Probing the Flavivirus Membrane Fusion Mechanism by Using Monoclonal Antibodies

Abstract

In this study, we investigated in a flavivirus model (tick-borne encephalitis virus) the mechanisms of fusion inhibition by monoclonal antibodies directed to the different domains of the fusion protein (E) and to different sites within each of the domains by using in vitro fusion assays. Our data indicate that, depending on the location of their binding sites, the monoclonal antibodies impaired early or late stages of the fusion process, by blocking the initial interaction with the target membrane or by interfering with the proper formation of the postfusion structure of E, respectively. These data provide new insights into the mechanisms of flavivirus fusion inhibition by antibodies and their possible contribution to virus neutralization.

Flaviviruses are small enveloped viruses and comprise important human pathogens, such as yellow fever virus, dengue virus, Japanese encephalitis virus, West Nile virus, and tick-borne encephalitis virus (TBEV) (3). They enter cells by receptor-mediated endocytosis, and the low pH in the endosome triggers the fusion of the viral with the endosomal membrane (reviewed in reference 26). Fusion is mediated by the major envelope protein E, a class II viral fusion protein that is also responsible for binding to the cellular receptor and forms an icosahedral lattice on the surface of mature virus particles (13, 18).

The X-ray structures of E from TBEV, dengue viruses, and West Nile virus have been determined in their prefusion conformations by using soluble truncated forms of E (sE) that lack the so-called stem-anchor region at the C terminus (Fig. 1A, F, and G) (10, 15, 17, 19, 22, 32). The native E protein forms a homodimer on the surface of mature virions, and each monomeric subunit of sE is composed of three distinct domains (DI, DII, and DIII) (Fig. 1A). A loop located at the tip of DII and buried at the dimer interface functions as an internal fusion peptide (FP) (Fig. 1A, F, and G) (1). Slightly acidic pH, as it occurs in endosomes, triggers specific rearrangements that convert the metastable E dimer into a stable postfusion trimer (Fig. 1H) (26). The structure of this low pH form was also determined by X-ray crystallography for sE from TBEV and dengue type 2 virus (2, 16) and suggested a mechanism for flavivirus membrane fusion, as depicted in Fig. 1A to E. The most dramatic changes during fusion concern the relocation of DIII and the stem: DIII moves from its original position at the end of the monomeric subunit to the side of DI and thus allows the formation of a hairpin-like structure by “zippering” the stem along the DII interfaces of the trimer (Fig. 1C and D). The formation of such a hairpin-like postfusion structure is a common feature of different classes of viral fusion proteins, indicating mechanistic similarities in the overall fusion process despite the structural unrelatedness of the proteins involved (11, 31). In this work, we used TBEV as a model to analyze the possible mechanisms of antibody-mediated fusion inhibition by the use of monoclonal antibodies (MAbs) that react with different sites in each of the three domains of the fusion protein E. As described in earlier studies, nine of these MAbs were neutralizing (A3, A4, A5, IE3, i2, IC3, B1, B2, B4) and three were nonneutralizing (A1, A2, B3) (8, 9, 27). The binding sites of all 12 MAbs have been mapped previously by neutralization escape mutants and mutagenesis of recombinant subviral particles (references 1, 9, and 14 and unpublished results) (Fig. 1F to H). Our results provide evidence for different mechanisms of antibody-mediated fusion inhibition and the interference of antibodies with early as well as late stages of the fusion process, depending on the MAb binding sites. These data are discussed in the context of the structural transitions of E during fusion and with respect to their possible implications for virus neutralization.

FIG. 1.

Schematic model of the flavivirus fusion process (A to E) and ribbon diagrams of the sE prefusion dimer (F and G) and postfusion trimer (H) of TBEV. E protein DI, red; E DII, yellow; E DIII, blue; FP, orange; stem, purple; transmembrane anchor, green; viral membrane, blue; target membrane, gray. (A) Metastable E dimer on the surface of native virions. (B) Low pH-induced dissociation of the dimer and insertion of the FP into the target membrane. (C) Relocation of DIII leading to hairpin formation, trimerization, and “zippering” of the stem along the body of the trimer. (D) Hemifusion intermediate. Only the contacting membrane leaflets are fused. (E) Fusion pore formation. In the final postfusion conformation, the FPs and the membrane anchors are juxtaposed in the fused membrane. (F) Side view and (G) top view of the sE dimer. (H) Side view of the sE trimer. The gray balls show the positions of mutations that affected binding of MAbs. The position of the carboxy terminus of sE is indicated by a purple arrow and labeled COO−.

We first investigated the effect of the MAbs on the overall fusion process by using an in vitro fusion assay with liposomes and fluorescence-labeled purified virions (4, 28). For this purpose, TBEV (strain Neudoerfl) was grown in primary chicken fibroblasts, metabolically labeled with 1-pyrenehexadecanoic acid (Molecular Probes, Leiden, The Netherlands), and purified by two cycles of gradient centrifugation (4, 28). Comparative titrations in BHK cells with four pyrene-labeled and four unlabeled virus preparations revealed that the incorporation of 1-pyrenehexadecanoic acid into the viral membrane had no significant effect on the infectivity of the virus (unpaired two-tailed t test, P = 0.4252). The mean infectivity of the labeled samples was 10^5.11 50% tissue culture infective doses (TCID₅₀)/ml (95% confidence interval [CI], 10^4.61 to 10^5.61), and that of the unlabeled samples was 10^4.93 TCID₅₀/ml (95% CI, 10^4.53 to 10^5.32). Liposomes consisted of phosphatidylcholine:phosphatidylethanolamine:cholesterol at a molar ratio of 1:1:2 which, as shown previously (28), has proven to be optimal for TBEV fusion. Phospholipids were from Avanti Polar Lipids, Alabaster, AL; cholesterol was from Sigma-Aldrich. For fusion inhibition experiments, pyrene-labeled TBEV was preincubated overnight with MAbs at 4°C (molar ratio of MAb to E, 6:1) in 10 mM HEPES, 140 mM NaCl, and 0.1% bovine serum albumin, pH 7.4, and then mixed with liposomes in a quartz cuvette of an LS50B fluorimeter (PerkinElmer) at 37°C with continuous stirring. The samples were acidified to pH 5.4 with 250 mM morpholineethansulfonic acid (MES), and the decrease in pyrene excimer fluorescence was recorded continuously for 60 s (4, 28). The 6:1 molar ratio of MAb to E was derived from standardization experiments with different concentrations of the strongly fusion-inhibiting MAb A3, which revealed that this ratio was the threshold for achieving complete fusion inhibition under the conditions used. In order to allow direct comparisons of the specific fusion-inhibiting activities of the different MAbs, the same molar ratio of antibody to E was used in all of the assays.

The presence of MAbs during the fusion reaction resulted in four distinct patterns of interference, as displayed in Fig. 2. Complete inhibition was observed with MAbs A3, A4, and IC3 (Fig. 2A), whereas B2 and B3 did not have a significant effect on the fusion reaction (Fig. 2B). All of the other MAbs had an incomplete inhibitory effect, but the resulting fusion curves were shaped differently: A1, A2, and A5 reduced the extent of fusion without impairing the initial rate of fusion (Fig. 2C); IE3, i2, B1, and B4, in contrast, affected both the rate and the extent of fusion during the 1-min observation period of the assay (Fig. 2D).

FIG. 2.

Low pH-induced fusion of pyrene-labeled TBEV in the absence (No mab) or presence of MAbs in a pyrene excimer fusion assay. Pyrene-labeled TBEV virions were preincubated with each of the MAbs overnight at 4°C, liposomes were added, the mixture was acidified, and the change in pyrene excimer fluorescence was monitored continuously for 60 s. The fusion curves were averaged, and the data shown are from at least three independent experiments. (A) Abolition of fusion (MAbs A3, A4, IC3). (B) No significant effect on fusion (MAbs B2, B3). (C) Reduction of the extent of fusion (MAbs A1, A2, A5). (D) Reduction of both rate and extent of fusion (MAbs IE3, i2, B1, B4).

The different patterns revealed in these experiments can be the result of differences in avidity but can also reflect different mechanisms of fusion inhibition by individual MAbs, possibly by interference with distinct steps of the fusion process (Fig. 1A to E). We have addressed this mechanistic question by investigating whether the binding of the MAbs already impaired the initiation of fusion, i.e., the interaction of the FP with target membranes. For this purpose, we analyzed the effect of each of the MAbs on acidic pH-induced coflotation of virions with liposomes, as described previously (24, 30). Briefly, TBEV was preincubated with MAbs, mixed with liposomes, and acidified as described for the fusion experiments. After 10 min at pH 5.4 and 37°C, the samples were back-neutralized with 150 mM triethanolamine, adjusted to 2-ml 20% sucrose (wt/wt) in 50 mM HEPES and 100 mM NaCl (pH 7.4) (HN buffer), applied to centrifuge tubes with a 50% (wt/wt) sucrose cushion, and overlaid with 5% (wt/wt) sucrose in HN buffer. Centrifugation was carried out for 1.5 h at 50,000 rpm at 4°C in a Beckman SW 55 rotor, and the fractions were collected by upward displacement (24).

Figure 3A shows representative examples of the patterns obtained (B2, no effect; A3, strong inhibition; A2, intermediate inhibition), and the quantitative evaluation of all experiments is displayed in Fig. 3B. Most of the MAbs had either no or only minor effects on coflotation (less than 50% inhibition), which can be explained by the fact that their binding sites in E are relatively distant from the FP loop (B1 to B4, DIII; i2 and IC3, DI; A4, A5, and IE3, DI-proximal part of DII) (compare to Fig. 1). The strongest inhibition of liposome binding (about 90%) was observed with MAb A3, consistent with its binding site in the DI-distal part of DII, in the vicinity of the FP loop (Fig. 1F and G).

FIG. 3.

Low pH-induced coflotation of TBEV with liposomes in the absence or presence of MAbs. TBEV virions were preincubated with each of the MAbs overnight at 4°C, liposomes were added, and the mixture was acidified, back-neutralized, and then subjected to centrifugation in sucrose step gradients. (A) Representative examples of coflotation experiments. From left, panel 1, coflotation without MAb (No mab) at pH 5.4 (solid line) and pH 7.4 (dotted line). Panels 2 to 4, coflotation in the presence of MAbs at pH 5.4. B2 (panel 2; no effect), A3 (panel 3; strong effect), and A2 (panel 4; intermediate effect). (B) Results of coflotation experiments with each of the MAbs, expressed as percentages of E protein bound to liposomes at pH 5.4 in comparison to the control without a MAb. The data are the averages from at least two independent experiments; the error bars represent the standard errors of the means.

The intermediate results obtained with MAbs A1 and A2 (about 50% fusion inhibition and reduction of liposome binding) were puzzling. On the one hand, these antibodies are known to react with the FP directly (Fig. 1) (1, 27) and therefore would have been expected to have the strongest inhibitory activities in both assays. On the other hand, the antibody-blocking experiments shown in Fig. 4A revealed an almost complete lack of activity of A1 and A2 with native virions, in contrast to the control MAbs A4, B2, and B4. Consistent with the previously described cryptic nature of the corresponding epitopes (27), both antibodies were thus barely capable of reacting with native virion preparations in solution as used in our assays. Therefore, even the intermediate activities observed here were unexpected. A possible interpretation of these results could be based on the fact that in the course of E dimer dissociation during acidification of the virus-antibody mixtures, the FP loops (and thus the epitopes recognized by A1 and A2) are exposed transiently and for a very short period of time before they become buried again through their insertion into a target membrane or, in the case of those not directly involved in the fusion reaction, into the viral membrane (25, 30). In addition, the FP is also shielded by the stem-anchor interactions with domain II that accompany the formation of the final trimeric postfusion structure (Fig. 1) (29). The change in accessibility of the FP for interactions with antibodies would therefore be restricted to a limited time window only under the conditions of the assay. For testing this interpretation, we simulated the same conditions as used in the fusion and liposome coflotation assays in a four-layer enzyme-linked immunosorbent assay (ELISA) without detergent (27) and analyzed the binding of A1 to TBE virions when allowed to react (i) at pH 7.4, (ii) simultaneously with acidification at pH 5.5, and (iii) after acidification at pH 5.5. The data displayed in Fig. 4B show a significantly higher reactivity of A1 when it is present at the time of acidification than that of the pH 7.4 control (paired two-tailed t test, P = 0.0005), but not when the MAb is added after acidification (paired two-tailed t test, P = 0.3106). These results are fully consistent with (i) the shielding of the FP in low pH-treated virions, presumably by the formation of postfusion trimers and FP insertion into the viral membrane, and (ii) the proposed transient time slot of epitope exposure as an explanation for the incomplete inhibition of fusion and coflotation by the FP-specific MAbs A1 and A2.

FIG. 4.

(A) Blocking ELISA in the absence of detergent with native virions and MAbs A1, A2, A4, B2, and B4 as described previously (27). A predetermined fixed dilution of the respective MAb was incubated with decreasing concentrations of native TBEV. The mixture was then added to microtiter plates that had been coated with purified virus at a concentration of 0.5 μg/ml, a procedure that leads to the exposure of the FP loop, allowing its reaction with FP-specific MAbs (27). Antibody that was not blocked by the antigen in solution bound to the solid-phase antigen and was detected using a peroxidase-labeled rabbit anti-mouse immunoglobulin G (27). Results are expressed as percentages of the absorbance value obtained with each MAb in the absence of a blocking antigen. The data are representative of results of at least two independent experiments. (B) Four-layer ELISA with TBEV and MAb A1 to analyze the transient exposure of the A1 binding site upon acidification. Native TBEV in phosphate-buffered saline (pH 7.4; protein concentration, 0.5 μg/ml) was captured by polyclonal anti-TBEV immunoglobulin G for 1 h at 37°C as described previously (27). A1 epitope exposure was tested with the following combinations of pH and biotin-labeled MAb: column A, A1 was added in phosphate-buffered saline (pH 7.4); column B, A1 was added in MES buffer (pH 5.5; 50 mM MES, 100 mM NaCl); column C, the captured virus was exposed to MES buffer (pH 5.5) for 10 min followed by MAb A1 in the same buffer. After incubation for 1 h at 37°C, bound antibodies were detected by using streptavidin-peroxidase (Sigma-Aldrich). The data are the averages from five independent experiments performed in duplicate, and the error bars represent the standard errors of the means.

Since they do not or only inefficiently inhibit coflotation (Fig. 3), the fusion-inhibitory activity of the other MAbs is most likely caused by an interference with the formation of the hairpin-like postfusion structure. An inspection of the location of the binding sites in the pre- and postfusion conformations suggests a grouping into categories of antibodies that may differ with respect to their mechanisms of fusion inhibition. (i) MAbs binding to epitopes that apparently become buried and part of the interfaces in the E trimer (MAbs A4, i2, and IC3) (Fig. 1H) most likely prevent the proper formation of the postfusion structure (including the “stem”) that is essential for full fusion to occur. (ii) The binding sites of several MAbs, however, appear to be surface exposed, in both pre- and postfusion conformations (Fig. 1F to H). Possible mechanisms of fusion inhibition in these cases include interference with (i) the relocation of DIII as a prerequisite for hairpin formation (MAbs B1 and B4), (ii) the proper positioning of the stem along DII, which is required for “zippering” as shown in Fig. 1C and D (MAb A5), and (iii) during fusion, cooperative interactions of E homotrimers that have been hypothesized to play a role in fusion pore enlargement (11). We have attempted to further define the mechanistic basis of fusion inhibition by these MAbs more directly by analyzing their effects on E trimerization. Such analyses first required the dissociation of the high-molecular-weight antibody-virus complexes formed by the application of pH values of <3.0 or >10.0. Both of these treatments, however, also change the oligomeric state of E, and conclusive results as to the oligomeric state of E in the presence of the bound MAbs could not be obtained. The MAbs B2 and B3 provide examples demonstrating that in some instances, fusion may not be affected at all by the presence of MAbs bound to the E protein.

Fusion inhibition is clearly a mechanism of virus neutralization with enveloped viruses that fuse at the plasma membrane, such as human immunodeficiency virus (33). However, even in the case of viruses that are taken up by receptor-mediated endocytosis and fuse from within endosomes, the inhibition of the fusion process by antibodies may contribute to virus neutralization (5, 12). In such instances, it has to be assumed that the antibody does not or only incompletely blocks receptor interactions and (at least at the concentrations applied) allows the internalization of the virus-antibody complex. Evidence for such a mechanism has indeed been provided for the neutralization of influenza virus (6, 21, 23) and a flavivirus (West Nile virus), both with polyclonal (7) and monoclonal antibodies (20).