The Disulfide-Bonded Structure of Feline Herpesvirus Glycoprotein I

Abstract

Alphaherpesvirus glycoproteins E and I (gE and gI, respectively) assemble into a hetero-oligomeric complex which promotes cell-to-cell transmission, a determining factor of virulence. Focusing on gI of feline herpesvirus (FHV), we examined the role of disulfide bonds during its biosynthesis, its interaction with gE, and gE-gI-mediated spread of the infection in vitro. The protein’s disulfide linkage pattern was determined by single and pairwise substitutions for the four conserved cysteine residues in the ectodomain. The resulting mutants were coexpressed with gE in the vaccinia virus-based vTF7-3 system, and the formation and endoplasmic reticulum (ER)-to-Golgi transport of the hetero-oligomeric complex were monitored. The results were corroborated biochemically by performing an endoproteinase Lys-C digestion of a [³⁵S]Cys-labeled secretory recombinant form of gI followed by tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis of the peptides under reducing and nonreducing conditions. We found that (i) gI derivatives lacking Cys⁷⁹ (C₁) and/or Cys²²³ (C₄) still assemble with gE into transport-competent complexes, (ii) mutant proteins lacking Cys⁹¹ (C₂) and/or Cys¹⁰² (C₃) bind to gE but are retained in the ER, (iii) radiolabeled endoproteinase Lys-C-generated peptide species containing C₁ and C₄ are linked through disulfide bonds, and (iv) peptides containing both C₂ and C₃ are not disulfide linked to any other peptide. From these findings emerges a model in which C₁ and C₄ as well as C₂ and C₃ form intramolecular disulfide bridges. Since the cysteines in the ectodomain have been conserved during alphaherpesvirus divergence, we postulate that the model applies for all gI proteins. Analysis of an FHV recombinant with a C₁→S substitution confirmed that the C₁-C₄ disulfide bond is not essential for the formation of a transport-competent gE-gI complex. The mutation affected the posttranslational modification of gI and caused a slight cold-sensitivity defect in the assembly or the intracellular transport of the gE-gI complex but did not affect plaque size. Thus, C₁ and the C₁-C₄ bond are not essential for gE-gI-mediated cell-to-cell spread, at least not in vitro.

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 Cys→Ser substitutions on gE-gI complex formation and intracellular transport and on gE-gI-mediated cell-to-cell spread are discussed.

MATERIALS AND METHODS

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-αgE) and gI (Ra-αgI), the cat antiserum against FHV (Cat-αFHV), and the plasmids pBS-gE, pBS-gI, pBS-gIΔM, 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-gIΔM.

pBS-gIΔM 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 G²³⁶→C substitution. The 5′ and 3′ halves of the gIΔM 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 gIΔMΔC₁ 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 gIΔM gene within the polylinker region. The 292-bp PmlI-HindIII fragment was gel purified and exchanged for the corresponding sequences in pBS-gIΔM, yielding pBS-gIΔMΔC₁. Mutagenesis of Cys⁹¹ (C₂) and Cys¹⁰² (C₃) was performed accordingly, with primer pairs 582-581 and 604-603, respectively. To obtain pBS-gIΔMΔC₂ and pBS-gIΔMΔC₃, 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 gIΔM 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-gIΔM, yielding pBS-gIΔMΔC₂ and pBS-gIΔMΔC₃, respectively. pBS-gIΔMΔC₂₃ was constructed by inserting the PmlI-HindIII fragment of pBS-gIΔMΔC₂ into PmlI- and HindIII-digested pBS-gIΔMΔC₃. To construct pBS-gIΔMC₀, C₁ in pBS-gIΔMΔC₂₃ 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.

TABLE 1.

Oligonucleotide primers used in these studies

Primer	Nucleotide sequence (5′ to 3′)	Gene	Positionsa	Polarity
569	TCTTCTTCCTATAAAATCGTTCAAGTA	gI	229–255	+
570	TTTATAGGAAGAAGAGTGATGGTTGTA	gI	217–243	−
581	TCATCATCTCCACGTGTACGCAATAAT	gI	265–291	+
582	ACGTGGAGATGATGAATATTCTATTAC	gI	253–279	−
603	CGGTCCTCTCTCCACAAGACCTCTATG	gI	298–324	+
604	GTGGAGAGAGGACCGGAAAGCATTATT	gI	286–312	−
630	TTAACGTCTCGCCCACGTTAT	gE	1168–1188	−
633	TTAGTGGTGGTGGTGGTGGTGCTTTCTGTCTGGTGTCCA	gI	850–867	−

^aNumerical positions on the genome of FHV B927 as counted from the initiation codon of gE or gI. 

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.

pBS-sgI was made as follows. pBS-gI (38) served as a template for PCR with the M13 universal primer and primer 633 (Table 1). The PCR product was blunt ended and subsequently cut with MluI. Thus, a 393-bp fragment was generated, which was gel purified and ligated to the 2.6-kb EcoRI (blunt)-MluI fragment of pBS-gI. As a result of the cloning procedure, nt 869 through 1155 of the gI gene were deleted and a termination codon was created downstream of the codon for Lys²⁸⁹. Sequence analysis of pBS-sgE and pBS-sgI confirmed that no inadvertent mutations had occurred during PCR and cloning.

Construction of recombinant virus FHV-gIΔC₁.

FHVΔgI-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 gIΔC₁, the Cys⁷⁹→Ser 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-gIΔMΔC₁ was first inserted into PmlI- and HindIII-digested pBS-gI, resulting in pBS-gIΔC₁. Subsequently, the XhoI-BamHI fragment from this plasmid was exchanged with the corresponding region in pUS1, yielding the transfer vector pUS1-gIΔC₁. Sequence analysis of the relevant region confirmed that pUS1-gIΔC₁ had the correct nucleotide substitution and that no inadvertent mutations had been introduced during the cloning procedures.

To generate recombinant FHV-gIΔC₁, 10⁶ CRFK cells, seeded in 35-mm-diameter dishes, were cotransfected with approximately 50 ng of FHVΔgI-LZ DNA and 1 μg of pUS1-gIΔC₁. The culture supernatants were harvested 7 days later, and plaque assays were performed. Recombinant viruses that had lost the encephalomyocarditis virus internal ribosomal entry site-lacZ expression cassette were identified by staining of plaques with X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside; Boehringer Mannheim) as a substrate and were plaque purified three times prior to the preparation of virus stocks. Proper introduction of the mutation was confirmed by Southern blot hybridization and sequence analysis of viral DNA.

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-αFHV and Ra-αgE, 3 μl each; Ra-αgI, 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% β-mercaptoethanol (β-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-gIΔMΔC₁. 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-αgI 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% β-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

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 FHVΔgI-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-αgI) (38), and the precipitated proteins were separated in SDS–7.5% polyacrylamide gels in the presence or absence of β-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.

FIG. 1.

Schematic representation of FHV gI, gIΔM, and the various gIΔM derivatives carrying Cys→Ser 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.

FIG. 2.

Analysis of gI and gIΔC₁ under reducing and nonreducing conditions. Subconfluent monolayers of CRFK cells were infected with either wild-type FHV strain B927 (B927) or FHV-gIΔC₁ (gIΔC₁). Cells infected with the gI-deficient recombinant FHVΔgI-LZ (ΔgI) (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-αgI. 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 β-ME (− β-ME and + β-ME, respectively). The closed arrowheads indicate immature reduced and oxidized gI species (igI_red and igI_ox, respectively) and mature reduced and oxidized gI species (mgI_red and mgI_ox, respectively). In the left-hand panel, mature gIΔC₁ is indicated by an open arrowhead. Molecular sizes are in kilodaltons.

To facilitate the analysis of mgI, the immunoprecipitates were treated with PNGase F, an amidase which cleaves both high-mannose and hybrid as well as complex N-linked oligosaccharides. Reduced igI was trimmed to 46 kDa, the expected size of the protein backbone, while reduced mgI was trimmed to yield a prominent 58-kDa species plus a number of minor products ranging from 56 to 62 kDa (Fig. 2, right panel, B927 +). The size difference between the PNGase F-treated igI and mgI species is indicative of extensive post-ER modifications, which include O-glycosylation (38). The apparent size variation among PNGase F-treated mgI species presumably reflects heterogeneity in the number or composition of the O-glycans.

Analysis of the PNGase F-treated samples under nonreducing conditions revealed three major products (Fig. 2, left panel, B927 +). Two of these are igI species, one virtually comigrating with the fully reduced protein (igI_red) and the other a faster-migrating oxidized form (igI_ox) running at 38K. The third product, migrating at 52K, represents oxidized mgI (mgI_ox). We did not detect products comigrating with fully reduced 58K mgI (mgI_red). A minor 41K igI species was consistently found, but at present it is not clear whether it is a bona fide folding intermediate or a misfolded, aberrantly oxidized form. The data indicate that gI acquires at least one intramolecular disulfide bond. gI oligomers linked through intermolecular disulfide bonds were not detected.

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 β-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.

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 FHVΔgI-LZ (ΔgI) 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-αgI. 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 β-ME. Immature reduced and immature oxidized gI species (igI_red and igI_ox, respectively) and mature reduced and mature oxidized gI species (mgI_red and mgI_ox, 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, gIΔM, 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, gIΔM 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-αFHV. 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-gIΔM hetero-oligomer (37). As shown in Fig. 4, substitution of C₁, as in gIΔMΔC₁, did not affect gE maturation. Thus, C₁, like C₄, is not essential for this process. In contrast, gIΔMΔC₂, gIΔMΔC₃, and gIΔMΔC₂₃, in which C₂ and/or C₃ had been replaced, and gIΔMC₀, which lacks all cysteine residues, failed to induce maturation of gE. Immunoprecipitation with the Ra-αgI antiserum confirmed that the gIΔM derivatives had been expressed to similar extents. Moreover, immature gE was found to coprecipitate with gIΔMΔC₂, gIΔMΔC₃, and gIΔMΔC₂₃, while in the reciprocal experiment, with a monospecific antiserum against gE, coprecipitation of the gIΔM derivatives was observed (data not shown). Apparently, replacement of C₂ and/or C₃ does not prevent the formation of the gE-gIΔM 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.

FIG. 4.

Coexpression of gE with cysteine mutants of gIΔM. vTF7-3-infected OST7-1 cells were cotransfected with plasmid pBS-gE and a plasmid encoding gIΔM or one of its derivatives (gIΔMΔC₁, -ΔC₂, -ΔC₃, -ΔC₂₃, 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-αFHV. 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 gIΔMΔC₁ 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-αFHV and Ra-αgI 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-αFHV serum predominantly detected mature sgE whereas the Ra-αgI serum readily precipitated both sgI and the 17K gIΔMΔC₁ 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).

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 gIΔMΔC₁ and gE (ΔC₁) 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-αFHV (CαFHV) or Ra-αgI (RaαgI). 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 β-ME. Of each Ra-αgI precipitate, only one-quarter was analyzed; the remainder was digested with EndoLys-C (see below). Mature reduced and oxidized sgI (sgI_red and sgI_ox, respectively), mature sgE (sgE), and gIΔMΔC₁ 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 gIΔMΔC₁ (see panel a) were digested with EndoLys-C (sgI and ΔC₁, respectively). Material immunoprecipitated from mock-infected cells was used as a control (m). The samples were analyzed under reducing (+ β-ME) and nonreducing (− β-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.

Immunoprecipitated PNGase F-treated sgI and gIΔMΔC₁ were digested with EndoLys-C. In the case of sgI, this should yield radiolabeled peptides of 63, 24, and 50 residues, the latter containing two potential O-glycosylation sites (Fig. 5b). Glycosylation at these sites would result in an increase in apparent molecular weight of approximately 4,000, and the product would be expected to run either as a diffuse smear or as a set of multiple bands in SDS-polyacrylamide gels. Digestion of gIΔMΔC₁ should yield only the 24-residue peptide containing C₂ and C₃. The sgI and gIΔMΔC₁ digests were separated in polyacrylamide gels (16.5% T, 6% C), employing the discontinuous tricine-SDS-PAGE system developed by Schägger and von Jagow (47), and radiolabeled peptides were visualized by fluorography. As shown in Fig. 5c, analysis of the sgI digest under reducing conditions yielded two distinct products of 3K and 5.5K plus a family of products migrating at 8 to 9K. The 5.5K and 8 to 9K peptides were absent from the mock-transfected cells and the gIΔMΔC₁ digest, but the latter did contain the 3K peptide. This product was thus identified as the 24-residue fragment containing C₂ and C₃. The 5.5K and 8 to 9K products apparently represent the unglycosylated 63-residue fragment containing C₁ and the O-glycosylated 50-residue fragment containing C₄, respectively. Under nonreducing conditions, the 3K peptide did not show a shift up in apparent molecular weight, indicating that it is not disulfide linked to other peptides (Fig. 5c, right panel, sgI). The 5.5 and 8K to 9K products, however, appear to be disulfide bonded, since they both were absent under nonreducing conditions and were replaced by a 16K product. These findings indicate that residues C₁ and C₄ form an intramolecular disulfide bridge.

Under both reducing and nonreducing conditions, a minor 8K peptide was detected in the gIΔMΔC₁ digest. This may represent a partial digestion product or, more likely, a digestion product of the high-molecular-weight polypeptide contaminating the gIΔMΔC₁ immunoprecipitate (Fig. 5a).

Characterization of a recombinant FHV expressing gIΔC₁.

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-gIΔC₁, was constructed in which C₁ of gI was replaced by Ser. The biosynthesis of gIΔC₁ 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 igIΔC₁ and mgIΔC₁ virtually comigrated with their respective fully reduced forms (Fig. 2, right panel). Furthermore, mgIΔC₁ 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 mgIΔC₁ was still 2K larger than mgI (Fig. 2, right panel), substitution of C₁ apparently also affected O-glycosylation. The more elaborate posttranslational processing of gIΔC₁ was not temperature dependent and occurred also at 32 and 39°C (Fig. 6a).

FIG. 6.

Biosynthesis of gE and gI in FHV-gIΔC₁-infected cells. (a) Comparative analysis of gIΔC₁. Subconfluent monolayers of CRFK cells were infected at 37°C with wild-type FHV strain B927 (WT), with FHV-gIΔC₁ (ΔC₁), or with FHVΔgI-LZ (ΔgI). 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-αgI. 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-gIΔC₁-infected cells. The experiment was performed as described above, except that RIPA was done with the gE-specific antiserum, Ra-αgE. 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.

As anticipated, gE still matured in FHV-gIΔC₁-infected cells, but maturation was less efficient than in cells infected with wild-type FHV strain B927 (Fig. 6b). Interestingly, this effect was most pronounced at 32°C. Apparently, the C₁→S substitution in gI results in a conditional cold-sensitivity defect affecting either the formation or subsequent ER-to-Golgi transport of gE-gI.

To determine whether substitution of C₁ affected the function of gE-gI, the plaque size of FHV gIΔC₁ was compared to that of the wild-type virus. The gI-deficient recombinant FHVΔgI-LZ served as a control. Plaque assays were performed in CRFK cells at 32, 37, and 39°C. Plaques were visualized immunohistochemically, and average plaque sizes were determined in square millimeters (Fig. 7). Whereas FHVΔgI-LZ displayed a small-plaque phenotype at all three temperatures, the average plaque size of FHV-gIΔC₁ equaled that of the wild-type strain B927. Apparently, elimination of the C₁-C₄ bond does not affect the function of gE-gI in vitro.

FIG. 7.

Plaque phenotype of FHV-gIΔC₁. A plaque assay was performed by infecting CRFK cells at 37°C with FHV B927, FHV-gIΔC₁, or FHVΔgI-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 πr². 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

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.

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-X_8–10-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-gIΔC₁, which carries a C₁→S 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 β-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-αgI 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 gIΔC₁, 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 gIΔC₁ are more accessible to glycosyltransferases and other sugar-modifying enzymes than those of wild-type gI. We cannot exclude the possibility, however, that in gIΔC₁ an additional O-glycosylation site was created inadvertently by replacing C₁ with a Ser residue.

Analysis of FHV-gIΔC₁ 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-gIΔC₁-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-gIΔC₁ 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 bond—i.e., the one apparently absent from gI—is 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.