form a hetero-oligomeric complex which is found in the viral envelope and at the surface of the infected cell (23, 38, 53, 54, 58, 61). Although dispensable for replication in cultured cells, the genes for gE and gI are conserved in all alphaherpesviruses studied to date (4, 11, 29, 36, 41, 43, 48, 54, 55). Infection experiments in both natural and experimental hosts indicate that the gE-gI complex is an important virulence factor: viruses deficient for gE and/or gI produce milder clinical signs, cause smaller primary lesions, and exhibit a lower degree of neuronal spread than the wild-type virus (8, 11, 12, 18, 27, 28, 40, 44, 48, 50, 53, 55).

Recently, Knapp and Enquist reported that the virulence of a pseudorabies virus (PRV) mutant deficient for gE and gI could be restored by complementation with gE and gI of bovine herpesvirus (26). This result suggests that gE-gI complexes of different alphaherpesviruses are functionally equivalent. However, the function of the gE-gI hetero-oligomer is not exactly known. For some herpesviruses, the gE-gI complex functions as a receptor for the Fc domain of immunoglobulin G and, consequently, may play a role in the evasion of humoral immunity (5, 6, 13, 14, 19, 20, 24, 30, 51). Most evidence, however, indicates that the gE-gI complex is primarily involved in cell-to-cell transmission, possibly by promoting cell fusion or virus release (4, 10-12, 60). Cell-to-cell spread differs in several respects from virus entry and apparently entails the transfer of the virus across cell junctions in a manner resistant to neutralizing antibodies (4, 11, 60). In vitro, virus mutants lacking gE and/or gI characteristically display a small-plaque phenotype (4, 11, 36, 37, 41, 48, 54, 60).

The biosynthesis of the gE-gI complex has been studied in detail for several alphaherpesviruses, and from this work the following picture emerges. The proteins are synthesized in the endoplasmic reticulum (ER) as N-glycosylated class I membrane proteins which readily assemble into noncovalently linked hetero-oligomeric complexes, most likely heterodimers (23, 25, 38, 53, 54, 58, 61). These are transported along the secretory pathway, concomitantly acquiring extensive posttranslational modifications: elaborate processing of the N-linked oligosaccharides, addition of O-linked oligosaccharides, and sulfatation, as well as phosphorylation (15, 16, 23, 30, 38, 43, 53, 54, 57, 58).

gE and gI both possess large cytoplasmic domains. In varicella-zoster virus, these domains contain signals that mediate cycling of the complex between the plasma membrane, the endosomes, and the trans-Golgi network (1, 42, 59). The cytoplasmic tails of gE and gI are dispensable, however, for complex formation (25, 37, 49). In the case of feline herpesvirus (FHV), a C-terminally truncated gI derivative of 166 residues (corresponding to the N-terminal half of the ectodomain) still assembles into a transport-competent complex with gE. An even shorter derivative, comprising the 93 N-terminal residues, can still bind to gE to yield a stable hetero-oligomer, but this complex is transport incompetent (37). These observations suggest that the N-terminal region of gI is involved in the interaction with gE, a notion supported by the observation that for varicella-zoster virus, mutations in the very N terminus of gI abolish complex formation (25).

Apparently, gE-gI function is primarily effectuated by the ectodomains. Deletion of the cytoplasmic tail of FHV gI only marginally affects plaque size (37). Furthermore, the cytoplasmic tail of PRV gE is dispensable for gE-gI-mediated neuronal spread. Upon retinal infection of rats, mutant viruses expressing truncated gE retained the ability to spread to all retinorecipient regions of the brain (49).

The ectodomains of gE and gI contain cysteine residues which are strictly conserved and are likely to form intramolecular disulfide bridges that stabilize the protein conformation. The cysteine map of gI is remarkably similar to those of two other short unique region-encoded alphaherpesvirus glycoproteins, gG and gD, in that they share a characteristic motif, C-X₁₁-C-X_8-10-C. This observation has led McGeoch (34) to postulate that gG, gD, and gI are evolutionary related and have arisen through gene duplication. The hypothesis predicts that gG, gD, and gI have similar disulfide-bonded structures (32). Another corollary of the hypothesis is that the formation of the conserved disulfide bonds must be important for the folding and/or function of these proteins: if gG, gD, and gI are related, they are separated by a large evolutionary distance, and the cysteine residues would not have been conserved unless their loss had entailed a decrease in viral fitness. The disulfide-bonded structure has been solved for the gD of herpes simplex virus (HSV) types 1 and 2 (32). Of the cysteines in the motif, the two C-terminal-most residues form a disulfide bond, yielding a small 8-residue loop; the N-terminal-most cysteine pairs with a fourth residue downstream of the motif.

Here, we have studied the disulfide bridges of a gI protein. We have probed the disulfide-bonded structure of FHV gI both by site-directed mutagenesis of cysteine residues and by biochemical analysis of the purified protein. The effects of CysSer substitutions on gE-gI complex formation and intracellular transport and on gE-gI-mediated cell-to-cell spread are discussed.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Cells, viruses, antisera, and plasmids. Cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Gibco BRL, Life Technologies, Inc.) supplemented with 10% fetal calf serum and 100 IU of penicillin and 100 µg of streptomycin per ml (DMEM-10% FCS). FHV strain B927 (22) was obtained from D. A. Harbour and propagated in Crandell feline kidney (CRFK) cells (American Type Culture Collection) (9). Recombinant vaccinia virus vTF7-3, expressing the bacteriophage T7 RNA polymerase (21), was obtained from B. Moss and propagated in RK-13 cells. Transient expression experiments were performed in OST7-1 cells (17). The monospecific rabbit antisera against FHV gE (Ra-gE) and gI (Ra-gI), the cat antiserum against FHV (Cat-FHV), and the plasmids pBS-gE, pBS-gI, pBS-gIM, and pUS1 have been described previously (37, 38).

Recombinant DNA techniques. Recombinant DNA techniques were performed according to the procedures of Sambrook et al. (46) and Ausubel et al. (3). Sequence analysis was performed with a T7 sequencing kit (Pharmacia Biotech). PCR was performed as described elsewhere (45), using the thermostable DNA polymerase of Thermus aquaticus (Taq polymerase; Gibco BRL, Life Technologies, Inc.) in accordance with the instructions of the manufacturer.

Mutagenesis of pBS-gIM. pBS-gIM was constructed by cutting pBS-gI with MluI, which cuts at nucleotide (nt) 497, and with XbaI, which cuts at a site downstream of the gI gene within the polylinker region of pBS-SK. The 3' recessive ends were filled in, using the large fragment of DNA polymerase I (Gibco BRL, Life Technologies, Inc.), and ligated. To replace Cys⁷⁹ (C₁) with Ser, two overlapping oligonucleotide primers of opposite polarity, no. 569 and no. 570 (Table 1), were designed, both of which directed a G²³⁶C substitution. The 5' and 3' halves of the gIM gene were PCR amplified with oligonucleotide 570 plus the M13 universal primer and with oligonucleotide 569 plus the M13 reverse primer, respectively. The PCR products were purified from low-melting-point agarose; 5 ng of each was mixed in a 50-µl volume of PCR buffer and denatured by incubation for 5 min at 95°C. After reannealing, heterologous DNA hybrids were elongated with Taq polymerase and a PCR was performed with the M13 universal and reverse primers to amplify the complete gIMC₁ gene. The resulting PCR product was digested with PmlI, which cuts at nt 277, and with HindIII, which cuts at a site upstream of the gIM gene within the polylinker region. The 292-bp PmlI-HindIII fragment was gel purified and exchanged for the corresponding sequences in pBS-gIM, yielding pBS-gIMC₁. Mutagenesis of Cys⁹¹ (C₂) and Cys¹⁰² (C₃) was performed accordingly, with primer pairs 582-581 and 604-603, respectively. To obtain pBS-gIMC₂ and pBS-gIMC₃, the PCR products obtained were cut with PmlI and HindIII or NotI, respectively; the NotI site is located within the polylinker region downstream of the gIM gene. The 292-bp PmlI-HindIII and 229-bp PmlI-NotI fragments were purified from the gel and exchanged with the corresponding sequences in pBS-gIM, yielding pBS-gIMC₂ and pBS-gIMC₃, respectively. pBS-gIMC₂₃ was constructed by inserting the PmlI-HindIII fragment of pBS-gIMC₂ into PmlI- and HindIII-digested pBS-gIMC₃. To construct pBS-gIMC₀, C₁ in pBS-gIMC₂₃ was replaced by Ser via PCR mutagenesis as described above. Sequence analysis of the relevant regions of each construct confirmed that no inadvertent nucleotide changes had occurred during PCR amplification and cloning.

Construction of plasmids pBS-sgE and pBS-sgI. To construct pBS-sgE, a PCR was performed with the M13 universal primer and primer 630 (Table 1), using plasmid pBS-gE (38) as a template. The PCR product was blunt ended with the large fragment of DNA polymerase I and digested with SnaBI. The resulting 204-bp fragment (nt 985 to 1188) was gel purified and inserted into the 4-kb SnaBI-SalI (blunt) fragment of pBS-gE, yielding pBS-sgE. In pBS-sgE, nt 1186 through 1599 of the gE gene were deleted and a termination codon was created downstream of the codon for Arg³⁹⁵.

Construction of recombinant virus FHV-gIC₁. FHVgI-LZ has been described previously (38). In this mutant, the gI gene had been disrupted by replacing nt 203 to 923 with an expression cassette, consisting of the lacZ gene downstream of the encephalomyocarditis virus internal ribosomal entry site. To construct an FHV recombinant that expresses gIC₁, the Cys⁷⁹Ser substitution was introduced into plasmid pUS1 (37), which contains a 7-kb EcoRV-BamHI fragment spanning the genes for gD, gI, gE, US8.5, US9, US10, and US1 (56). To this end, the 292-bp PmlI-HindIII fragment of pBS-gIMC₁ was first inserted into PmlI- and HindIII-digested pBS-gI, resulting in pBS-gIC₁. Subsequently, the XhoI-BamHI fragment from this plasmid was exchanged with the corresponding region in pUS1, yielding the transfer vector pUS1-gIC₁. Sequence analysis of the relevant region confirmed that pUS1-gIC₁ had the correct nucleotide substitution and that no inadvertent mutations had been introduced during the cloning procedures.

Transfection of vTF7-3-infected cells and metabolic labeling. Subconfluent monolayers of OST7-1 cells grown in 35-mm-diameter dishes were washed once with DMEM and infected with vaccinia virus vTF7-3 at a multiplicity of infection of 3 in DMEM at 37°C. One hour postinfection (p.i.), the cells were washed with DMEM and transfected with a mixture consisting of 2 to 5 µg of plasmid DNA, 500 µl of DMEM, and 10 µl of Lipofectin (Gibco BRL, Life Technologies, Inc.). After a 5-min incubation at room temperature, 500 µl of DMEM was added, and incubation was continued at 37°C. At 2 h p.i., the temperature was lowered to 32°C. From 4 to 5 h p.i., the cells were incubated with 1 ml of minimum essential medium with Earle's salts, lacking cysteine and methionine (Gibco BRL, Life Technologies, Inc.). Then, 100 µCi of Redivue L-[³⁵S] in vitro cell labeling mix ([³⁵S]Met plus [³⁵S]Cys; Amersham) was added to the culture medium and incubation was continued for 1 h. The cells were harvested either immediately or after a 2-h chase with DMEM-10% FCS containing 5 mM (each) L-methionine and L-cysteine.

Metabolic labeling of FHV-infected cells. Subconfluent monolayers of CRFK cells in 35-mm-diameter dishes were washed once with DMEM and infected with either wild-type FHV strain B927 or recombinant FHV at a multiplicity of infection of 5 at 37°C. At 1 h p.i., the culture medium was replaced by DMEM-10% FCS, and the incubation was continued at 37°C. Metabolic labeling was done as described for vTF7-3-infected cells, except that the cysteine-methionine depletion and subsequent labeling procedures were performed 2 h later and at 37°C, unless indicated otherwise.

RIPA and SDS-PAGE. Metabolically labeled cells were washed once with ice-cold phosphate-buffered saline and then lysed on ice in 600 µl of lysis buffer (20 mM Tris-Cl [pH 7.5], 1 mM EDTA, 100 mM NaCl, 1% Triton X-100) containing 1 µg of pepstatin A, 40 µg of aprotinin, and 1 µg of leupeptin per ml. For analysis under nonreducing conditions, the cells were washed with ice-cold phosphate-buffered saline containing 20 mM N-ethylmaleimide (NEM; Pierce) for 5 min and then lysed in lysis buffer supplemented with 20 mM NEM, in order to block free sulfhydryl groups. Nuclei and cell debris were removed by centrifugation for 1 min at 10,000 × g and 4°C. Two hundred microliters of the supernatant was mixed with 1 ml of detergent mix (50 mM Tris-Cl [pH 8.0], 62.5 mM EDTA, 0.4% sodium deoxycholate, 1% Nonidet P-40), and sodium dodecyl sulfate (SDS) was added to a final concentration of 0.25%. After 15 min on ice, the antisera were added (Cat-FHV and Ra-gE, 3 µl each; Ra-gI, 5 µl) and incubation was continued for 16 h at 4°C. Immune complexes were collected by adding 50 µl of a 10% (wt/vol) suspension of formalin-fixed Staphylococcus aureus cells (Pansorbin; Calbiochem) in detergent mix. After a 30-min incubation at 4°C, the precipitates were washed three times with radioimmunoprecipitation assay (RIPA) buffer (10 mM Tris-Cl [pH 7.4], 150 mM NaCl, 0.1% SDS, 1% sodium deoxycholate, 1% Nonidet P-40). Digestion of immunoprecipitated proteins with peptide:N-glycosidase F (PNGase F; New England Biolabs) was performed in accordance with the instructions of the manufacturer but in the absence of reducing agents. Finally, the proteins were dissolved in 30 µl of modified Laemmli sample buffer (7) containing 5% -mercaptoethanol (-ME) unless indicated otherwise, heated for 5 min at 95°C, and analyzed by SDS-polyacrylamide gel electrophoresis (PAGE).

Purification and EndoLys-C digestion of secretory gI (sgI). vTF7-3-infected OST7-1 cells were transfected with pBS-sgE and either pBS-sgI or pBS-gIMC₁. Mock-transfected cells served as a negative control. Cells were depleted for cysteine from 4 to 5 h p.i.; [³⁵S]Cys (ICN) was then added to the tissue culture supernatant to a final concentration of 150 µCi/ml, and the cells were metabolically labeled from 5 to 8 h p.i. The culture supernatant was harvested, and detached cells and cell debris were removed by centrifugation at 12,000 × g for 2 min at 4°C. SDS was added to the supernatants to a final concentration of 1.3%; this was followed by a 15-min incubation at room temperature to dissociate gE-gI hetero-oligomers. Subsequently, the samples were diluted with detergent mix to an SDS concentration of 0.25%, and a standard RIPA with Ra-gI serum was performed. Immune complexes were eluted from the formalin-fixed S. aureus cells with PNGase F buffer (50 mM sodium phosphate [pH 7.5], 0.5% SDS) and then treated with PNGase F. The reaction mixtures were diluted fivefold with distilled water and 10× endoproteinase Lys-C (EndoLys-C) buffer to end concentrations of 25 mM Tris-Cl (pH 8.5) and 1 mM EDTA. Proteolytic digestion was performed in a reaction volume of 250 µl at estimated enzyme/substrate ratios of 1:40 to 1:100 by adding 1.7 µg of rehydrated EndoLys-C (sequencing grade; Boehringer Mannheim). Incubation was for 14 h at 37°C. The digestion was stopped by adding 125 µl of modified Laemmli sample buffer (7) with or without 5% -ME and then heating for 5 min at 95°C. The peptides were separated by using the discontinuous tricine-SDS-PAGE system developed by Schägger and von Jagow (47), with the separating gel containing 13.3% (wt/vol) glycerol and a monomer/cross-linker ratio of 16.5% T/6% C (where T denotes the total percentage of both acrylamide and bisacrylamide and C denotes the percentage concentration of the cross-linker relative to the total concentration T [see also reference 47]). Gels were run for 2 h at 30 V followed by 9 h at 30 mA, fixed for 30 min in 50% methanol-10% acetic acid, and vacuum dried at 80°C. Radiolabeled peptides were visualized by fluorography, using TranScreen phosphor screens (Kodak) in combination with Kodak BioMax MS films. The Rainbow ¹⁴C-methylated protein low-molecular-weight marker (Amersham) was used to determine apparent molecular weights.

Plaque assays and immunohistochemistry. Plaque assays in CRFK cells, visualization of plaques by immunohistochemistry, and calculation of average plaque size have been described previously (37).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

SDS-PAGE analysis of FHV gI under reducing and nonreducing conditions. The four cysteine residues in the ectodomain of FHV gI are located at positions 79 (C₁), 91 (C₂), 102 (C₃), and 223 (C₄) (Fig. 1). To determine whether they are involved in the formation of intra- and/or intermolecular disulfide bonds, gI was analyzed under reducing and nonreducing conditions. CRFK cells were infected with wild-type FHV strain B927 or, as a negative control, with the gI-deficient recombinant virus FHVgI-LZ (38). The cells were metabolically labeled from 7 to 8 h p.i. and lysed in the presence of 20 mM NEM to alkylate free sulfhydryl groups. The lysates were subjected to immunoprecipitation with a rabbit antiserum raised against residues 20 to 36 of FHV gI (Ra-gI) (38), and the precipitated proteins were separated in SDS-7.5% polyacrylamide gels in the presence or absence of -ME. In some experiments, the immunoprecipitates were treated with endoglycosidase H (EndoH) to distinguish immature gI (igI) species from mature post-ER forms (data not shown). As reported previously (37, 38), under reducing conditions, igI migrated as a 67,000-molecular-weight (67K) protein whereas mature gI (mgI) produced an EndoH-resistant 80 to 100K smear (Fig. 2, right panel, B927 ). When gI was analyzed under nonreducing conditions (Fig. 2, left panel, B927 ), three species were observed, one of which comigrated with reduced igI. The other products, migrating at 60K and at 76 to 95K, apparently represent oxidized forms of igI and mgI, respectively.

View larger version (14K): [in this window] [in a new window]   FIG. 1.   Schematic representation of FHV gI, gIM, and the various gIM derivatives carrying CysSer substitutions. The proteins are represented by open boxes. The N-terminal signal peptide and the transmembrane region of each protein are indicated by hatched boxes. C1, C2, C3, and C4, representing the cysteine residues in the ectodomain of gI at positions 79, 91, 102, and 223, respectively, are indicated by black bars. A fifth cysteine residue, C5, present in the cytoplasmic domain at position 322, is also shown. The position of Arg166 is marked by an arrowhead.

View larger version (35K): [in this window] [in a new window]   FIG. 2.   Analysis of gI and gIC1 under reducing and nonreducing conditions. Subconfluent monolayers of CRFK cells were infected with either wild-type FHV strain B927 (B927) or FHV-gIC1 (gIC1). Cells infected with the gI-deficient recombinant FHVgI-LZ (gI) (38) served as a negative control. The cells were metabolically labeled from 7 to 8 h p.i. at 37°C. The cell lysates were prepared in the presence of 20 mM NEM to block free sulfhydryl groups and were then subjected to RIPA with the gI-specific antiserum Ra-gI. The proteins were either treated with PNGase F (+) or mock treated () and separated in SDS-7.5% polyacrylamide gels in the absence and presence of -ME ( -ME and + -ME, respectively). The closed arrowheads indicate immature reduced and oxidized gI species (igIred and igIox, respectively) and mature reduced and oxidized gI species (mgIred and mgIox, respectively). In the left-hand panel, mature gIC1 is indicated by an open arrowhead. Molecular sizes are in kilodaltons.

The kinetics of disulfide bond formation and folding during FHV gI synthesis. To study the kinetics of disulfide bond formation and subsequent gI maturation, FHV-infected CRFK cells were pulse-labeled for 5 min at 8 h p.i. and then harvested either immediately or after chase periods of up to 120 min. Prior to analysis under reducing and nonreducing conditions, the immunoprecipitated gI species were treated with PNGase F. As shown in Fig. 3, mgI first appeared after a 30-min chase. Analysis under nonreducing conditions revealed that after a 5-min pulse-labeling, the majority of newly synthesized gI comigrated with igI_red and thus was either fully reduced or had acquired intramolecular disulfide bonds that did not affect migration in SDS-polyacrylamide gels (Fig. 3, left panel, B927 0'). However, as determined by -scanning, 25% of the labeled gI was already in the 38-kDa igI_ox form. During the subsequent 10-min chase, the amount of igI_ox increased to 60%. Thus, the conversion of igI_red into igI_ox occurred predominantly posttranslationally, with an estimated half-life (t_1/2) of 5 to 7 min. Subsequently, igI_ox declined with an estimated t_1/2 of 35 min, concomitant with the appearance of mgI_ox. The folding of gI appeared to be an inefficient process. After a 120-min chase, about 35% of gI still remained in the ER, and most of it comigrated with igI_red. Only 12% of the labeled gI appeared to have been converted into mgI_ox.

View larger version (41K): [in this window] [in a new window]   FIG. 3.   Pulse-chase analysis of FHV gI under reducing and nonreducing conditions. Subconfluent monolayers of CRFK cells were infected with wild-type FHV strain B927 (B927). At 8 h p.i., the cells were metabolically labeled for 5 min at 37°C and harvested either immediately (0') or after chase periods of 5, 10, 30, 60, and 120 min (5', 10', 30', 60', and 120', respectively). To assure quantitative precipitation of labeled gI also after a 60-min or 120-min chase, the amount of unlabeled gI was reduced by adding 0.5 mM cycloheximide to the culture supernatant 30 min after labeling. Cells infected with FHVgI-LZ (gI) and harvested after a 5-min pulse-labeling (0') or following a subsequent 120-min chase period (120') served as negative controls. Cell lysates, prepared in the presence of 20 mM NEM, were subjected to RIPA with Ra-gI. To facilitate the analysis, all samples were treated with PNGase F. The proteins were separated in SDS-7.5% polyacrylamide gels in the absence () and presence (+) of -ME. Immature reduced and immature oxidized gI species (igIred and igIox, respectively) and mature reduced and mature oxidized gI species (mgIred and mgIox, respectively) are indicated. Molecular sizes are in kilodaltons.

C₂ and C₃ of gI are required for maturation of gE. We have previously shown that a gI derivative, gIM, which consists of only the N-terminal 166 residues and thus lacks C₄, still induces gE maturation (37). Apparently, the formation of an intramolecular disulfide bridge involving C₄ is not essential for the interaction with gE or for ER-to-Golgi transport of the resulting complex. To study the role of the remaining cysteine residues, gIM derivatives in which C₁, C₂, or C₃ was replaced by Ser were constructed. These mutant proteins were coexpressed with gE in the vaccinia virus-based vTF7-3 expression system (21) and metabolically labeled for 1 h, followed by a 2-h chase. Immunoprecipitation was performed with an FHV-specific feline hyperimmune serum, Cat-FHV. Conversion of the immature ER-resident 83K gE species into the mature 95K form was interpreted to indicate the assembly of a transport-competent gE-gIM hetero-oligomer (37). As shown in Fig. 4, substitution of C₁, as in gIMC₁, did not affect gE maturation. Thus, C₁, like C₄, is not essential for this process. In contrast, gIMC₂, gIMC₃, and gIMC₂₃, in which C₂ and/or C₃ had been replaced, and gIMC₀, which lacks all cysteine residues, failed to induce maturation of gE. Immunoprecipitation with the Ra-gI antiserum confirmed that the gIM derivatives had been expressed to similar extents. Moreover, immature gE was found to coprecipitate with gIMC₂, gIMC₃, and gIMC₂₃, while in the reciprocal experiment, with a monospecific antiserum against gE, coprecipitation of the gIM derivatives was observed (data not shown). Apparently, replacement of C₂ and/or C₃ does not prevent the formation of the gE-gIM hetero-oligomer. However, the resulting complex is no longer transport competent and is retained in the ER. These observations fit a model for the disulfide-bonded structure of gI in which C₂ and C₃ as well as C₁ and C₄ form intramolecular disulfide bridges and in which the C₂-C₃ bond (but not the C₁-C₄ bond) is essential for ER-to-Golgi transport of the gE-gI hetero-oligomer.

View larger version (40K): [in this window] [in a new window]   FIG. 4.   Coexpression of gE with cysteine mutants of gIM. vTF7-3-infected OST7-1 cells were cotransfected with plasmid pBS-gE and a plasmid encoding gIM or one of its derivatives (gIMC1, -C2, -C3, -C23, or -C0). The cells were metabolically labeled from 5 to 6 h p.i., followed by a 2-h chase. The cell lysates were subjected to RIPA with an FHV-specific feline hyperimmune serum, Cat-FHV. The samples were analyzed in SDS-7.5% polyacrylamide gels. The immature, EndoH-sensitive 83-kDa gE species (igE) and the mature, EndoH-resistant, 95-kDa gE species (mgE) are indicated with arrowheads. Molecular sizes are in kilodaltons.

Biochemical evidence for a C₁-C₄ disulfide bond. To test our hypothesis that C₁ and C₄ form a disulfide bridge, we followed a strategy entailing proteolytic digestion of purified radiolabeled gI and analysis of the resulting peptides under reducing and nonreducing conditions. To this end, secretory forms of gI (sgI) and gE (sgE), truncated immediately N terminal of their predicted transmembrane regions, were coexpressed in the vTF7-3 expression system. Pilot experiments had shown that sgE and sgI, when expressed separately, are retained in the cell. However, when coexpressed, they form a complex which is secreted into the culture supernatant (39), thus providing a source of mature gI virtually devoid of immature forms that could complicate the analysis. Cells, mock transfected or cotransfected to express gIMC₁ and sgE, were used as controls. The proteins were metabolically labeled with [³⁵S]Cys, the tissue culture supernatant was harvested, and the sgE-sgI complex was dissociated with SDS. RIPA was performed with the Cat-FHV and Ra-gI antisera; this was followed by PNGase F digestion to remove N-glycans. PNGase F-treated mature sgE and sgI migrate at 53K and 44K, respectively, in SDS-polyacrylamide gels, whereas their immature deglycosylated forms run at 44K and 34K, respectively (39). The size differences between the intracellular and secreted forms indicate that the truncated proteins are posttranslationally modified to extents similar to those of full-length gE and gI. As shown in Fig. 5a, the Cat-FHV serum predominantly detected mature sgE whereas the Ra-gI serum readily precipitated both sgI and the 17K gIMC₁ product (Fig. 5a, left panel). Under nonreducing conditions (Fig. 5a, right panel), sgI migrated at 38K and thus showed a 6K shift in apparent molecular weight, identical to that of full-length gI (Fig. 2).

View larger version (28K): [in this window] [in a new window]   FIG. 5.   Protein-chemical analysis of the disulfide-bonded structure of FHV gI. (a) Secretory forms of gE and gI (sgI) were coexpressed in vTF7-3-infected OST7-1 cells. Mock-transfected cells (m) and cells coexpressing gIMC1 and gE (C1) were used as controls. The cells were metabolically labeled with [35S]Cys from 5 to 8 h p.i. The culture supernatants were harvested, treated with SDS to dissociate the sgE-sgI hetero-oligomers, and subjected to RIPA with either Cat-FHV (CFHV) or Ra-gI (RagI). N-linked oligosaccharides were removed by treating the immunoprecipitates with PNGase F. The samples were analyzed in SDS-15% polyacrylamide gels in the presence (+) or absence () of -ME. Of each Ra-gI precipitate, only one-quarter was analyzed; the remainder was digested with EndoLys-C (see below). Mature reduced and oxidized sgI (sgIred and sgIox, respectively), mature sgE (sgE), and gIMC1 are all indicated. (b) The EndoLys-C restriction map of sgI. The upper panel shows a schematic representation of full-length gI. The signal peptide and the transmembrane anchor are indicated by hatched boxes, the cysteine residues in the ectodomain (C1 to C4) are indicated by black bars, and potential O-glycosylation sites are indicated by "lollipops." The lower panel shows a schematic representation of sgI. The protein is depicted as a horizontal bar, and the relative positions of the four cysteine residues and Arg166 (arrowhead) are indicated. EndoLys-C cuts C terminal to lysine residues. The positions of these residues in sgI are marked by vertical lines below the horizontal bar. The predicted sizes of the fragments containing C1 to C4 are given in amino acids (aa). (c) Schägger-von Jagow SDS-PAGE analysis of EndoLys-C-generated radiolabeled peptides. Immunoprecipitated, PNGase F-treated sgI and gIMC1 (see panel a) were digested with EndoLys-C (sgI and C1, respectively). Material immunoprecipitated from mock-infected cells was used as a control (m). The samples were analyzed under reducing (+ -ME) and nonreducing ( -ME) conditions in 16.5% T-6% C polyacrylamide gels, employing the discontinuous tricine-SDS-PAGE system (47). Radiolabeled peptides were visualized by fluorography. The sgI-derived peptides are indicated at the right. Peptide sizes were estimated by using the Rainbow low-molecular-weight marker (Amersham). Molecular sizes are in kilodaltons.

Characterization of a recombinant FHV expressing gIC₁. The experiments described above suggest that the C₁-C₄ bond is not essential for the formation of the gE-gI complex or for its release from the ER. However, since C₁ and C₄ are conserved in all gI proteins characterized thus far, these residues could be important for gE-gI function. To study this, an FHV recombinant, FHV-gIC₁, was constructed in which C₁ of gI was replaced by Ser. The biosynthesis of gIC₁ differed from that of wild-type gI in several respects. Under nonreducing conditions, igI_ox and mgI_ox forms were absent (Fig. 2, left panel) and igIC₁ and mgIC₁ virtually comigrated with their respective fully reduced forms (Fig. 2, right panel). Furthermore, mgIC₁ was considerably larger than mgI (85 to 115K instead of 80 to 100K; see also Fig. 6a). PNGase F treatment suggested that this size increase could be attributed to a more extensive processing of N-linked glycans. However, since PNGase F-treated mgIC₁ was still 2K larger than mgI (Fig. 2, right panel), substitution of C₁ apparently also affected O-glycosylation. The more elaborate posttranslational processing of gIC₁ was not temperature dependent and occurred also at 32 and 39°C (Fig. 6a).

View larger version (45K): [in this window] [in a new window]   FIG. 6.   Biosynthesis of gE and gI in FHV-gIC1-infected cells. (a) Comparative analysis of gIC1. Subconfluent monolayers of CRFK cells were infected at 37°C with wild-type FHV strain B927 (WT), with FHV-gIC1 (C1), or with FHVgI-LZ (gI). The cells were metabolically labeled from 7 to 8 h p.i. at 32, 37, or 39°C. Following a 2-h chase period at the respective temperatures, the cells were harvested and the cell lysates were subjected to RIPA with Ra-gI. The proteins were treated with PNGase F (+) or mock treated () and subsequently analyzed in SDS-7.5% polyacrylamide gels. The immature and mature gI forms are indicated (igI and mgI, respectively), as are the corresponding PNGase F-treated species (igI' and mgI', respectively). (b) Maturation of gE in FHV-gIC1-infected cells. The experiment was performed as described above, except that RIPA was done with the gE-specific antiserum, Ra-gE. The immature and mature gE forms are indicated (igE and mgE, respectively), as are the corresponding PNGase F-treated species (igE' and mgE', respectively). Molecular sizes are in kilodaltons.

View larger version (17K): [in this window] [in a new window]   FIG. 7.   Plaque phenotype of FHV-gIC1. A plaque assay was performed by infecting CRFK cells at 37°C with FHV B927, FHV-gIC1, or FHVgI-LZ and then incubating for 72 h at 32, 37, or 39°C. Plaques were visualized immunohistochemically, and plaque sizes were quantitated by measuring 25 randomly selected plaques along the x and y axes at a 20-fold magnification. The average plaque size in square millimeters was then calculated from the mean radius (r) by using the term r2. The histogram shows the plaque sizes relative to that of FHV B927 incubated at 37°C. Standard deviations are indicated by bars. The experiment was performed three times, yielding essentially identical results.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Alphaherpesvirus gE and gI constitute important virulence factors (8, 11, 12, 18, 27, 28, 40, 44, 48, 50, 53, 55), apparently promoting cell-to-cell transmission in mucosal and neuronal tissues (4, 10-12, 49, 60). In infected cells, gE and gI form a noncovalently linked hetero-oligomer (23, 38, 53, 54, 58, 61), and it is assumed that this complex, expressed at the cell surface, represents the functional unit. For release from the ER and transport along the exocytotic route, FHV gE must oligomerize with gI (38). Complex formation and subsequent maturation of the proteins thus provide convenient parameters by which to assess different aspects of the fate and function of either glycoprotein (37).

In the present study, we probed the disulfide-bonded structure of gI by employing single and pairwise substitutions of the four cysteine residues in the ectodomain of FHV gI. The resulting mutant proteins were coexpressed with gE in the vTF7-3 system (21), and the effect on gE-gI interaction and gE maturation was monitored. We found that gI derivatives lacking C₁ and/or C₄ assembled into a transport-competent complex with gE. In contrast, derivatives lacking C₂ and/or C₃ still bound to gE but the hetero-oligomer was retained in the ER. In addition, we took a protein-chemical approach and performed an endoproteolytic digestion of a [³⁵S]Cys-labeled secretory form of gI followed by PAGE analysis of the radiolabeled peptides under reducing and nonreducing conditions. Radiolabeled EndoLys-C-generated peptide species of 5K and 8 to 9K, predicted to contain C₁ and C₄, respectively, were found to be disulfide bonded. However, a 3K peptide, identified to contain both C₂ and C₃, was not disulfide linked to any other peptide. The biological properties of the cysteine mutants, in combination with the direct protein-chemical evidence for the C₁-C₄ bond, led us to a model for gI in which C₁ and C₄ as well as C₂ and C₃ form intramolecular disulfide bridges (Fig. 8a). Since the cysteines in the ectodomain of gI proteins have been conserved during alphaherpesvirus divergence (2, 29, 34, 39), we predict that this model applies for all gI proteins.

View larger version (31K): [in this window] [in a new window]   View larger version (7K): [in this window] [in a new window]   FIG. 8.   (a) The disulfide-bonded structure of FHV gI. Shown is a schematic model in which the gI polypeptide is represented by a thick line. The cysteine residues at positions 79, 91, 102, and 223 are indicated by C1, C2, C3, and C4, respectively. Disulfide bonds between C1 and C4 as well as between C2 and C3 are indicated by thin lines. The arrowhead indicates the position of Arg166. N-linked oligosaccharides are indicated by hexagons on sticks, and putative O-linked glycans are depicted as "lollipops." Also shown is cysteine residue C5 at position 322 in the cytoplasmic domain of gI; we speculate that this residue is a target for acylation. (b) Comparison of the disulfide linkage patterns of HSV gD (32) and FHV gI. Proteins are depicted by horizontal bars, and the transmembrane regions are indicated by hatched boxes. The C-X11-C-X8-10-C motifs in gD and gI are aligned. Disulfide bonds are indicated by brackets.

In FHV-infected cells, the disulfide-bonded gI structure is generated posttranslationally, at least in part. Newly synthesized gI is converted into an immature, EndoH-sensitive oxidized form, igI_ox. The formation of igI_ox can be followed under nonreducing conditions in SDS-polyacrylamide gels since it is accompanied by a 6K decrease in apparent molecular weight. Comparative analysis of the recombinant virus FHV-gIC₁, which carries a C₁S substitution, indicated that this shift in mobility is due to the formation of the C₁-C₄ bridge. It is of note that we did not obtain gel electrophoretic evidence of a disulfide bond between C₂ and C₃. However, closure of this bond would produce a loop of only 10 residues, which might not appreciably affect migration in SDS-polyacrylamide gels.

C₁-C₄ disulfide bond formation occurs with an estimated t_1/2 of 5 to 7 min. In comparison, the release of gI from the ER is slow, occurring with an estimated t_1/2 of 35 min. Also, gI folding appears to be an inefficient process: of the molecules synthesized during a 5-min pulse, more than 35% were still in the ER after a 120-min chase, as determined by -scanning. Most had not even acquired the C₁-C₄ bond and might represent irretrievably misfolded gI species. Only some 12% of the FHV gI acquired EndoH resistance. This estimate, however, should be regarded with caution, since mature gI may not have been quantitatively detected due to a different avidity of the Ra-gI serum for immature gI (38), to a loss of gI following incorporation into secreted virions, and to the extensive and heterogeneous posttranslational modifications which cause mgI to migrate as a diffuse smear in SDS-polyacrylamide gels. Our results are consistent with observations made for PRV (53) and bovine herpesvirus (54). In cells infected by these viruses, significant amounts of gI are also retained in the ER and not recruited into transport-competent hetero-oligomeric complexes.

Elimination of the C₁-C₄ disulfide bond, as in gIC₁, affected the posttranslational modification of gI, leading to an even more exuberant processing of N-linked glycans and possibly also to the addition of an extra O-linked oligosaccharide. These observations are not without precedent. Hyperglycosylation is also observed when gI is expressed in the absence of gE and vice versa (38, 58). Furthermore, it was recently reported that the elimination of a disulfide bond in the hemagglutinin-neuraminidase glycoprotein of Newcastle disease virus resulted in the usage of a normally inaccessible N-linked glycosylation site (35). Apparently, both the oligosaccharide chains and potential O-glycosylation sites of gIC₁ are more accessible to glycosyltransferases and other sugar-modifying enzymes than those of wild-type gI. We cannot exclude the possibility, however, that in gIC₁ an additional O-glycosylation site was created inadvertently by replacing C₁ with a Ser residue.

Analysis of FHV-gIC₁ confirmed the results of our heterologous expression studies in that C₁, and thus the formation of the C₁-C₄ bond, was found to be dispensable for intracellular transport of the gE-gI hetero-oligomer. Because disruption of disulfide bonds often results in a temperature-sensitive phenotype (31), we studied gE biosynthesis at 32, 37, and 39°C. At 32°C, the efficiency of gE maturation in FHV-gIC₁-infected cells was indeed significantly reduced compared to that in wild-type FHV-infected cells. Apparently, this defect could be overcome at the elevated temperatures. Perhaps the C₁-C₄ bond facilitates the occurrence of a thermodynamically unfavorable conformational change in gE and/or gI that is essential for ER-to-Golgi apparatus transport of the complex.

Alphaherpesviruses lacking gE and/or gI generally display a small-plaque phenotype (4, 11, 36, 37, 41, 48, 54, 60). We have previously shown that deletion of the gI gene from the FHV genome results in a 85% reduction in plaque size in CRFK cells compared to that of the parental wild-type FHV strain B927. FHV recombinants expressing mutant gI proteins produce plaques of intermediate size (37). Quantitation of plaque size thus provides an easy and useful assay by which to assess gE-gI function in vitro (33, 37, 52). Despite the low overall level of amino acid sequence identity among gI proteins of different alphaherpesvirus species, which is on the order of 20 to 30% and is mainly restricted to the N-terminal half of the ectodomain, the cysteine residues are strictly conserved (2, 29, 34, 39). It was thus anticipated that disruption of the C₁-C₄ disulfide bond would affect the function of gE-gI. Surprisingly, however, the plaques of FHV-gIC₁ were indistinguishable from those of the wild-type strain B927 at 39, 37, and 32°C. We conclude that C₁ and the formation of the C₁-C₄ bond are not essential for gE-gI-mediated cell-to-cell spread in vitro. Why, then, have C₁ and C₄ been conserved? One obvious explanation is that the loss of the C₁-C₄ bond would interfere with efficient viral spread during natural infection. Alternatively, disruption of the C₁-C₄ bond may affect the stability of the gE-gI complex, making it more vulnerable to proteolytic degradation in vivo, or may increase its antigenicity. In all of these cases, the loss of the C₁-C₄ bond would result in a decrease in viral fitness.

The presence of a characteristic cysteine motif, C-X₁₁-C-X_8-10-C, in the ectodomains of gG, gD, and gI has been interpreted to indicate that these glycoproteins have arisen through gene duplication (34). The ectodomain of gD contains six cysteine residues, with C₂, C₃, and C₄ apparently corresponding to C₁, C₂, and C₃ of gI (34). For gD of HSV types 1 and 2, the disulfide-bonded structure has been resolved. Disulfide bridges are formed between C₁ and C₅, C₂ and C₆, and C₃ and C₄ (Fig. 8b) (32). As predicted by Long et al. (32), the disulfide-bonded structure of gI, as determined in this paper, conforms to that of gD. Interestingly, of the three disulfide bridges in HSV gD, the C₁-C₅ bondi.e., the one apparently absent from gIis the least important for function (32). Our results are consistent with McGeoch's hypothesis that gD and gI are derived from a common ancestral protein (34). However, the question of whether these proteins are indeed evolutionarily related can be answered only by a comparison of their three-dimensional structures.