Analysis of Epstein-Barr Virus Glycoprotein B Functional Domains via Linker Insertion Mutagenesis

Abstract

Epstein-Barr Virus (EBV) glycoprotein B (gB) is essential for viral fusion events with epithelial and B cells. This glycoprotein has been studied extensively in other herpesvirus family members, but functional domains outside of the cytoplasmic tail have not been characterized in EBV gB. In this study, a total of 28 linker insertion mutations were generated throughout the length of gB. In general, the linker insertions did not disrupt intracellular expression and variably altered cell surface expression. Oligomerization was disrupted by insertions located between residues 561 and 620, indicating the location of a potential site of oligomer contacts between EBV gB monomers. In addition, a novel N-glycosylated form of wild-type gB was identified under nonreducing Western blot conditions that likely represents a mature form of the protein. Fusion activity was abolished in all but three variants containing mutations in the N-terminal region (gB30), within the ectodomain (gB421), and in the intracellular C-terminal domain (gB832) of the protein. Fusion activity with variants gB421 and gB832 was comparable to that of the wild type with epithelial and B cells, and only these two mutants, but not gB30, were able to complement gB-null virus and subsequently function in virus entry. The mutant gB30 exhibited a low level of fusion activity with B cells and was unable to complement gB-null virus. The mutations generated here indicate important structural domains, as well as regions important for function in fusion, within EBV gB.

Epstein-Barr virus (EBV) is a gammaherpesvirus that infects up to 90% of the human population by adulthood (32). Asymptomatic infections generally occur during childhood, whereas adolescent infections may result in the development of infectious mononucleosis. After the primary infection is resolved by the host immune system, the virus establishes a latent state, and the host remains an infected carrier for life. EBV is implicated in malignancies of epithelial and lymphoid origin, including nasopharyngeal carcinoma, Hodgkin's disease, and Burkitt's lymphoma (45, 71, 74). Roles for EBV in the development of gastric carcinoma and multiple sclerosis have also been suggested (20, 32, 45, 70, 71, 74).

The two main target cell types of EBV infection are epithelial cells and B lymphocytes. Fusion between the viral envelope and the cellular membrane is an essential event in the entry of all herpesviruses (25). For EBV, the viral glycoproteins necessary for fusion of the viral envelope with the target cell are glycoprotein B (gB), the complex of gH and gL (gH/gL), and gp42 (65). There are two complexes of glycoproteins that form the fusion machinery, depending on which cell type is infected. Entry of EBV into B cells is mediated by gB, gH/gL, and gp42, whereas fusion with epithelial cells requires gB and gH/gL only (4, 48). Entry of virus into B cells occurs after the initial binding of gp350/220, the major EBV envelope glycoprotein, to CD21 expressed on the surfaces of B cells (29). This interaction increases the efficiency of infection but is not required for cell-based fusion assays (30). gp42 then plays two roles in the entry process: it binds to its receptor major histocompatibility complex type II expressed on B cells and also facilitates membrane fusion, likely through interaction with the gH/gL complex (21, 41, 49, 63, 67, 72). Less is known about the sequence of events during epithelial cell fusion, since no epithelial cell receptors for EBV have been identified. It has been hypothesized that a receptor does exist, as a soluble gH/gL complex was shown to bind to the epithelial cell surface (5). There is also strong evidence that a switch in fusion complex components occurs based on the cell type in which EBV replication occurs. EBV produced in epithelial cells is enriched for gp42 and therefore infects B cells more efficiently, whereas virus produced in B cells has low levels of gp42 and is better at infecting epithelial cells (4).

The viral envelope glycoprotein gB is conserved throughout the herpesvirus family. EBV gB has been shown to be essential for virus-cell fusion events and egress of virions from infected cells (18, 22, 27, 39, 40, 48). In cells undergoing lytic replication, EBV gB is primarily localized to the perinuclear membrane and the endoplasmic reticulum and is found in reduced amounts on the cell surface (17, 18, 39, 40, 52). EBV gB is also found in the virion envelope; moreover, the levels of gB expressed in the virion envelope vary between EBV strains and influence infectivity (18, 31, 52). EBV virions that contain high levels of gB are more infectious than those with low levels of gB (52).

Cleavage of EBV gB occurs during protein maturation, and two fragments of 78 kDa and 58 kDa are found in mature virions, as well as in recombinant protein expressed in insect cells (2, 18, 31). Cleavage of gB has been demonstrated in several herpesviruses, including bovine herpesvirus and human cytomegalovirus (HCMV); however, elimination of the cleavage site appears to have no functional consequences (37, 68). The relevance of cleavage in EBV gB function is not known.

Knowledge of the functionally important EBV gB domains is limited. To investigate the importance of different regions for the fusion activity of EBV gB, we generated a panel of 28 random linker insertion variants so that the insertions were spread throughout the length of the protein. All of these mutants were expressed in CHO-K1 cells, and 24 were detected at the cell surface. Next, the abilities of these mutants to form oligomers were assessed. We also observed a form of gB with a larger molecular weight than the monomeric form of gB when sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed under nonreducing conditions. This form of EBV gB is detectable in EBV-infected cells, as well as epithelial and B-cell lines transfected with gB. We analyzed the glycosylation state of this higher molecular weight form and determined that it is an N-glycosylated form of gB. The abilities of the gB linker insertion mutants to mediate fusion in the presence and absence of gH/gL were tested using a cell-cell fusion assay with B cells and epithelial cells. Finally, we assessed three gB linker insertion mutants that function in the cell-based fusion assay in a viral entry assay. Complementation of a gB-null lymphoblastoid cell line (LCL) by these mutants was utilized, with the endpoint of the assay being transformation of primary B cells. Of these three mutants, two were functional in the viral entry assay.

MATERIALS AND METHODS

Generation of gB mutants by random linker insertion mutagenesis.

The GPS-LS Linker Scanning System from New England Biolabs was utilized to generate random insertions of 15 bp, which resulted in the insertion of 5 amino acids. The insertion generated a unique PmeI restriction site, which was used for diagnostic purposes. The target plasmid for the random insertions contained full-length gB in pSG5 (Stratagene) (27). The linker insertion contained the PmeI site; the gB gene had an NheI site, whereas the XbaI site was present within the pSG5 vector plasmid. Insertions in the gB plasmid were identified by restriction enzyme digestion and subsequently sequenced (Northwestern University Biotechnology Core Facility). Linker insertion mutations that resulted in a stop codon were discarded. The mutants were named based on the site of insertion in the amino acid sequence. Table 1 provides a summary of this information, as well as the amino acid sequence inserted by each linker.

TABLE 1.

Summary of EBV gB linker insertion mutants^a

Mutant nomenclature	Site(s) of mutation (aa)b	Insertion sequencec
gB	None	None
gB30	30-31	AP-CLNTP-PA
gB88	88-89	DN-CLNNN-II
gB115	115-116	AD-LFKQD-SV
gB190	190-191	DA-LFKHA-PG
gB198	198-199	TY-CLNTY-RT
gB208	208-209	CL-MFKHL-IT
gB236	236-237	FY-CLNIY-DG
gB263	263-264	YD-MFKHD-NR
gB307	307-308	TI-CLNTI-AT
gB320	320-321	TD-CLNTD-EG
gB324	324-325	TS-CLNTS-SF
gB353	353-356	HE-CLNNE-KY
gB361	361-362	DR-LFKHR-YT
gB380	380-381	LL-VFKQL-AW
gB421	421-422	GS-MFKQS-TP
gB486	486-487	LE-CLNME-QK
gB561	561-562	FS-CLNIS-FI
gB597	597-598	YF-LFKHF-QS
gB602	602-603	EI-LFKQI-HV
gB620	620-621	IA-CLNIA-TL
gB632	632-633	SL-MFKHL-IE
gB646	646-647	SR-CLNTR-DE
gB675	675-676	LR-MFKQR-KD
gB703	703-704	SV-FKHSV-GQ
gB717	717-718	GL-LFKHL-FS
gB734	734-735	FG-CLNTG-GM
gB802	801-802	AL-LFKQL-HE
gB832	832-833	RF-LFKHF-PG
gBΔ798	Truncated at 798	Noned

^aThe nomenclature of mutants used throughout the paper was determined by the first amino acid prior to the site of linker insertion. All mutants were sequenced at the Northwestern University sequencing facility, and the amino acid composition of the insertion was confirmed.

^baa, amino acids.

^cThe 2 amino acids before and after the hyphens are the wild-type EBV gB sequence that flanks the linker insertion. A unique PmeI restriction site was introduced into all gB linker insertion mutants and was used for diagnostic purposes.

^dThe gBΔ798 mutant possesses a unique BclI restriction site.

Cells and antibodies.

Chinese hamster ovary (CHO-K1) cells were cultured in Ham's F-12 medium (BioWhittaker) supplemented with 10% Fetalplex serum (Gemini Bioproducts) and 1% penicillin-streptomycin (BioWhittaker). 293T epithelial cells (ATCC) stably expressing T7 RNA polymerase were cultured in Dulbecco's modified Eagle's medium (BioWhittaker) supplemented with 10% Fetalplex serum, 1% penicillin-streptomycin, and 100 μg/ml zeocin (54). Daudi B lymphocytes expressing T7 RNA polymerase, which also express HLA class II and CD21, were cultured in RPMI 1640 medium (BioWhittaker) supplemented with 10% Fetalplex, 1% penicillin-streptomycin, and 900 μg/ml G418 (63).

Polyclonal gB antibody was generated by genetic immunization of rabbits with EBV gB expression vectors (Aldevron). The monoclonal gB antibody CL55 was generously provided by Lindsay Hutt-Fletcher (Louisiana State Health Sciences Center, Shreveport, LA). Polyclonal and monoclonal actin antibodies were purchased from Sigma.

Anti-rabbit and anti-mouse immunoglobulin G (IgG) conjugated to horseradish peroxidase (HRP) were purchased from Cell Signaling Technologies. Akt antibody was obtained from Cell Signaling Technologies. Bio-Rad avidin conjugated to HRP was utilized for biotinylation experiments. Ultralink immobilized neutravidin from Pierce and protein G-Sepharose from GE Healthcare were used for immunoprecipitation.

Transfection.

The Lipofectamine 2000 transfection reagent protocol was followed for all transfections (Invitrogen). For fusion assays and biotinylation and oligomerization experiments, CHO-K1 cells were seeded into six-well plates 18 h prior to transfection so that they were 60 to 70% confluent upon transfection. For fusion assays, cells were transfected with 0.8 μg of gB, 0.5 μg of gH and gL, 0.85 μg of gp42, and 0.8 μg of luciferase reporter plasmid. Cells to be used for biotinylation and oligomer formation studies were transfected with 0.8 μg of gB alone. Six hours after transfection, the cells were washed once with warm phosphate-buffered saline (PBS), and serum-containing Ham's F-12 medium was added.

For intracellular expression analysis and enzymatic digestion, cells were seeded into T25 flasks at a lower density so that they were 60 to 70% confluent upon transfection. The cells were transfected using the same Lipofectamine protocol with the only difference being the use of 10 μg of gB and no other glycoproteins. The cells were washed 6 h after transfection, and serum-containing Ham's F-12 medium was added to each flask.

SDS-PAGE and Western blot analysis.

CHO-K1 cells were transfected as described above; 24 h after transfection, the cells were washed once with Ham's F-12 medium and twice with PBS. The cells were detached with 1 mM EDTA in PBS, pelleted by centrifugation, washed again with PBS, and counted using a Z1 Coulter particle counter (Beckman Coulter) so that samples contained a concentration of 10 million cells per ml. The samples were lysed with 1% Triton X-100 lysis buffer for 30 min and centrifuged at 4°C to remove cellular debris (54, 63). For intracellular expression determination, SDS buffer containing β-mercaptoethanol was added. For analysis of oligomer formation, SDS buffer lacking β-mercaptoethanol was used, and in both instances, the samples were boiled for 10 min at 70°C. The samples were then separated on 7.5% Criterion SDS-PAGE gels (Bio-Rad), transferred to Millipore Immobilon-P membranes, and blocked in 5% milk in Tris-buffered saline with Tween (TBST) for 30 min at room temperature. The membranes were then probed with polyclonal gB antibody or Akt antibody diluted at 1:1,000 in 5% milk-TBST at 4°C overnight, washed five times with TBST, and probed for 30 min with anti-rabbit or anti-mouse IgG conjugated to HRP diluted at 1:2,500 and 1:3,000, respectively, in 5% milk-TBST. After five more TBST washes, enhanced-chemiluminescence reagents were added, and the blots were exposed to Hyperfilm (Amersham Biosciences).

Biotinylation, immunoprecipitation, and Western blotting of biotinylated samples.

The biotinylation protocol was modified slightly from those used previously (54). As mentioned above, CHO-K1 cells were plated for transfection in six-well plates, washed with PBS, and then detached with 1 mM EDTA in PBS 24 h after transfection. The time of harvest for biotinylation matched the time that the cell-cell fusion results were read, as described below. After being detached, the cells were washed twice with ice-cold PBS and counted using the Z1 Coulter particle counter; 2.5 × 10⁷ cells/ml were used for biotinylation. The cells were incubated with a final concentration of 2 mM EZ-Link Sulfo-NHS-LC-Biotin (Pierce) and rotated at 4°C for 30 min. To stop the reaction, the cells were then washed three times with PBS containing 100 mM glycine. The cells were pelleted and resuspended in 1% Triton X-100 lysis buffer so that each sample had 10 million cells/ml, as described for SDS-PAGE above. The cells were lysed for 30 min, centrifuged to remove cell debris, and then added to beads for immunoprecipitation. Protein G-Sepharose beads were loaded with CL55 gB antibody or polyclonal actin antibody to precipitate gB and actin. One microgram of antibody was added to 20 μl of beads diluted in 40 μl of PBS. For immunoprecipitation with polyclonal gB, 2 μl of antibody was used per sample. The lysates were rotated with the beads overnight at 4°C and washed four times the following day with 1 ml of Triton X-100 lysis buffer. The bound proteins were then eluted off of the beads by adding SDS buffer without β-mercaptoethanol and boiling the beads for 10 min at 70°C. Separation of these proteins was done on 7.5% Criterion gels, which were transferred to Immobilon-P membranes and probed with 1:2,500 avidin-HRP in 5% milk-TBST. The membranes were stripped and reprobed to detect the presence of actin with a monoclonal actin antibody. Alternatively, neutravidin beads were used to capture all biotinylated proteins. The same immunoprecipitation protocol was followed, minus the addition of antibody, and following incubation overnight, the samples were washed with Triton X-100 lysis buffer supplemented with 0.35 M NaCl. Bound proteins were again eluted from the beads by adding SDS buffer without β-mercaptoethanol and boiled for 10 min at 70°C. SDS-PAGE on these samples was performed as described above. The membranes were probed with either 1:1,000 polyclonal gB in 5% milk-TBST or 1:1,000 monoclonal actin antibody in 5% milk-TBST overnight and then incubated with secondary antibodies and developed as described above.

Induction of EBV-GFP⁺ Akata cells.

EBV-GFP⁺ Akata cells (EBV-TKdel; obtained from Lindsey Hutt-Fletcher, Lousiana State University) (50) were induced to produce virus by incubating 2 × 10⁶ cells with 8 μl of human anti-IgG (MP Biomedicals) for 3 h at 37°C. Following incubation, the cells were washed twice with 2 ml of fresh RPMI minus G418 and then resuspended in 2 ml of RPMI supplemented with 500 mg/ml G418. The induced Akata cells were then grown in six-well plates for 24 h, harvested, and lysed for SDS-PAGE analysis as described for CHO-K1 cells under nonreducing conditions.

Removal of N- and O-linked sugars.

CHO-K1 cells were transfected with pSG5 vector or wild-type gB plasmid as described above. The cells were harvested as for the SDS-PAGE and Western blotting experiments, and the lysates were incubated with CL55 gB antibody bound to protein G beads for 3 h at 4°C. For endoglycosidase H (EndoH) (New England BioLabs) treatment, 5× SDS loading buffer without a reducing agent was added to the beads, and then the beads were boiled for 10 min at 70°C. After being boiled, the beads and lysates were spun down, and the liquid was removed and divided into separate tubes. The lysates were boiled at 100°C for 10 min with and without denaturing buffer according to the kit instructions; 10× G5 reaction buffer was added to each sample, and either water or 2 μl of EndoH was also added to each reaction mixture. Samples were then incubated at 37°C for 1 h. Western blot analysis followed, using 7.5% Criterion gels as described above. The blots were probed with polyclonal anti-gB antibody. Further glycosylation removal was done using the E-Degly kit (Sigma), which contains PNGase F and O-glycosidase. Whole-cell lysate was immunoprecipitated with CL55 anti-gB antibody as described previously. Protein was then eluted off the beads with 0.2 M glycine at pH 2.5 for 5 min. These eluates were then treated under native or denaturing conditions as suggested by the kit protocol. In short, under denaturing conditions, 5× reaction buffer was added to the eluted protein, along with denaturation solution, and heated for 5 min at 100°C. These samples were then cooled to room temperature, and Triton X-100 solution was added. One microliter of each enzyme provided or 1 μl of water was added to each sample, and the mixture was incubated for 3 h at 37°C. For native digestion, 5× reaction buffer and 1 μl of enzyme or water was added to each sample. The samples were then incubated at 37°C for 5 days. After incubation at 37°C, 5× SDS buffer lacking β-mercaptoethanol was added, and SDS-PAGE was performed as described above.

Cell-cell fusion assay.

Cells were plated and transfected as described above; 6 h after transfection, when the medium would be changed for cells being utilized for SDS-PAGE, effector CHO-K1 cells were mixed with the target cell population. The epithelial target cells used were 293T cells, which stably express T7 polymerase (54). CHO-K1 and 293T cells were washed with PBS, detached with 1 mM EDTA in PBS, and counted with the Coulter Z1 particle counter. The target B cells used were Daudi cells, which also stably express T7 polymerase, and they were also counted using the cell counter (63). A total of 0.4 × 10⁶ cells (0.2 × 10⁶ of both the target and effector cell types) were mixed together and plated in a 24-well plate in Ham's F-12 medium (22, 48, 54). Eighteen hours after being overlaid, the cells were washed with PBS and lysed with 100 μl of passive lysis buffer (Promega) at room temperature. Luciferase activity was measured in triplicate samples by the addition of 100 μl of Promega luciferase assay reagent on a Perkin-Elmer Victor plate reader.

Complementation of gB-null LCLs.

The gB-null LCL M.2, which has a hygromycin cassette inserted into the gB reading frame, has been previously characterized (27, 40). We utilized this cell line and the gB-containing LCL10 line for the complementation assay. Cells were counted using trypan blue exclusion and were nucleofected when they reached the log phase of growth (between 600,000 and 800,000 cells/ml). The Amaxa Nucleofector II was utilized to nucleofect the cells according to the manufacturer's instructions. Five million LCLs resuspended in Amaxa solution C were mixed with 4 μg of total DNA (2 μg of BZLF1 and 2 μg of pSG5 vector, gB, or gB linker insertion mutant). Two samples were nucleofected per mutant using program X-001. Peripheral blood mononucleocytes (PBMCs) were isolated from whole blood (LifeSource) using a Ficoll gradient. Two days after nucleofection, the LCLs were irradiated using a gamma cell irradiator at 7,000 rad. The irradiated LCLs were then cocultivated with PBMCs; 50,000 irradiated LCLs were mixed with 100,000 PBMCs in 150 μl of RPMI containing a 1:10,000 dilution of cyclosporine A (Sigma) and plated in a 96-well plate. The cocultured cells were then grown for 8 weeks and analyzed for the formation of foci of transformed B cells. The transformation results were confirmed by PCR amplification of the hygromycin cassette inserted into the gB open reading frame of M.2 LCLs as previously described (27).

RESULTS

Generation and expression of EBV gB linker insertion mutants.

A total of 28 EBV gB linker insertion mutants containing 5-amino-acid insertions were generated using the GPS-LS Linker Scanning System as described in Materials and Methods. The results of the sequencing, as well as the nomenclature of the mutants that were used throughout this study, are shown in Table 1. Figure 1, top, depicts the locations of these mutations in the EBV gB amino acid sequence, as well as the locations of potentially important structural and functional regions identified by the PredictProtein program (62).

FIG. 1.

EBV gB amino acid sequence and locations of EBV gB linker insertion mutations on the HSV-1 gB crystal structure. (Top) The 857-amino-acid EBV gB sequence was obtained from the B95-8 genome (accession number YP_401713). Potentially important structural and functional domains are illustrated, as determined by the PredictProtein program. The 22-amino-acid signal sequence is highlighted in gray. N-linked glycosylation sites are represented by green lettering, and conserved cysteine residues are depicted by underlined pink letters. The fusion loops identified in reference 1 are shown with red underlined letters. The putative furin cleavage site is highlighted in magenta, the amphipathic helices are highlighted in green, and the transmembrane domain is represented by yellow highlighting. The probable coiled-coil regions are shown as underlined blue lettering. In the C-terminal tail, the four-arginine endoplasmic reticulum retention signal is identified with underlined orange letters. The sites of linker insertion mutants are identified with orange arrows. The lime green arrow indicates the site of the Δ798 truncation. Category I mutants are identified by normal text, category II mutants are distinguished by underlined text, and category III mutants are identified in boldface letters. (B) EBV gB and HSV-1 gB amino acid sequences were aligned using the MultiAlin program (http://bioinfo.genotoul.fr/multalin/multalin.html), and linker insertion mutants were located in the HSV-1 gB amino acid sequence and mapped onto the HSV-1 gB crystal structure (Protein Data Bank ID, 2gum) using PyMol. The numbers indicate locations within the EBV gB amino acid sequence. Insertions predicted by the PHD program to be in specific structural elements of EBV gB (see Table 3 for clarification) are identified by color. Loops and unstructured regions are indicated in green (gB88, gB236, gB190, gB320, gB420, gB561, gB620, gB632, and gB675), those predicted to be in α-helical structures are depicted in red (gB353 and gB486), and β-sheets are represented by blue (gB115, gB198, gB208, gB324, gB361, gB597, and gB602). An asterisk indicates the site of the 421 mutation, which could be mapped to only one strand of the HSV-1 gB crystal structure trimer.

The protein expression of all gB mutants was examined by transient expression of the constructs in CHO-K1 cells. This analysis indicated that all of the linker insertion mutants were expressed, although at different levels (Fig. 2 and data not shown). The gB mutants gB198, gB561, gB597, and gB632 were consistently detected at low levels, whereas the remaining mutants were usually expressed at levels comparable to those of wild-type gB.

FIG. 2.

Expression of EBV gB linker insertion mutants. CHO cells were transfected with gB linker insertion mutants as described in Materials and Methods. Western blot (WB) analysis of gB linker insertion mutant expression in whole cell lysates was determined under reducing conditions. Molecular size markers in kDa are identified to the left of the blot. pSG5 represents the vector control; gB is wild-type gB and contains no linker insertion. Akt was utilized as a loading control in the bottom panel. A representative experiment is shown; expression of each mutant was tested three to five times.

A cleavage product with a molecular size of approximately 50 kDa was detected with wild-type gB, gBΔ798, and the following linker insertion mutants: gB30, gB88, gB115, gB190, gB208, gB324, gB421, gB602, gB620, gB646, gB703, and gB802 (Fig. 2), as well as gB320 and gB832 (data not shown). This cleavage product is similar in size to the 58 kDa C-terminal gB fragment found in EBV virions, which was suggested to be generated by cleavage at the putative furin cleavage site in EBV gB (31). Recombinant EBV gB protein was indeed shown to be processed by insect cells at the predicted furin site (2). The N-terminal cleavage product of 78 kDa was not detected by the polyclonal gB antibody, possibly because the gB epitopes in the reduced N-terminal fragment were not recognized. To investigate this possibility, recombinant EBV gB ectodomain containing residues 23 to 683 (described as EctoS FL in reference 2) was analyzed in the absence and presence of reducing agent. While both N- and C-terminal fragments were visible on the Coomassie blue-stained gel (Fig. 3A), only the smaller C-terminal fragment could be detected by Western blotting (Fig. 3B). These observations indicate that the gB variants listed above are processed in CHO-K1 cells in much the same fashion as was previously seen in vivo.

FIG. 3.

Polyclonal gB antibody does not detect the N-terminal gB cleavage fragment. The EBV gB variant EctoS FL (2) was loaded onto 12% SDS-PAGE gels. (A) An EBV gB variant (0.5 μg) was loaded without and with dithiothreitol (DTT) and stained with Coomassie blue. Molecular sizes in kDa are shown on the right of the gel. (B) The same variant (0.25 μg) loaded in the absence and presence of DTT was probed with polyclonal anti-gB antibody as described for gB isolated from cells. Molecular size markers in kDa are shown on the right of the blot. WB, Western blotting.

The cell-based enzyme-linked immunosorbent assay method generally utilized for the detection of surface-expressed EBV glycoproteins was not sensitive enough to detect EBV gB at the cell surface. Therefore, to determine if the gB mutants were expressed on the surfaces of CHO-K1 cells, biotinylation using sulfosuccinimidyl-1-6-(biotinamido)hexanoate was employed. gBΔ798 is truncated at residue 798 and lacks the endoplasmic reticulum retention signal previously shown to be a determinant of cell surface expression (39, 48). This truncated form of gB served as a positive control for cell surface expression. The intracellular protein actin was used as a negative control to verify that cell lysis and the labeling of intracellular proteins were not occurring (Fig. 4C). CHO-K1 cells were transfected with pSG5, wild-type gB, or gBΔ798, and whole cell lysates were incubated with actin beads; the immunoprecipitated proteins were separated on SDS-PAGE gels and blotted with avidin-HRP (Fig. 4C). No biotinylated actin was detected by this method, indicating that intracellular proteins were not labeled. The converse was also performed: biotinylated proteins from whole cell lysates were immunoprecipitated with neutravidin beads, and blotting was done with either anti-actin or anti-gB antibody (Fig. 4B and A). Biotinylated gB was detected on the cell surface by this method, whereas actin was not (Fig. 4E). The presence of actin, which also served as a loading control, was shown in the whole cell lysates (Fig. 4D and E).

FIG. 4.

Biotinylation control experiments. CHO cells were transfected with pSG5 vector, full-length gB, and the gBΔ798 truncation as previously described. Actin was chosen as a representative intracellular protein to show that only biotinylation of cell surface proteins was detected with this assay. Molecular size markers are on the left of the blot. (A and B) All biotinylated proteins were immunoprecipitated with neutravidin and then the membranes were probed for gB (A) or actin (B). WB, Western blotting. (C) Biotinylated protein was immunoprecipitated (IP) with actin antibody and then probed with avidin-HRP. (D) Whole cell lysates were analyzed for the presence of actin using a polyclonal actin antibody. (E) Biotinylated protein was immunoprecipitated with actin antibody and then probed with actin antibody.

After we established that the integrity of the membrane was not perturbed during the biotinylation process, the presence of gB and gB linker insertion mutants on the cell surface was assessed (Fig. 5A). gB and gBΔ798 were expressed on the cell surface as expected (Fig. 5). gB198 and gB561, which were expressed at barely detectable levels in whole cell lysates, were not detected on the cell surface (data shown for gB198 only) (Fig. 5A, top). Notably, gB236, gB263, and gB380, which were seen in whole cell lysates, were not detectable on the cell surface (Fig. 5A, middle, and data not shown). We suspected that structural changes resulting from linker insertion could have prevented the recognition of these mutants by the conformation-specific CL55 monoclonal anti-gB antibody. To test this hypothesis, CHO-K1 cells expressing these five mutants were biotinylated, and gB from cell lysates was immunoprecipitated with a polyclonal gB antibody (gB421 and gB486 were utilized as additional controls for positive cell surface expression [Fig. 5B]). A faint band for linker insertion mutant gB263 was detected, indicative of the mutant's presence on the surfaces of CHO-K1 cells, while the other four variants were absent (Fig. 5B). The four variants (gB198, gB561, gB236, and gB380) were likely not expressed on the cell surface due to impaired folding or trafficking of the protein.

FIG. 5.

Cell surface expression of gB linker insertion mutants. (A) CHO cells were transfected with gB linker insertion mutants as described previously. The cell surface expression of gB linker insertion mutants was determined by biotinylation and then immunoprecipitation (IP) with a monoclonal gB antibody. pSG5, vector control; gBΔ798, gB truncated at amino acid 798. Molecular size markers are on the left of the blots. WB, Western blotting. (B) CHO cells were transfected with gB linker insertion mutants that were not detected on the cell surface with monoclonal gB antibody. The cell surface expression of these gB linker insertion mutants was determined by biotinylation and immunoprecipitation with a polyclonal gB antibody. pSG5, vector control; gBΔ798, gB truncated at amino acid 798. Molecular size markers are on the left of the blots.

In summary, of the 28 linker insertion mutants analyzed, all were expressed intracellularly and 24 were expressed on the cell surface (Table 2).

TABLE 2.

Summary of results

Linker insertion	Expression by WBa	Cell surface expression	N-linked glycosylation product	Ability To oligomerize	Epithelial fusion		B-cell fusion		Category
					Without gH/gL	With gH/gL	Without gH/gL	With gH/gL
Wild-type gB	Yes	Yes	Yes	Yes	No	Yes	No	Yes
gBΔ798	Yes	Yes	Yes	Yes	Yes	Yes	No	No
gB30	Yes	Yes	Yes	Partialc	No	Yes	No	Yes	III
gB88	Yes	Yes	No	Yes	No	No	No	No	I
gB115	Yes	Yes	No	Yes	No	No	No	No	I
gB190	Yes	Yes	No	Yes	No	No	No	No	I
gB198	Yes	No	No	No	No	No	No	No	I
gB208	Yes	NTb	NT	NT	No	No	No	No	I
gB236	Yes	Yes	No	Yes	No	No	No	No	I
gB263	Yes	Yesd	Yes	Yes	No	No	No	No	II
gB307	Yes	Yes	No	Yes	No	No	No	No	I
gB320	Yes	Yes	Yes	Yes	No	No	No	No	II
gB324	Yes	Yes	No	Yes	No	No	No	No	I
gB353	Yes	Yes	No	Yes	No	No	No	No	I
gB361	Yes	Yes	No	Yes	No	No	No	No	I
gB380	Yes	No	No	Yes	No	No	No	No	I
gB421	Yes	Yes	Yes	Yes	No	Yes	No	Yes	III
gB486	Yes	Yes	No	Yes	No	No	No	No	I
gB561	Yes	No	No	No	No	No	No	No	I
gB597	Yes	Yes	No	No	No	No	No	No	I
gB602	Yes	Yes	No	No	No	No	No	No	I
gB620	Yes	Yes	No	No	No	No	No	No	I
gB632	Yes	Yes	No	Yes	No	No	No	No	I
gB646	Yes	Yes	No	Yes	No	No	No	No	I
gB675	Yes	Yes	No	Yes	No	No	No	No	I
gB703	Yes	Yes	No	Yes	No	No	No	No	I
gB717	Yes	Yes	No	Yes	No	No	No	No	I
gB734	Yes	Yes	No	Partial	No	No	No	No	I
gB802	Yes	Yes	No	Yes	No	No	No	No	I
gB832	Yes	Yes	Yes	Yes	No	Yes	No	Yes	III

^aDetectable protein is indicated as Yes regardless of the level of expression compared to control. WB, Western blotting.

^bNT, not tested.

^cPartial indicates that a less complex form of oligomer was detected.

^dDetectable on the cell surface only with polyclonal gB antibody.

Oligomerization of EBV gB linker insertion mutants.

Herpes simplex virus (HSV) gB has been shown to oligomerize, forming dimers in vivo, and most recently, trimers were identified in the crystallized recombinant gB ectodomains (10, 19, 26). The oligomers formed by HSV type 1 (HSV-1) gB are generally heat sensitive, but linker insertion mutation appears to provide some structural stability, which leads to the detection of oligomers upon boiling (10, 44). HSV-1 gB can be found in three forms in vivo: a precursor 110-kDa form, a mature glycosylated 120-kDa form, and oligomers of various sizes (10). Recent evidence from a study using linker insertion mutagenesis of HSV-1 gB indicates that insertion in several β-sheet regions in the ectodomain can prevent oligomerization (44). Simple EBV gB ectodomain trimers were detected when hydrophobic residues in the putative fusion loops were mutated into the corresponding less hydrophobic residues from HSV-1 gB (1). We assessed the abilities of gB and gB linker insertion mutants to form oligomers in vivo.

Oligomer formation was analyzed in whole cell lysates for wild-type gB and all linker insertion mutants (Fig. 6). Oligomers formed by wild-type EBV gB and gBΔ798 are shown in Fig. 6, including the band of a smaller oligomer, possibly a dimer, and those that we believe are more complex and that are seen at higher molecular weight. These complex oligomers may encompass a variety of forms of gB oligomers.

FIG. 6.

Oligomer formation of gB linker insertion mutants. Oligomer formation of gB linker insertion mutants determined in whole cell lysates under nonreducing conditions. pSG5, vector control; gB and gBΔ798, positive controls. Oligomers are identified by brackets to the right of the blot; the small bracket with the asterisk identifies an oligomeric band; the larger bracket denotes the location of complex oligomers containing multiple bands. Monomeric gB is indicated by an open arrow, and the novel band is designated by a closed arrow. Molecular sizes are shown on the left of the blot.

The vast majority of gB linker insertion mutants formed oligomers, with the exceptions of gB198, gB561, gB597, gB602, and gB620 (Fig. 6). As mentioned previously, gB198 and gB561 are expressed at very low levels in CHO-K1 cells, and thus, the lack of detectable amounts of oligomers was not surprising. gB597, gB602, and gB620 were all expressed at detectable levels but were unable to form complex oligomers (Fig. 6). In the representative blot shown in Fig. 6, gB602 exhibits a faint band at the level of a complex oligomer that might have been detected with a longer exposure, but this band was not present on other blots.

gB30 and gB734 are expressed but did not appear to form higher molecular weight oligomers like those observed for wild-type gB (Fig. 6). The aforementioned mutants that were unable to form complex oligomers are indicated in the summary table as forming partial oligomers (Table 2).

Although the linker insertion mutants gB597, gB602, and gB620 were expressed at the cell surface, they did not appear to oligomerize in a manner similar to that observed for wild-type gB. These data suggest that oligomerization is not essential for cell surface expression of EBV gB and that the region encompassed by gB597 to gB620 is an area important for oligomerization. The loss of oligomerization caused by structural changes due to insertions in regions around residues 597, 602, and 620 in EBV gB is consistent with the possible involvement of this region in the formation of the oligomeric interface.

Examination of the novel higher molecular mass band detected under nonreducing SDS-PAGE conditions.

HSV-1 gB is detected as an immature form of 100 kDa and a mature glycosylated form of 120 kDa (10, 44). EBV gB has also been reported to be either a 110-kDa or a 125-kDa protein (13, 17, 35, 39, 55). Wild-type gB, gBΔ798, and the linker insertion mutants gB30, gB263, gB320, gB421, and gB832 exhibited a higher molecular size band that migrated just above monomeric gB (Fig. 6) under nonreducing conditions. We explored the possibility that this higher molecular size band was a cell-type-specific effect of gB overexpression, as well as the possibility that it could represent a mature form of EBV gB produced through more extensive glycosylation than that for monomeric gB.

We first determined if the higher molecular size gB variant detected in CHO-K1 cell lysates was a result of CHO-K1-specific processing of gB. The analogous form of gB was not detectable in whole cell lysates of EBV⁺ Akata cells induced to produce virus or the gB-null LCL M.2 transfected with wild-type gB (Fig. 7A and data not shown). This band may be present in these cell types but difficult to detect. Whole cell lysates from transiently transfected 293T epithelial cells and an EBV-infected LCL induced to undergo lytic replication were blotted with polyclonal anti-gB antibody, and the presence of the higher molecular size form was confirmed (data not shown and Fig. 7B), indicating that the band was likely not due to overexpression or differential processing in CHO-K1 cells.

FIG. 7.

Characterization of a novel band identified for EBV gB under nonreducing conditions. (A) The presence of the novel band was determined in uninduced (−IgG) and induced (+IgG) EBV-GFP⁺ Akata cells. SDS-PAGE was performed with lysates that did not contain a reducing agent and probed with a polyclonal anti-gB antibody. Molecular size markers are on the left of the gel. The monomeric form of gB is identified by the open arrow. WB, Western blotting. (B) Expression of the novel band was analyzed in the gB⁺ EBV-transformed LCL10, which was transfected with pSG5 and BZLF1 or gB and BZLF1. SDS-PAGE was performed with lysates that did not contain a reducing agent and probed with a polyclonal anti-gB antibody. Molecular size markers are on the left of the gel. The novel band is identified by the closed arrow; monomeric gB is indicated with an open arrow. (C) EndoH treatment of cells transfected with pSG5 vector control or wild-type gB. Whole cell lysate was first immunoprecipitated (IP) with CL55 monoclonal anti-gB antibody. The left gel shows the initial immunoprecipitation, with gB oligomers identified with a bracket, the novel band with a closed arrow, and monomeric gB with an open arrow. The middle gel depicts treatment of immunoprecipitations under reducing conditions without enzyme (−EH) or with enzyme (+EH). The monomeric form of gB with and without enzyme treatment is indicated with open arrows. The right gel shows treatment of gB immunoprecipitations under nonreducing conditions without enzyme (−EH) and with enzyme (+EH). gB oligomers, which were not detected under reducing conditions, are identified with a bracket, the novel band is shown with a closed arrow, and the monomeric form of gB in the absence and presence of enzymatic digestion is shown with open arrows. (D) PNGase F and O-glycosidase treatment of cells transfected with gB. The left gel shows lysates treated with enzymes under reducing conditions with dithiothreitol. The right gel contained lysates treated under nonreducing conditions. All samples, under reducing or nonreducing conditions, were immunoprecipitated with CL55 anti-gB monoclonal antibody, and Western blots were then probed with polyclonal anti-gB antibody. Lane 1 in each panel shows untreated gB eluent, whereas lane 2 is the gB eluent with buffer added and subjected to the same incubation conditions as those samples containing enzyme. The enzymes added to gB immunoprecipitates were as follows: lanes 1, initial IP; lanes 2, no enzyme; lanes 3, PNGase F; lanes 4, PNGase F plus sialidase; lanes 5, PNGase F, sialidase, β-1,4-galactosidase, β-N-acetylglucosaminidase, and O-glycosidase; lanes 6, O-glycosidase; lanes 7, O-glycosidase, sialidase, β-1,4-galactosidase, and β-N-acetylglucosaminidase. The bracket on the right of the blot indicates the location of gB oligomers, the closed arrow represents the novel band detected under nonreducing conditions, and the open arrows point to the locations of the monomeric form of gB.

To further understand the nature of the higher molecular size gB variant, cells were transfected with wild-type gB and then treated with EndoH to determine if this variant corresponded to a more glycosylated form of gB. EndoH removes high-mannose N-linked carbohydrates. Both EBV gB and HSV-1 gB are modified during protein maturation with N-linked glycosylation that can be removed by EndoH treatment (10, 38, 39, 55, 73). There are nine potential sites for N-linked glycosylation modifications (consensus sequence, NXT/S) within the EBV gB ectodomain; these residues are indicated in Fig. 1, top. Whole cell lysates from CHO-K1 cells transfected with EBV gB and pSG5 were immunoprecipitated with CL55 gB antibody and treated with EndoH as described in Materials and Methods. Under reducing conditions, the higher molecular size band is lost, and a decrease in the molecular weight of EBV gB can be seen upon treatment with EndoH, confirming that gB is modified by high mannose N-linked glycosylations (Fig. 7C, middle). EndoH treatment under nonreducing conditions led to a downward shift in the molecular weight of monomeric gB, as well as gB oligomers, and a barely detectable shift was seen with the higher molecular size band, indicating that this form of gB is not solely a result of high mannose N-linked glycosylation modifications (Fig. 7C, right).

PNGase F treatment was also utilized to analyze the glycosylation modifications occurring on EBV gB; the enzyme removes a wider panel of N-linked modifications than EndoH. Under reducing conditions, PNGase F alone altered the migration pattern of wild-type gB (Fig. 7D, left). PNGase F, in addition to sialidase, which removes modifications to allow PNGase F enzymatic access to glycosylation sites, also decreased the size of gB compared to untreated sample (Fig. 7D, left). These results were consistent with what was seen with EndoH treatment under reducing conditions, indicating that gB may be modified by multiple forms of N-linked glycosylation.

The presence of O-linked glycosylation modifications was assessed using O-glycosidase treatment following the treatment of lysates with PNGase F. Treatment with PNGase F, O-glycosidase, sialidase, β-1,4-galactosidase, and β-N-acetylglucosaminidase led to a shift in the size of wild-type gB comparable to that produced by treatment of PNGase F alone (Fig. 7D, left). If gB was indeed O glycosylated, we would have predicted an additional decrease in size. O-Glycosidase treatment alone or in conjunction with all enzymes except PNGase F did not alter the size of monomeric gB under reducing conditions (Fig. 7D, left, lanes 6 and 7). Under nonreducing conditions, PNGase F treatment led to a shift of the upper band downward, presumably to the size of monomeric gB at 110 kDa (Fig. 7D, right). PNGase F treatment of gB with sialidase or other enzymes, including O-glycosidase, did not further alter the size of monomeric or oligomeric gB under nonreducing conditions (Fig. 7D, right). O-Glycosidase treatment did not alter the migration pattern of the upper band under nonreducing conditions (Fig. 7D, right). Treatment with PNGase F, O-glycosidase, or the other enzymes shown in Fig. 7 also did not alter the migration pattern of EBV gB oligomers. Based on these studies, it appears that the unique upper band identified represents an N-glycosylated modified form of monomeric gB that is sensitive to PNGase F.

Evaluation of the ability of EBV gB linker insertion mutants to mediate fusion with epithelial and B cells.

A previously developed cell-cell fusion assay was used to determine if gB linker insertion mutants could elicit fusion with epithelial and B cells (22, 34, 48, 54, 63). The abilities of the linker insertion mutants to fuse with the two EBV target cell types, both with and without gH/gL, were tested. For epithelial cells, the level of fusion mediated by wild-type gB, gH, and gL (positive control) was set to 100%, and the fusion results with linker insertion mutants were then normalized to the positive control. Positive fusion ability was designated as fusion of greater than 20% compared to the positive control. Of those mutants that were expressed in CHO-K1 cells as determined by Western blotting (Fig. 2), only gB30, gB421, and gB832 were able to promote fusion with 293 epithelial cells, and this occurred only in the presence of the gH/gL complex. Epithelial fusion measured in the presence of gH/gL was approximately 55% for gB30, 65% for gB421, and 90% for gB832 (Fig. 8A). Fusion with B cells was observed with these three mutants in the presence of gH/gL and gp42. For B-cell fusion, CHO-K1 cells cotransfected with gB, gH/gL, and gp42 served as the positive control, and the level of fusion mediated by this group of glycoproteins was set to 100%. gB30 exhibited approximately 30% fusion activity, gB421 exhibited 65% fusion activity, and gB832 functioned at or above wild-type levels (Fig. 8B). These three mutants, gB30, gB421, and gB832, which functioned in fusion with B cells, were functional only with the gH/gL complex. The 25 remaining linker insertion mutants were not functional in fusion with either cell type.

FIG. 8.

Function in fusion of EBV gB linker insertion mutants. (A) Fusion function of EBV gB linker insertion mutants with epithelial cells. Wild-type gB was set at 100% fusion, and the percent fusion was then calculated relative to this value for all linker insertion mutants. A minimum of three fusion assay experiments were performed for each mutant. The error bars represent standard errors. (B) Fusion function of EBV gB linker insertion mutants with B cells. Wild-type gB was set at 100% fusion, and the percent fusion was then calculated relative to this value for all linker insertion mutants. A minimum of three fusion assay experiments were performed for each mutant. The error bars represent standard errors.

gB linker insertion mutant complementation of EBV gB-null virus.

One of the unique aspects of EBV infection is the ability of the virus to transform primary B cells that can give rise to multiple forms of B-cell lymphoma, as well as LCLs. In order to assess the ability of EBV virions containing linker insertion mutant gBs to infect and transform primary B cells, we took advantage of a previously characterized gB-null LCL in which a hygromycin cassette was inserted into the gB reading frame (27, 40). gB-null LCLs were complemented with pSG5, gB, and the three linker insertion mutants that had been identified as functional in the cell-cell fusion assay (gB30, gB421, and gB832). The gB-expressing wild-type LCL10 line was used as a positive control. LCL10 cells were nucleofected as described in Materials and Methods with either pSG5 or gB to determine if gB overexpression affected focus formation. Overexpression of gB in these cells did not increase the number of wells in which foci formed (data not shown). Of the three functional fusion mutants tested in the viral entry assay, only gB421 and gB832 were able to complement the gB-null virus (data not shown). The mutant gB30 did not lead to the transformation of primary B cells, and it is likely that this poor ability to elicit fusion with B cells translated into an entry defect, as seen in the complementation assay.

DISCUSSION

EBV gB is an essential component of the viral fusion machinery, and functional regions of this protein, apart from the C-terminal tail, which regulates fusion, have not been characterized (22). Therefore, the goal of this study was to investigate the roles of potential functional domains within EBV gB. Previous reports utilizing random linker insertion mutagenesis within HSV-1 gB indicated sensitivity to structural alteration by amino acid insertions, which prevented either expression of gB or its proper folding (53, 57). The lack of expression or inability to fold into a functional conformation was shown to be primarily due to insertions into α-helix or β-sheet secondary structural elements (53, 57). In our current study, mutants were expressed despite the site of insertion in the secondary structure of EBV gB and most mutants formed oligomers, but the majority were not fully glycosylated, and only three mutants were functional in a cell-based fusion assay.

Studies of HSV-1, as well as HSV-2, have indicated that gB may be more amenable to linker insertions, as well as to smaller insertions, if they are located in loop structures between α-helices or β-sheets (42, 53). Recent evidence using a global random linker insertion method identical to ours in HSV-1 gB yielded results very similar to those in our study (44). In both linker insertion mutagenesis studies, the vast majority of linker insertion mutants were expressed, could form oligomers, and were present on the cell surface, regardless of whether the insertions were located in α-helices, β-sheets, or loop regions. Our data also correlate well with the study describing characterization of HCMV gB by linker insertion mutagenesis, indicating that HCMV gB tolerates moderate structural disruptions without altering oligomer formation and trafficking to the cell surface (64). Unlike these previous studies, however, in some cases, linker insertion mutations in EBV gB prevented full glycosylation of the protein and function in fusion.

Based on the results presented here, the 28 EBV gB linker insertion mutants can be classified into three categories. Category I mutants were expressed in whole cell lysates, and the majority were detected on the cell surface but were not fully glycosylated and were nonfunctional in fusion. Category II mutants, which were expressed in whole cell lysates and on the cell surface, were fully glycosylated and were nonfunctional in fusion. Category III mutants were fully glycosylated and functional in fusion. Using the HSV-1 gB crystal structure in concert with secondary structure predictions obtained for EBV gB from the PHD neural network, mutations in all categories described here were mapped to the HSV-1 gB crystal structure (Fig. 1, bottom). In addition, the predicted secondary structural regions of EBV gB and HSV-1 gB were compared to the actual crystal structure of HSV-1 gB in Table 3. The predicted secondary structure elements of EBV and HSV-1 gB are strikingly similar, consistent with the functional conservation of gB within the herpesvirus family (26).

TABLE 3.

Locations of EBV gB linker insertions in the predicted EBV gB structure

Linker site	Location in predicted EBV secondary structurea	Location in predicted HSV-1 secondary structureb	Location in HSV-1 crystal structurec
gB30	L	L	NA
gB88	L	E/L	β3
gB115	E	E/H	β5
gB190	L	L	US
gB198	E	E	β11
gB208	E	E	β11
gB236	L	E/L	US
gB263	L	L	US
gB307	H	H	β18
gB320	E	L	US
gB324	E	E	β20
gB353	H	H	αB
gB361	US	E/L	β21
gB380	E	E	US
gB421	L	L	US
gB486	H/US	H/L	αC
gB561	E	E/L	β29
gB597	E	E	β33
gB602	E	E	β34
gB620	E	E	β35
gB632	E/US	L	US
gB646	US	L/H	αE
gB675	H	H	NA
gB703	H	US	NA
gB717	E	US	NA
gB734	H	US	NA
gB802	H	H	NA
gB832	H/L	H	NA

^aThe PredictProtein program (http://www.predictprotein.org) was used for analysis of EBV gB, and the PHD neural network was utilized to predict gB secondary structure. E, β-sheet; H, α-helix; L, loop; US, unstructured.

^bThe PHD neural network method was also employed to estimate the secondary structure of HSV-1 as previously described (53). E, β-sheet; H, α-helix; L, loop; US, unstructured.

^cThe results of secondary structure predictions were also compared to the HSV-1 gB crystal structure to assess the accuracy of the predictions (26). NA, not applicable; this region was not determined in the HSV-1 gB crystal structure.

Category I mutants.

Category I mutants were detectable in whole cell lysates under reducing and nonreducing conditions and were variably expressed on the cell surface (Table 2). The mutants that fall into this category are gB88, gB115, gB190, gB198, gB236, gB307, gB324, gB353, gB361, gB380, gB486, gB561, gB597, gB602, gB620, gB632, gB675, gB703, gB717, gB734, and gB802. All lacked the N-linked glycosylated form found for wild-type gB that may represent mature gB. When mapped to the HSV-1 gB crystal structure, five mutants classified in category I (gB88, gB307, gB361, gB380, and gB646) map to regions in EBV gB predicted to form secondary structures different from what has been found for HSV-1 gB (Table 3 and Fig. 1, bottom).

Mutants in this category were also located near predicted structural and functional elements. Linker insertion mutants were found in the three predicted amphipathic helices, the transmembrane domain, and the approximately 100-residue-long cytoplasmic tail (Fig. 1, top). One linker insertion mutant, gB734, lies within the proposed transmembrane domain. This mutant is expressed on the cell surface but does not form higher-order oligomeric structures. This finding indicates a putative role for the transmembrane domain in oligomerization that may be unique to EBV gB. HSV-1 gB linker insertion mutants located in the transmembrane domain, residues 774 to 796, did not alter oligomer formation (44).

In addition to predicted structural domains, mutants classified in category I lie within predicted functional domains, including coiled-coil domains, the cytoplasmic tail, and near the putative fusion loop residues (Fig. 1, top). Coiled-coil domains are secondary structure motifs important for oligomerization of protein subunits and are identified in viral proteins as structural elements essential for fusion (7, 12, 33). Peptides targeted to heptad repeat regions predicted to form coiled coils in HCMV gB and gH prevented virus entry (46). EBV gB has three predicted coiled-coil regions, residues 459 to 482, 656 to 684, and 789 to 816, based on the COILS program (http://www.ch.embnet.org/software/COILS_form.html) (47). Linker insertion mutants gB486, gB675, and gB832 mapped within or directly adjacent to the three predicted coiled-coil regions (Fig. 1, top). The final predicted coiled-coil region in the C-terminal tail contains the insertion gB802. Previous truncations made in this region at residue 801 or 798 allowed fusion with epithelial cells independently of gH/gL at approximately 60% of wild-type fusion levels (48), but the linker insertion in this region abolished fusion entirely with and without gH/gL. The fusion loops of HSV-1 gB (residues 173 to 179 and 258 to 265), as well as EBV gB (residues 112 to 113 and 193 to 196), were recently characterized and are critical for gB fusion activity (1, 23). These fusion loops, which are rich in hydrophobic residues in EBV gB, are similar in sequence to fusion peptides found in class I and class II fusion proteins (14, 33). Three mutants described in this study had insertions in close proximity to the fusion loops identified within EBV gB (gB115, gB190, and gB198) (Fig. 1, top) (1). HSV-1 gB linker insertion mutants located near the fusion loops (A261 and Y265) also prevented fusion without affecting cell surface expression or oligomer formation (44). Disturbances in these crucial fusogenic regions of gB could have resulted in nonfunctional proteins.

The most interesting discovery within category I was the inability of gB597, gB602, and gB620 to form complex oligomers. Residues 597 to 620 in EBV gB correspond to residues 642 to 665 in HSV-1 gB. These insertions were mapped to domain IV (formed by residues 111 to 117 and 573 to 661) of HSV-1 gB, which contains an epitope recognized by virus-neutralizing antibodies (Fig. 1, bottom) (26, 28). Linker insertion mutants generated in HSV-1 gB that were mapped to domain IV did not appear to alter oligomer formation but were unable to be processed to the mature form of gB (44). The HSV-1 gB β-strand (residues 665 to 668), to which the gB620 insertion was mapped, forms a sheet with β-strands provided by an adjacent gB monomer in the trimer complex. If the structure of EBV gB is indeed similar to that of HSV-1 gB, the same strand in EBV gB could be used in a similar fashion to form trimeric interfaces. Linker insertions in the analogous region of HSV-1 gB, T665 and V667, also abrogated oligomer formation (44). Within domain IV of the crystal structure is a region that is homologous to the AD-1 region found within HCMV gB. The AD-1 homologous region encompasses residues 584 to 658 in HSV-1 gB and 538 to 613 in EBV gB (6). In HCMV gB, AD-1 has been demonstrated to be important for oligomerization, as well as antibody recognition (6, 66). EBV gB linker insertion mutants that were localized to the AD-1 homologous region, gB561, gB597, and gB602, were unable to oligomerize as well.

Category II mutants.

Two mutants from the current study fall into category II. Both of these mutants, gB263 and gB320, are detectable on the cell surface, are fully glycosylated, and are nonfunctional in fusion. gB263 maps to a predicted loop region in the EBV structure and an unstructured region in the HSV-1 gB crystal structure (Table 3). This mutant is detectable on the cell surface only by using a polyclonal antibody, identifying a possible epitope for the CL55 monoclonal antibody. Interestingly, gB320 is one of six mutants that map to a different structural element in HSV-1 gB than was predicted for EBV gB (Table 3). Linker insertion in these two regions may have disrupted potential interactions with gH/gL, gp42, the site of interaction with a gB receptor, or blocked a conformational change in gB required for activation that thus prevented function in fusion. These mutants are of great interest and are currently under investigation.

Category III mutants.

The three category III mutants, gB30, gB421, and gB832, were fully glycosylated, were expressed in whole cell lysates and on the cell surface, and were functional in fusion with epithelial and B cells. Of the three, gB30 functioned poorly in fusion with B cells and was unable to complement gB-null virus. This result indicates that regions in the gB N terminus might be a determinant of cell-type specificity or interaction with gp42. It is interesting that all category III mutants map to regions in EBV gB that are predicted to form loops (26). This suggests that these particular insertions are located in highly flexible regions that are unlikely to form core secondary structure elements, and thus, function in fusion is not disrupted by these mutations. In the crystal structure of HSV-1 gB, which is believed to be a postfusion conformation, these mutations would localize to regions that were not present in the gB structure (gB30 or gB832) or were in a potentially unstructured region removed by tryptic digestion of the crystallized protein (gB421) (Table 3). In support of these findings, linker insertion mutations generated in HSV-1 gB that fell into the extreme N or C terminus of the protein, as well as the linker region, were functional in fusion (44). gB832 is located in the cytoplasmic tail of EBV gB. Mutations within and truncation of the cytoplasmic tails of multiple herpesvirus gBs have been demonstrated to increase the fusogenic ability of gB (3, 8, 9, 11, 15, 16, 22, 24, 36, 48, 51, 56). With the close proximity of the C-terminal tail of HSV-1 gB to the putative fusion loops in the postfusion crystal structure, it is not surprising that disruption of structures within the C-terminal tail of EBV gB may influence fusion capability (26).

Summary.

gB has been postulated to play the role of a fusogen in herpesvirus entry. Recent structural data for HSV-1 gB is compatible with this supposition due to its similarity to the postfusion form of vesicular stomatitis virus (VSV) G protein (26). Comparison of the predicted secondary structure elements of EBV gB and the HSV-1 gB ectodomain crystal structure suggests that EBV gB has a three-dimensional structure similar to that of HSV-1 gB and therefore also that of VSV G. Coincidentally, analysis of VSV G fusion domains by linker insertion mutation yielded results similar to those of our study in that the majority of mutants were expressed, their cell surface expression levels varied, and oligomers were formed by mutant proteins (43). Our linker mutation analysis indicated that disruption of EBV gB structure, albeit by the insertion of only 5 amino acids, drastically affected membrane fusion, even though all mutants were expressed and most were able to traffic to the cell surface. Category I mutants were most likely nonfunctional because processing to the mature N-glycosylated form of gB did not occur. Category II mutants were properly processed but were also nonfunctional in the cell-based fusion assay, possibly due to incomplete “triggering” of the protein to its fusion-active form. It is known that triggering of conformation changes in VSV G requires low pH, and upon exposure to low pH, the compact prefusion form of the protein unfolds into the extended postfusion form (59-61). Unlike VSV G, herpesvirus gB is not the only glycoprotein required for entry, and it is likely that the trigger for conformation change from the pre- to postfusion form is an interaction, direct or indirect, with the other glycoproteins required for entry into target cells or with a cell surface receptor(s) (25, 58). Recent evidence utilizing HSV-1 entry proteins has produced a model in which sequential and concerted action of gD bring the cell and virus membranes together; gH/gL then allows lipid mixing between the two membranes, and gB allows formation of the final pore and the completion of the fusion process (69). EBV fusion most likely occurs in a similar stepwise manner with gp42 playing the role of gD in B-cell fusion and gB and gH/gL cooperating in the process for epithelial cell fusion. The category II mutants, gB263 and gB320, can be used to identify interactions of EBV gB with other EBV-encoded glycoproteins, as well as a potential cellular receptor for gB. The two mutants may also be instrumental in identifying structural conformational changes that occur in EBV gB during the fusion process. Finally, the category III mutant gB30 was also intriguing because function in fusion was greatly diminished with B cells and the mutant was completely negative for complementation of a gB-null virus. This suggests a putative role for the N terminus of EBV gB in B-cell entry and in determining cell tropism. It is possible that mutation disrupted a conformational change in gB that would normally occur upon interaction with the other EBV glycoproteins. Further analysis of the mutants described in the current study awaits the determination of the EBV gB structure in both the pre- and postfusion forms.