An N-Terminal Domain of Herpes Simplex Virus Type I gE Is Capable of Forming Stable Complexes with gI

Abstract

Using limited proteolytic analyses, we show that gE present in soluble herpes simplex virus type 1 gE-gI complexes is cleaved into a C-terminal (CgE) and an N-terminal (NgE) domain. The domain boundary is in the vicinity of residue 188 of mature gE. NgE, but not CgE, forms a stable complex with soluble gI.

The glycoproteins gE and gI of alphaherpesviruses form stable complexes, which have been implicated in multiple functions. These include immune system evasion via the ability to bind to the Fc domains of human immunoglobulin G (IgG), enhancement of viral cell-to-cell spread, and virulence (reviewed in references 8, 13, 16, and 17). Much remains to be understood about which regions of the gE-gI complex are important for each function and the overall molecular basis for each function. To allow for molecular characterization of gE-gI functions, we previously expressed soluble forms of herpes simplex virus type 1 (HSV-1) gE and gI in CHO cells and showed that the glycoproteins assembled into stable complexes (5). We determined that the stoichiometry of the gE-gI complex was 1:1. We also demonstrated that soluble gE-gI complexes bound human immunoglobulins with a 1:1 stoichiometry and with K_d values in the range of 200 to 400 nM. In the present studies, we undertook investigations of the domain structure of gE-gI complexes, with the goals of obtaining further insights into protein domains important for the formation of the gE-gI complex and for the function of the gE-gI complex in viral spread and Fc binding. Other studies have identified segments of gE and gI that are important for the gE-gI interaction and the gE-gI–IgG interaction (1, 2, 11). While the studies map gE-gI and gE-gI–IgG interactions to the linear sequence of gE or gI, little insight is available about gE-gI interactions in the context of the three-dimensional structure of the protein complex. Limited-proteolysis experiments have been valuable for providing structure-function correlations and information about domain organization in other systems (4, 14, 15). Here we used limited proteolytic analysis to obtain insights into the domain structure of soluble gE-gI. We showed that the extracellular domain of gE contains a C-terminal and an N-terminal domain, with the domain boundary in the vicinity of residue 188 of mature gE. Subsequently, we analyzed the ability of each gE domain to form complexes with gI, as well as to interact with the Fc domains of immunoglobulins. We interpret the results of these studies using sequence alignments of gE from several alphaherpesviruses.

Proteolytic digestion of gE-gI complexes yields information about a domain boundary.

Soluble gE-gI was purified from transfected CHO cells using human IgG-based affinity chromatography as previously described (5), and was subjected to limited proteolytic analysis at 4°C with three different proteases. Five to 20 μg of the gE-gI protein was digested with either 0.12 to 0.5 μg of trypsin (in a buffer containing 100 mM Tris-HCl [pH 8.5]), 0.25 to 1 μg of chymotrypsin (in 100 mM Tris-HCl–10 mM CaCl₂ [pH 7.8]), or 1 to 4 μg of endoproteinase Glu-C (in 50 mM phosphate buffer [pH 7.8]). All proteolytic enzymes were obtained from Roche Molecular Biochemicals. Reactions were quenched by addition of the protease inhibitor N-tosyl-l-lysine chloromethyl ketone (TLCK) (at 50 μg/ml, for trypsin) or aprotinin (at 1 μg/ml, for chymotrypsin and endoproteinase Glu-C). Samples were boiled in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) buffer and analyzed using an SDS–12% PAGE gel. Digestion of soluble gE-gI with all three enzymes resulted in the degradation of gE into smaller fragments in the molecular size range of 20 to 30 kDa, as well as some fragments smaller than 20 kDa (Fig. 1). Remarkably, gI was stable to digestion by all three enzymes.

FIG. 1.

Digestions of soluble gE-gI with three different proteases show that gI is stable while gE is degraded into 20- to 25-kDa fragments. Soluble gE-gI complexes were subjected to limited proteolytic digestion with trypsin, endoproteinase Glu-C (endo Glu-C), or chymotrypsin at 4°C for 30, 30, or 120 min, respectively. Products of tryptic digestion were analyzed by SDS-PAGE. Lane 1, intact gE-gI complexes (20 μg); lane 2, trypsin-digested gE-gI (20 μg); lane 3, endoproteinase Glu-C-digested gE-gI (5 μg); lane 4, chymotrypsin digested gE-gI (5 μg).

Anion-exchange chromatography was used to establish the identities of the tryptic digestion products. For this analysis, 400 μg of gE-gI protein in 100 mM Tris-HCl (pH 8.5) was digested with 5 μg of trypsin for 30 min at 4°C. The reaction was quenched by addition of TLCK, and digested protein loaded on a Mono Q column (Amersham Pharmacia Biotech). The column was washed with 20 mM Tris-HCl (pH 8.5), and proteins were eluted using a gradient generated with 20 mM Tris-HCl–1 M NaCl (pH 8.5). Two major chromatographic peaks were resolved (Fig. 2A). SDS-PAGE analysis of fractions corresponding to the two peaks indicated the presence of intact soluble gI (40 to 45 kDa), as well as several lower-molecular-weight bands in the 20- to 30-kDa range (Fig. 2B). The identities of the proteins were established after transfer of proteins to polyvinylidene difluoride membranes and N-terminal sequence analysis using an Applied Biosystems model 494 sequencer. Peak 1 (Fig. 2B, lanes 1 and 2) contains intact gI (with the N-terminal sequence LVVRG), as well as C-terminal 20- to 30-kDa fragments of gE, all of which initiate at residue 189 of mature gE (N-terminal sequence, SWPSA). Peak 2 (Fig. 2B, lanes 3 and 4) contains predominantly gI (N-terminal sequence, LVVRG), but also ∼20- to 24-kDa fragments of gE, all initiating at (GTPKT) or near (GPTQK; residue 23 of mature gE) the N terminus of mature gE (Fig. 2B). The soluble gE construct expressed in CHO cells is truncated at position 399 (mature gE numbering [5]). The molecular weight of the largest C-terminal gE fragment observed (lanes 1 and 2) is consistent with the expected size for the gE 189-to-399 fragment, including one glycosylation site within this stretch. The smallest and largest C-terminal gE fragments observed in peak 1 are expected to differ by ∼50 residues (at the C terminus). These results suggested the existence of a domain boundary for gE in the vicinity of residue 188 of mature gE.

FIG. 2.

Analysis of tryptic digestion products yields information about a domain boundary in gE. (A) Soluble gE-gI complexes were digested with trypsin (substrate/enzyme ratio, 80:1 by weight) at 4°C, and digested proteins were separated on a Mono Q column. (B) Chromatographic peaks 1 and 2 from panel A were analyzed by SDS-PAGE followed by N-terminal sequencing. Lanes 1 and 2, proteins contained in peak 1, including gI (with the N-terminal sequence LVVRG) and gE fragments with SWPSA as the N-terminal sequence (initiating at residue 189 of mature gE). Lanes 3 and 4, proteins contained in peak 2, including gI (with the N-terminal sequence LVVRG) and gE fragments with the N-terminal sequence GTPKT (initiating at the N terminus of mature gE) or GPTQK (initiating at residue 23 of mature gE).

The observation that gI fractionated into two distinct peaks upon tryptic digestion of gE-gI complexes (Fig. 2) suggested that distinct gE-gI complexes, or gE-gI complexes and free gI, were being resolved in the two peaks. This result raised the question of whether the N-terminal gE domain, the C-terminal gE domain, or both could form complexes with gI. In peak 1, C-terminal gE peptides appeared to be in stoichiometric excess relative to gI. In peak 2, N-terminal gE peptides appeared to be substoichiometric relative to gI. Thus, based on the expected 1:1 stoichiometry (5), it was difficult to assess which of the gE peptides in the two sets, if any, were complexed to gI. By immunoprecipitations with gI-specific antibodies, we could not demonstrate that either the C-terminal or the N-terminal gE fragments were complexed to gI (data not shown). For the N-terminal gE fragments, interpretations of the results of these immunoprecipitation experiments were complicated by the low overall recovery of the peptides (possibly due to further digestion under the conditions of the tryptic digestion) and by their comigration with antibody light chains. Thus, as described below, it was necessary to address complex formation between gI and C-terminal or N-terminal gE peptides by coexpressing these combinations in CHO cells and assessing the interactions relative to that observed for gI complexes with full-length soluble gE.

2E9 is a monoclonal antibody specific for gI and for gE-gI complexes.

To facilitate analyses of complex formation between gE and gI, we raised monoclonal antibodies against gE-gI complexes by immunizing mice with 50 μg of purified soluble gE-gI complexes. The bleeds and hybridoma lines were screened by enzyme-linked immunosorbent assays (ELISA) using ELISA plates coated with 10 μg of purified soluble gE-gI/ml. Goat anti-mouse IgG conjugated to horseradish peroxidase (Bio-Rad) at a 1/1,000 dilution was used as the secondary antibody. Mouse immunoglobulins do not bind to gE-gI via the Fc domains, and goat IgG binds only weakly to gE-gI (9); thus, the screen was designed to identify antibodies with Fab (rather than Fc) specificity for a component of the gE-gI complex. Hybridoma lines that were positive by ELISA relative to control lines were further screened by immunoprecipitation with metabolically labeled CHO cells expressing either soluble gE, gI, or both (5). A wash protocol more rigorous than that previously described (5) was used (50 mM Tris-HCl, 150 mM NaCl, 0.02% NaN₃, 1 mM EDTA, 0.1% NP-40, and 0.25% gelatin, pH 7.5) in the immunoprecipitation assays in order to minimize nonspecific binding. All the hybridoma supernatants, including those from control cell lines and media alone, immunoprecipitated soluble gE-gI complexes from CHO cells expressing gE and gI, whereas none of the hybridoma lines immunoprecipitated gE from CHO cells expressing soluble gE (data not shown). Since soluble gE-gI, but not soluble gE alone, binds to the Fc domains of IgG with high affinity (5), these observations indicated that the Fc regions of bovine immunoglobulins present in the cell culture medium interacted with gE-gI complexes and interfered with the assay to identify the hybridoma lines with Fab reactivity toward a component of the gE-gI complex. One hybridoma line, 2E9, immunoprecipitated gI from CHO cells expressing gI. Ascites fluid was generated using the 2E9 hybridoma line. In metabolic-labeling experiments with this ascites fluid, 2E9 immunoprecipitated soluble gE-gI complexes from CHO cells expressing those proteins whereas ascites fluid generated using a second hybridoma line (3D3) did not (Fig. 3, fifth and eighth lanes). Thus, Fc-mediated immunoprecipitation of gE-gI complexes is not observed using antibodies contained in ascites fluids from mice, as expected from previous observations that the Fc domains of murine immunoglobulins do not interact with gE-gI complexes (9). 2E9 ascites fluid could also immunoprecipitate soluble gI from a CHO cell line expressing soluble gI alone, but not soluble gE from a cell line expressing soluble gE alone (Fig. 3, sixth and seventh lanes). 2E9 ascites fluid was used for further analyses of gI complexes with truncated gE fragments.

FIG. 3.

The 2E9 hybridoma line recognizes gI and gE-gI complexes. CHO cells expressing both soluble gE and gI, or either soluble gE or soluble gI alone, or untransfected CHO-K1 cells as negative controls were metabolically labeled with [³⁵S]methionine/cysteine (ICN Biomedicals) for 5 h. Cell supernatants were immunoprecipitated with 10 μg of rabbit anti-HSV IgG to visualize the expressed HSV proteins, or with 0.05 ml of 2E9 ascites fluid, or with 0.05 ml of a control ascites fluid (generated using the 3D3 hybridoma line). The 30-kDa band observed in the second lane is a spontaneously derived proteolytic fragment of gE, often visualized in cells that express gE.

NgE but not CgE forms a stable complex with gI.

To establish whether either the N-terminal or the C-terminal gE domain could associate with gI, we coexpressed gI along with truncation mutants of gE encoding either residues 1 to 188 appended to a hexahistidine tag (NgE) or residues 189 to 399 (CgE) (numbering corresponds to mature gE). The DNA sequence encoding HSV-1 gE in the PCRII vector (5) was modified by PCR to generate sequences encoding NgE and CgE. For NgE, the gE DNA sequence was truncated at the position corresponding to residue 188 of mature gE and a sequence encoding a hexahistidine epitope tag was appended. For generating CgE, bridge PCR was used to fuse the sequence encoding the gE leader peptide with that encoding CgE. PCR products encoding NgE and CgE were cloned individually into the unique XhoI and NotI sites of the pBJ5-GS expression vector (5). Each of these constructs was transfected into CHO cells along with the previously described pBJ5-GS-gI (5). Cells resistant to the drug methionine sulfoximine were selected as previously described (5). Cells secreting NgE and gI (NgE+gI) or CgE and gI (CgE+gI) were identified by immunoprecipitation of supernatants of metabolically labeled cells using the anti HSV-1 polyclonal antibody (ScyTek Laboratories), and clonal lines were obtained (Fig. 4, anti-HSV immunoprecipitations). CgE migrates faster than NgE on SDS-PAGE gels (Fig. 4), even though CgE is larger by 17 amino acids. We believe that this might arise due to charge differences between the two proteins that result from the introduction of a hexahistidine tag on NgE but not CgE.

FIG. 4.

NgE but not CgE forms stable complexes with gI, as shown by coimmunoprecipitation analyses with antibodies specific for gE and gI. CHO cells expressing soluble gE and gI, or cells expressing NgE and gI, or CgE and gI, were metabolically labeled with [³⁵S]methionine/cysteine (ICN Biomedicals) for 5 h. Cell supernatants were immunoprecipitated with the indicated antibodies, and proteins were visualized by SDS-PAGE and phosphorimaging analyses. Antibodies used were anti-gH (an irrelevant antibody), to assess nonspecific binding; anti-HSV, to visualize expressed HSV proteins; anti-His, which recognizes the hexahistidine epitope tag present on NgE; 1108, which recognizes an unknown epitope present in gE and CgE; and 2E9, which is directed against gI.

Using the NgE+gI and CgE+gI cell lines or the previously described cell line expressing soluble gE and gI (gE+gI) (5), we further investigated complex formation. Two different antibodies, 2E9 and anti-His (Covance), were able to coprecipitate NgE-gI complexes (Fig. 4). By contrast, 2E9 failed to coprecipitate CgE along with gI. A commercially purchased gE-specific antibody, 1108 (Goodwin Institute), which recognizes CgE, was also unable to coprecipitate gI with CgE (Fig. 4). By passing supernatants from the NgE+gI cell line over a nickel column, both NgE and gI could be isolated, as assessed by immunoblot analyses of fractions eluted from the columns with anti-His and 2E9 antibodies, respectively (data not shown). Taken together, these observations indicated that the N-terminal domain of gE contains a binding site for interaction with gI.

To compare the relative interaction propensities of gI with gE versus NgE, the intensities corresponding to gI were quantitated from 2E9-based immunoprecipitations of supernatants from the gE+gI and NgE+gI cell lines, and intensities corresponding to coprecipitating gE or NgE were normalized relative to gI. The normalization procedure took into account the different methionine/cysteine contents in the different proteins, using previously described calculations (12). Association levels were calculated as the ratio of normalized gE intensity to gI intensity or the ratio of normalized NgE intensity to gI intensity). The observed association level for the gE-gI complex was 0.24 ± 0.04, averaged over six independent experiments. The observed association level for the NgE-gI complex was slightly lower at 0.20 ± 0.07, averaged over five independent experiments. The finding that the relative association levels for NgE-gI and gE-gI are very similar indicates that NgE folds into a native-like conformation and that the affinities of gI for gE and NgE are comparable. The slightly lower values for NgE-gI may arise at least in part from the lower overall expression of NgE. Expression levels of gE and NgE were determined by quantifying and normalizing intensities corresponding to each protein relative to gI intensities in anti-HSV-1 antiserum-based immunoprecipitations of supernatants from gE+gI and NgE+gI cell lines. Based on these analyses, we estimated that nearly equal levels of gE and gI were being expressed in the gE+gI cell line (the ratio of normalized gE intensity to gI intensity was 0.89 ± 0.11 in five independent experiments). This ratio was lower in the NgE+gI cell line (the normalized NgE intensity/gI intensity ratio was 0.78 ± 0.17 in five independent experiments), correlating with the slightly reduced association levels for NgE-gI complexes. We have previously estimated the stoichiometry of the purified soluble gE-gI complex to be 1:1. The lower-than-stoichiometric recoveries of both gE and NgE relative to gI in the present experiments indicate that both the gE-gI and NgE-gI complexes are dissociating under the conditions of the immunoprecipitation experiments. It is possible that the presence of a detergent or antibody, or other conditions of the immunoprecipitation assays, reduces the stability of the gE-gI complex relative to that observed for the purified complex (5).

Our observations that CgE does not form a stable complex with gI do not preclude the possibility of CgE residues participating in gE-gI interactions; rather, our data suggest that residues contained in CgE are not sufficient to mediate a stable CgE-gI interaction. We cannot exclude the formal possibility that folding constraints in CgE account for the lack of observable interactions with gI. However, CgE peptides are recovered in high yields following tryptic digestion (Fig. 2B, peak 1), suggesting that CgE is stable as an isolated domain. Furthermore, CgE expression in the CgE+gI cell line was equal to that of gI, as assessed by quantitation and normalization of CgE and gI intensities in anti-HSV immunoprecipitation analyses of supernatants from CgE+gI cell lines (normalized CgE intensity/gI intensity was 0.98 ± 0.19 in four independent experiments). Thus, it is unlikely that misfolding of the CgE domain is responsible for the lack of association of this domain with gI.

Consistent with our observations that an N-terminal domain of gE can form stable complexes with gI are recent studies of gE-gI complexes from bovine herpesvirus 1 (BHV-1), which have indicated that residues 1 to 246 of BHV-1 gE are sufficient for complex formation with gI (18). The BHV-1 gE fragment contains residues corresponding to the entire HSV-1 NgE domain but also has a 40 residue segment corresponding to the N terminus of HSV-1 CgE. Other studies have reported that a 106-residue segment containing residues 163 to 268 of mature gE (183 to 288 of the gE precursor with the signal sequence) contains the minimal gI interaction site (1). Further linker insertion mutagenesis identified residues in the vicinity of residues 215 and 244 of mature gE (235 and 264, respectively, of the gE precursor) as being important for the gE-gI interaction (1). Our observation that NgE, truncated at residue 189, forms stable complexes with gI raises the possibility that the previously reported effects of mutations at gE residues 215 and 244 might have an effect on gE structure or folding rather than the gE-gI interaction per se.

Soluble gE-gI complexes, but not NgE-gI or the CgE+gI combination, function as an Fc receptor for IgG.

Early studies have suggested that cells transfected with genes encoding both gE and gI have enhanced IgG binding activity compared to cells transfected with gE alone (3, 6, 7, 10). Using linker insertion mutagenesis, it has also been shown that mutations in the C-terminal region of the extracellular domain of gE, at positions 215, 244, 265, 304, 313, 319, 335, 351, 360, and 369 of the mature gE sequence, could disrupt the gE-gI–IgG interaction (1). All of these residues fall within the CgE domain. Using an immunofluorescence-based monomeric IgG binding assay, Basu et al. have also reported that gE residues 5 to 397, 91 to 397, and 163 to 397, when fused between residues 244 and 246 of gD, could bind to monomeric IgG in the absence of gI (1). Thus, we compared the relative abilities of soluble gE, soluble gI, gE-gI complexes, and the CgE+gI and NgE+gI combinations to bind Fc using immunoprecipitation-based experiments with metabolically labeled proteins. The use of intact human IgG in these experiments is complicated by our observations that most commercial human IgG preparations contain low levels of anti-HSV antibodies that react with gE and gI by binding via the Fab ends (data not shown). Thus, we used human Fc (Jackson ImmunoResearch) or IgG purified from normal rabbit serum (Jackson ImmunoResearch) in our assays. Neither the CgE+gI combination nor the NgE+gI combination is able to bind human Fc or rabbit IgG under conditions where soluble gE-gI binds (Fig. 5). Additionally, neither gE alone nor gI alone shows specific binding to human Fc or rabbit IgG. Based on these observations, it appears that if soluble gE or CgE is able to bind IgG, these interactions must be significantly reduced in affinity compared to the interaction of soluble gE-gI with IgG. Thus, as previously suggested (3, 6, 7), gI plays a critical role in the binding of monomeric IgG by gE-gI complexes, either by directly participating in Fc binding or, alternatively, through an indirect effect on gE domain structure and conformation.

FIG. 5.

Soluble gE-gI complexes, but not the NgE-gI complexes or CgE+gI combination, function as an Fc receptor for IgG. gE-gI complexes have previously been shown to interact with the Fc domains of human and rabbit immunoglobulins but not mouse immunoglobulins. CHO cells expressing soluble gE and gI, or cells expressing either NgE plus gI, CgE plus gI, gE alone, or gI alone, were metabolically labeled with [³⁵S]methionine/cysteine (ICN Biomedicals) for 5 h. Cell supernatants were immunoprecipitated with either anti-HSV-1, human Fc [hIgG(Fc)], rabbit IgG (rIgG) (purified from normal rabbit serum), or mouse IgG (mIgG). Immunoprecipitated proteins were visualized by SDS-PAGE followed by phosphorimaging analyses. The 30-kDa band observed in the second panel (with the gE-expressing cell line) is a spontaneously derived proteolytic fragment of gE, often visualized in cells that express gE.

In this report we used a rational approach to design truncated gE constructs that were based on a knowledge of the domain structure of gE from limited-proteolysis experiments. When we aligned sequences of gE from several alphaherpesviruses, we found that CgE showed significantly higher sequence conservation than NgE (approximately 37 and 19% sequence identity, respectively). However, the experiments reported here indicate that it is the NgE domain rather than the CgE domain that appears to play a prominent role in gE-gI interactions. These observations raise the possibility that the more highly conserved CgE residues participate in functions pertaining to Fc binding, as previously suggested by the studies of Basu et al. (1), and also in interactions important for viral cell-to-cell spread. The studies described here will facilitate high-resolution structural analyses of gE domains and complexes with gI.

Appendix

The soluble gI used in the experiments described here was derived from the HSV-1 KOS strain as previously described (5). The sequence of gI from HSV-1(KOS) has several differences from the published strain 17 gI sequence. In the extracellular domain, these are G73→V, Q135→R, Y179→H, S189→Y, Q206→P, I221→T, P222→S, A223→T, and P248→H (residue numberings correspond to immature gI). The soluble gI construct used in the experiments described here and in previous experiments (5) lacks the last of three NNNPSTT repeat regions present in full-length gI of HSV-1(KOS), but the soluble gI sequence is otherwise identical to the extracellular domain of the gI of HSV-1(KOS). The deleted region corresponds to residues 221 to 227 of the immature sequence (residues 201 to 207 of the mature protein). Based on the results described in this report and previous results (5), residues 201 to 207 of gI do not appear to be important for the gE-gI interaction or for the gE-gI—IgG interaction.

gE used in the experiments described here was also derived from the HSV-1 KOS strain as previously described (5). The amino acid sequence of gE from HSV-1(KOS) is identical to the published strain 17 sequence. The full-length gE construct, but not the soluble gE construct that we expressed in CHO cells (5), was found to contain an A→P mutation at residue 273 (mature sequence numbering) of the extracellular domain. Both constructs, in combination with full-length and soluble gI, respectively, bound IgG with high affinity (5), suggesting that the A273P mutation on gE does not interfere with the gE-gI interaction or the gE-gI–IgG interaction. The corresponding A→P mutation is also present on CgE (residue 85 of the mature sequence) expressed in the CgE + gI cell line described here.