Immunodominance and Functional Activities of Antibody Responses to Inactivated West Nile Virus and Recombinant Subunit Vaccines in Mice

Abstract

Factors controlling the dominance of antibody responses to specific sites in viruses and/or protein antigens are ill defined but can be of great importance for the induction of potent immune responses to vaccines. West Nile virus and other related important human-pathogenic flaviviruses display the major target of neutralizing antibodies, the E protein, in an icosahedral shell at the virion surface. Potent neutralizing antibodies were shown to react with the upper surface of domain III (DIII) of this protein. Using the West Nile virus system, we conducted a study on the immunodominance and functional quality of E-specific antibody responses after immunization of mice with soluble protein E (sE) and isolated DIII in comparison to those after immunization with inactivated whole virions. With both virion and sE, the neutralizing response was dominated by DIII-specific antibodies, but the functionality of these antibodies was almost four times higher after virion immunization. Antibodies induced by the isolated DIII had an at least 15-fold lower specific neutralizing activity than those induced by the virion, and only 50% of these antibodies were able to bind to virus particles. Our results suggest that immunization with the tightly packed E in virions focuses the DIII antibody response to the externally exposed sites of this domain which are the primary targets for virus neutralization, different from sE and isolated DIII, which also display protein surfaces that are cryptic in the virion. Despite its low potency for priming, DIII was an excellent boosting antigen, suggesting novel vaccination strategies that strengthen and focus the antibody response to critical neutralizing sites in DIII.

The availability of high-resolution structures is a prerequisite for understanding structural determinants of immunogenicity and immunodominance. Knowledge of factors that control and/or influence these properties of antigens can lead to an improvement of existing vaccines and the rational design of new vaccine antigens and regimens (13). Flaviviruses are among those human-pathogenic viruses for which detailed structural information is available through the combined use of X-ray crystallography and cryo-electron microscopy (cryo-EM) (25, 30, 36-38, 40, 49, 65). The most important representatives are the mosquito-borne yellow fever (YF), dengue (DEN), Japanese encephalitis (JE), and West Nile (WN) viruses, as well as tick-borne encephalitis (TBE) virus (14). These viruses have a significant impact on public health and the potential for emergence in new geographic regions, as exemplified by the expansion of WN virus in the Americas since its first introduction into the United States in 1999 (17). Human vaccines are in general use against YF virus (live attenuated), JE virus (live attenuated and inactivated whole virus), and TBE virus (inactivated whole virus) but not yet against DEN and WN viruses (48).

Mature flavivirions are composed of an isometric capsid containing the positive-stranded RNA genome and a lipid envelope with two membrane-associated proteins, E and M (33). As revealed by cryo-EM, 90 copies of E-protein dimers (oriented parallel to the viral membrane) form a tight shell at the virion surface in a herringbone-like arrangement (30, 38) (Fig. 1 A). Because of its dual function in receptor binding and acid pH-induced membrane fusion (33), E is the major target of virus-neutralizing antibodies that mediate protection and long-lived immunity, after both natural infection and vaccination (12, 46, 47, 50). Each of the monomeric subunits of E contains three distinct domains, designated domain I (DI), DII, and DIII (Fig. 1B). The central domain, DI, is flanked by DII, carrying the highly conserved internal fusion peptide (FP) loop, and by DIII, which has an immunoglobulin-like fold. DIII protrudes slightly off the relatively smooth surface of mature virions and has been implicated in receptor binding (6, 9, 32). The rest of the molecule, which is absent from the crystal structures, comprises the so-called stem that follows DIII and leads to the double membrane-spanning anchor at the C terminus of the molecule.

FIG. 1.

Immunogens and immunization schedules. (A) Schematic model of a flavivirus virus particle based on cryo-EM reconstructions of dengue and WN viruses (30, 38). The viral surface is covered by a densely packed shell consisting of 90 copies of E-protein dimers which are arranged in a herringbone-like lattice, consisting of 30 rafts of three E dimers. One of these rafts, an E dimer, and an E monomer are depicted separately at the bottom of this panel. (B) Ribbon diagrams (top and side views) of the WN virus soluble E-protein monomer (40) (Protein Data Bank accession number 2HG0). (C) Ribbon diagrams (top and side view) of the WN virus DIII (40) (Protein Data Bank accession number 2HG0). In panels A, B, and C, DI, DII, and DIII of E are shown in red, yellow, and blue, respectively. The FP loop at the tip of DII is highlighted in orange, the surface-exposed DIII lateral ridge epitope (DIII-lr) in cyan, disulfide bridges in green, and the carbohydrate side chain (CHO) in gray. (D) Analysis of the immunogens used in the study by SDS-PAGE (15% gel) and Coomassie staining. Lane 1, formalin-inactivated purified WN virus; oligomeric bands of C and E are a result of protein cross-links as a result of formalin treatment; lane 2, recombinant WN virus sE; lane 3, recombinant WN virus DIII. (E) Schematic of the immunization schedule with virion (inactivated purified virus), sE, and DIII. With each of these antigens, a total of 30 mice (divided into groups of 10) were immunized with two doses at an interval of 4 weeks. Blood samples were taken 8 weeks after the second immunization, and the sera of each group of 10 mice were pooled for further analyses. (F) Schematic of the immunization schedule for eight different prime-boost regimens. The schedule consisted of two primary immunizations at an interval of 4 weeks, followed by two booster immunizations after 8 weeks, again at an interval of 4 weeks. For each regimen, 10 mice were used. Blood samples were taken 8 weeks after the second primary immunization as well as 2 weeks after the second booster immunization and were pooled for further analyses. In panels E and F, black arrows indicate time points of immunization and red arrows indicate time points of blood sampling.

Epitopes involved in virus neutralization have been identified in each of the three domains, at sites exposed at the virion surface (3, 10, 16, 20, 29, 41, 44, 50, 51, 58, 59, 63). Studies with mouse monoclonal antibodies (MAbs) suggest that the most potent neutralizing antibodies are virus type specific and bind to the highly variable upper lateral ridge of DIII (16, 42). Antibodies to the conserved FP loop, on the other hand, are broadly flavivirus cross-reactive but display comparatively low neutralizing activity (44, 57).

A significant degree of heterogeneity in the specificities of antibody populations was observed with mouse, human, and horse postinfection and/or postimmunization sera (11, 31, 43, 52, 59, 62). Antibodies to DIII made up a significant fraction of the IgG response in mice but only a small proportion in humans, in whom the response was dominated by antibodies to the FP loop (11, 31, 43, 59). Factors that control the immunodominance of antibody responses are ill defined but, in addition to host-specific factors, may be influenced by the physical organization of the antigen and differences in the presentation of antigenic protein surfaces during immunization. The specific design and selection of antigens can also provide clues to the development of vaccines against highly variable pathogens, such as HIV and influenza viruses (15, 55, 61).

Considering the many unresolved questions of immunodominance, we conducted a mouse immunization study, using the WN virus system, in which we investigated whether and in which way the fine specificity of antibody responses to E, especially the contribution of DIII antibodies, was influenced by the quaternary organization of the immunogen, i.e., E in its isolated monomeric form (soluble protein E [sE]) compared to E tightly packed in whole inactivated virus. We also assessed whether it is possible to specifically focus and strengthen the antibody response to DIII by (i) boosting with isolated DIII after priming with inactivated virion or sE and (ii) priming with isolated DIII followed by boosting with inactivated virion or sE. We demonstrate that inactivated virions induced DIII-reactive antibodies with higher specific neutralizing activity than sE and that in both cases antibodies to this domain dominated the total neutralizing activity. We also show that the isolated DIII is an excellent booster antigen which restimulates DIII-specific and neutralizing antibodies after priming with virion or sE more strongly than the homologous priming antigens. Paradoxically, priming with DIII followed by boosting with virion or sE had no beneficial effect but, in contrast, resulted in dampening of the antibody response. Our data suggest that the neutralizing immune response to flavivirus vaccines may be focused and optimized, after appropriate priming, by booster vaccinations with preparations of recombinant DIII.

MATERIALS AND METHODS

Production of purified inactivated WN virus.

Virus production was carried out essentially as described in reference 21 for tick-borne encephalitis virus. In brief, primary chicken embryo cells were infected with WN virus strain NY99 (GenBank accession no. AF196835). At 24 h postinfection, the cell supernatant was harvested, clarified by centrifugation, and inactivated with formalin (1:2,000) for 24 h at 37°C. The inactivated virus was concentrated by ultracentrifugation and purified by rate zonal centrifugation, followed by equilibrium sucrose density gradient centrifugation. SDS-PAGE analysis indicated the presence of a small proportion of uncleaved prM, as is characteristic for purified preparations of WN virus (8) and for flaviviruses in general (24), as well as oligomeric bands of C and E as a result of formalin cross-linking (Fig. 1D).

Expression and purification of WN and TBE virus sE.

Recombinant E proteins from WN virus strain NY99 and TBE virus strain Neudoerfl (GenBank accession no. U27495) were expressed in Schneider 2 (S2) cells using the pMTBip/V5-His vector (Invitrogen) and inserts containing the coding sequences for prM and E (C-terminally truncated after amino acid 400). For immunization, the WN virus sE was produced without a His tag by the introduction of a stop codon after amino acid 400. Blasticidin resistance was used for the selection of stably transfected cells, according to the manufacturer's instructions. Expression was induced by CuSO₄, and the supernatants were harvested 7 to 11 days postinduction. Recombinant proteins were purified by immunoaffinity chromatography using the flavivirus cross-reactive monoclonal antibody 4G2 (ATCC HB-112). The purity of the recombinant sE proteins was >95%, as determined by SDS-PAGE and densitometric analysis (Fig. 1D).

Expression and purification of WN virus recombinant DIII.

DIII of WN virus strain NY99, encompassing amino acids 300 to 399 of E, was expressed with and without a C-terminal His tag (including a stop codon after amino acid 399) in Escherichia coli strain BL21 using the pET 32a Xa/LIC vector (Novagen) as a fusion protein with thioredoxin. For purification of soluble DIII without a C-terminal His tag, the fusion protein in clarified cell lysates was bound to a Ni²⁺-immobilized affinity column (GE Healthcare Life Sciences) via the internal His tag located between thioredoxin and DIII, and DIII was eluted from the column after proteolytic cleavage from thioredoxin by factor Xa protease. For expressing C-terminally His-tagged DIII, the internal His tag was deleted from the expression cassette so that the fusion protein bound to the affinity column via the C-terminal His tag of DIII. The thioredoxin fusion partner was cleaved off by factor Xa and removed by washing the column, and DIII was specifically eluted by an imidazole gradient. Final purification was accomplished by anion-exchange chromatography using Sepharose Q columns (GE Healthcare Life Sciences), according to the manufacturer's instructions. By SDS-gel analysis, a single band corresponding to DIII was visible by Coomassie staining (Fig. 1D).

Mouse immunization experiments.

Mouse experiments were carried out in strict accordance with the guidelines of the Federation of European Laboratory Animal Science Associations (FELASA) and Austrian federal law. The protocol was approved by the ethics committee of the Medical University of Vienna and the Austrian Federal Ministry of Science and Research (permit number BMWF-66.009/0068-II/10b/2008). Six- to 8-week-old C57BL/6 mice (Charles River Laboratories) were immunized subcutaneously with inactivated virus, sE, or DIII adsorbed to 0.2% Al(OH)₃ (aluminum hydroxide gel; Sigma) in 100 μl buffer containing 0.05 M triethanolamine, and 0.1 M NaCl, pH 8.0, at doses of 1 μg, 5 μg, and 20 μg per immunization, respectively. Blood was taken from the tail vein at the time points indicated in Fig. 1E and F, using microvette 200 capillaries (Sarstedt). Antibody analyses were carried out using pools of sera from the groups of mice with the same immunization schedule.

Virion enzyme-linked immunosorbent assay (ELISA).

Microtiter plates were coated with formalin-inactivated purified WN virus at a concentration of 0.35 μg/ml in carbonate buffer, pH 9.6, at 4°C. Serial 3-fold dilutions of pooled mouse serum, starting at a dilution of 1:100, were added to the plates and the plates were incubated for 1 h at 37°C. As negative controls, sera from naïve mice were included. The bound antibodies were then detected using peroxidase-labeled rabbit anti-mouse immunoglobulin G as described in reference 57. Titration curves were established using Prism (version 5) software (GraphPad Software Inc., San Diego, CA), and serum titers were defined at an absorbance (490 nm) cutoff of 0.35, as deduced from the analysis of negative-control sera. Each serum pool was tested in at least three independent experiments, and results are presented as the geometric means of these repetitive experiments.

Analysis of avidity in virion ELISA.

Samples were analyzed in the standard virion ELISA, except that an additional washing step with and without 6 M urea in the washing buffer was applied after serum incubation, essentially as described previously (56). Titers were determined at an absorbance cutoff of 1.0 after curve fitting using a four-parameter logistic regression (Prism software, version 5.0; GraphPad Software Inc.). Relative avidities are expressed as the percentage of the titer obtained in the presence of urea compared to the titer obtained in the absence of urea.

sE and DIII ELISAs.

Microtiter plates were coated with rabbit anti-His tag IgG antibody (QED Bioscience) at 4°C in carbonate buffer, pH 9.6. DIII-His and sE-His proteins were added at concentrations of 1 and 0.5 μg/ml, respectively, and the plates were incubated for 1 h at 37°C. Threefold serial dilutions (starting at 1:100) of pooled mouse serum were then added for 1 h at 37°C. As negative controls, sera from naïve mice were included. The bound antibodies were detected using peroxidase-labeled goat anti-mouse immunoglobulin G (Pierce) for 1 h at 37°C. Visualization of bound antibodies and determination of serum titers were done as described for the virion ELISA.

WN virus neutralization assay.

WN virus neutralization assays were carried out in microtiter plates using Vero cells (ECACC accession no. 84113001). Pooled mouse sera were heat inactivated at 56°C for 30 min, and 2-fold serial dilutions, starting at a dilution of 1:20, were incubated with 20 to 40 50% tissue culture infective doses of WN virus strain NY99 for 1 h at 37°C. The virus-antibody mixture was then transferred onto Vero cells, and the cells were incubated for 4 days at 37°C. The cells were fixed with 4% paraformaldehyde for 30 min at room temperature. After fixation, the cells were blocked and permeabilized with Tris-buffered saline (TBS; 50 mM Tris and 150 mM NaCl, pH 7.6, containing 3% nonfat dry milk, 0.5% Triton X-100, and 0.05% Tween 20) for 30 min at 37°C, and a WN virus-specific rabbit serum was added for 1.5 h at 37°C. Bound antibodies were detected with alkaline phosphatase-labeled anti-rabbit IgG (Sigma) and SigmaFast pNNp (Sigma) as a substrate. The enzymatic reaction was stopped with 1.5 N NaOH, and the absorbance was measured at 405 nm. Titers were determined after curve fitting using a four-parameter logistic regression (Prism software, version 5.0; GraphPad Software Inc.) and a cutoff 90% reduction of the absorbance in the absence of antibody (90% neutralization titer [NT₉₀]). A cutoff 50% reduction (NT₅₀) was not as robust in the lower-titer range, since some negative sera would have yielded false-positive titers between 10 and 20. Each serum pool was tested in at least two independent experiments.

Antibody depletion using WN virus DIII and TBE virus sE.

One microgram of His-tagged WN virus DIII or TBE virus sE was incubated with 1 mg paramagnetic Dynabeads His-Tag Isolation and Pull Down beads (Invitrogen) for 30 min at room temperature on an orbital shaker (1,000 rpm). After pelleting of the beads by magnetic force, the beads were resuspended in pull-down buffer, according to the manufacturer's instructions, and incubated for 1 h at 37°C with a 1:5 dilution of serum pools in phosphate-buffered saline (PBS). The beads were pelleted again by magnetic force, and the depleted serum was collected. To achieve quantitative depletion, this procedure was performed three times. Lack of nonspecific binding of antibodies to the beads was confirmed by the incubation of mouse immune sera with unloaded beads and beads loaded with an unrelated His-tagged protein.

Antibody depletion using WN virus.

Ten micrograms of inactivated WN virus was incubated with pooled sera (diluted 1:50 in PBS, pH 7.4, and 0.1% bovine serum albumin) for 1 h at 37°C. The virus-antibody complexes were pelleted by ultracentrifugation in a Beckman Ti90 rotor at 50,000 rpm at 4°C for 1 h. The supernatant was collected and used for further analyses.

Statistical analyses.

Data were analyzed with GraphPad Prism software, version 5. Two-tailed t tests were used to compare antibody titers, ratios of antibody titers, and ELISA avidities. Differences were considered significant when the P value was less than 0.05.

RESULTS

Characteristics of immunogens.

Three different antigens of WN virus were used in our immunization studies (Fig. 1A to D) (see also Materials and Methods). The first antigen was highly purified formalin-inactivated virus. As revealed by cryo-EM studies with tick-borne encephalitis virus (6a; unpublished data), formalin inactivation does not affect the specific herringbone-like arrangement of 90 E-protein dimers at the surface of mature virions. The second antigen was purified recombinant sE produced in S2 cells. Recombinant sE was shown to be a monomer by cross-linking and sedimentation analyses. Evidence for correct folding was obtained by the demonstration of the presence of S-S bridges by SDS-PAGE under reducing and nonreducing conditions (data not shown). The third antigen was purified recombinant DIII produced in bacteria. The presence of the single disulfide bridge was confirmed by mass spectrometry (data not shown).

All three antigen preparations displayed similarly strong reactivities with conformation-sensitive MAbs E16 and E24 (kindly provided by Michael S. Diamond [42]) in the virion as well as the sE and DIII ELISAs, as described in Materials and Methods (data not shown).

Consistent with previous data (19), preliminary experiments had revealed that the inactivated virus had an at least 6-fold higher specific immunogenicity per μg of antigen than sE (data not shown). DIII alone did not yield significant titers even when it was used at a dose of 50 μg. Therefore, in order to obtain measurable antibody responses and to be able to assess modulations in fine specificity and immunodominance, Al(OH)₃ was used as an adjuvant and different antigen doses per mouse were applied at each immunization as follows: inactivated virion, 1 μg; sE, 5 μg; DIII, 20 μg.

Antibody response to virion, sE, and DIII.

For investigating the influence of the immunogens' structural context (i.e., the same antigen tightly packed in virions or in soluble form) on fine specificity, dominance, and functional activity of the antibody response, we immunized groups of mice twice with whole inactivated virus, the isolated sE, and the isolated DIII (see the immunization schedule in Fig. 1E). Eight weeks after the second immunization, blood samples were taken and pooled sera were analyzed in neutralization assays as well as in ELISAs using inactivated virus, sE, DIII, and the heterologous TBE virus sE as antigens (Fig. 2). Both the inactivated virion and sE induced neutralizing antibodies, but the titer was higher after virion immunization, despite the fact that the dose applied was only one-fifth of that of sE (Fig. 2A). DIII, on the other hand, induced antibodies that were only partially neutralizing (NT₅₀ titer, 35), and the extrapolation of the neutralization curve indicated that the NT₉₀ titer was below 10. In ELISAs, the sera exhibited strikingly different patterns of reactivity with virion, sE, DIII, and TBE virus sE (Fig. 2B), suggestive of differences in the fine specificities of the antibody responses obtained. Most importantly, the E protein in the context of whole virions induced a lower proportion of DIII antibodies, relative to the total amount of virion-reactive antibodies, than the isolated sE (Fig. 2C).

FIG. 2.

Serum antibody titers obtained after immunization with virion, sE, and DIII. (A) Virus neutralization titers (NT₉₀); (B) ELISA titers against different antigens, as indicated by the color codes in the inset; (C) relative contents of DIII antibodies, displayed as the ratio of DIII-specific versus virion-specific ELISA titers. The data presented were derived from the analysis of pooled sera obtained from three groups of 10 mice per antigen. The figure displays the means of the results obtained with these three serum pools. The error bars represent the standard errors of the means.

Contribution of DIII-specific antibodies to virus neutralization.

Because of the different ELISA ratios obtained after immunization with virion and sE (Fig. 2B and C) and the fact that the most potent neutralizing MAbs described so far are directed to DIII (12, 46), we investigated the contribution of DIII-specific antibodies to neutralization by their removal with recombinant DIII bound to magnetic beads (Materials and Methods). As shown by DIII ELISA, this procedure resulted in the virtually complete depletion of DIII-reactive antibodies from the serum pools (Fig. 3 A); in the virion ELISA, the drop of reactivity was 32% with the virion sera but as high as 76% with the sE sera (Fig. 3B). In both instances, however, about 75% of the neutralizing activity was removed and was thus attributable to DIII-reactive antibodies (Fig. 3C). This discrepancy reflects significant differences in the specific neutralizing activities of DIII-reactive antibodies induced by virion and sE, expressed as the ratio of DIII-associated NT versus the DIII ELISA titer × 100 (Fig. 3D). These values were 1.5 and 0.4 after immunization with virion and sE, respectively, and <0.1 after immunization with DIII (no measureable NT₉₀ titer). One possible explanation of these findings would be that E in its soluble form and, even more so, the isolated DIII would also induce antibodies to sites that are not highly exposed at the virion surface and thus contribute less to virus neutralization (46).

FIG. 3.

ELISA and NT assay analysis of virion, sE, and DIII postimmunization sera after depletion with DIII. (A) Percent DIII-reactive antibodies in postdepletion sera, determined by DIII ELISA; (B) percent virion-reactive antibodies in postdepletion sera, determined by virion ELISA; (C) percent virus neutralizing antibodies in postdepletion sera, determined by neutralization assay. Results are expressed as the percentages of the titers before depletion. The DIII-induced neutralizing antibody titer was already <10 before depletion (Fig. 2A). The numbers above the bars in panels A, B, and C indicate the calculated percentage of the contribution of DIII antibodies to total reactivity. (D) Ratio of DIII-associated neutralizing titers versus DIII-specific ELISA titers × 100 as a measure of the specific neutralizing activity and thus the functional quality of DIII-reactive antibodies. The DIII-associated neutralizing titers were deduced from the data in Fig. 2A and 3C and were calculated to be 123 and 58 for virion and sE sera, respectively. *, the NT₉₀ titer of DIII-induced antibodies was <10, and therefore, no exact ratio could be calculated. The error bars represent the standard errors of the means obtained in three independent determinations.

To corroborate this hypothesis and to identify such antibodies, we depleted the three serum pools with whole inactivated virus and analyzed the resulting sera in virion and DIII ELISAs (Fig. 4 A and B). With all three samples, the virus-reactive antibodies were almost completely removed by the procedure used (Fig. 4A). The DIII ELISA revealed that about 50% of the antibodies induced by DIII did not react with the virus in solution (Fig. 4B). Unexpectedly however, no difference between the virus- and sE-induced DIII-specific antibodies was found in this assay (Fig. 4B). A possible additional explanation of the lower functional activity of DIII-specific antibodies induced by sE and DIII compared to the activity of DIII-specific antibodies induced by the virion could be related to differences in avidity for the virus. Such differences would be obscured in the depletion procedure because of the high excess of virus. We therefore analyzed the relative avidities of the sera in virion ELISAs using a urea wash step (see Materials and Methods). As displayed in Fig. 5, significant differences in the expected order were found, with the highest avidity being for the antibodies induced by virus, followed by those induced by sE and DIII.

FIG. 4.

ELISA analysis of virion, sE, and DIII postimmunization sera after depletion with virion. (A) Percent virion-reactive antibodies in postdepletion sera, determined by virion ELISA; (B) percent DIII-reactive antibodies in postdepletion sera, determined by DIII ELISA. Results are expressed as the percentages of the titers before depletion. The error bars represent the standard errors of the means obtained in three independent determinations.

FIG. 5.

Relative avidities of virion, sE, and DIII postimmunization sera determined in virion ELISA with or without urea. The error bars represent the standard errors of the means obtained in three independent determinations.

Induction of broadly cross-reactive antibodies.

Broadly cross-reactive antibodies directed to the highly conserved FP loop can make up a significant proportion of the polyclonal immune response to flaviviruses (11, 31, 43). For determining such antibodies, we analyzed the postimmunization sera in an ELISA with the heterologous TBE virus sE as an antigen. Overall, the WN virion and WN virus sE postimmunization sera displayed only very low cross-reactive titers in this assay (Fig. 2B), and depletion with TBE virus sE did not measurably diminish their reactivities with WN virus sE (data not shown). There was also no evidence for cross-neutralization in assays with TBE virus (data not shown). As expected, due to the lack of the conserved FP site, WN virus DIII did not induce any detectable cross-reactive antibodies (Fig. 2B), consistent with the findings of previous studies on the antigenic specificity of DIII (4).

Modulation of immunodominance by prime-boost regimens with DIII.

In order to increase and strengthen the antibody response toward DIII and concomitantly the neutralizing potency of postimmunization sera, we conducted two sets of experiments: (i) mice primed with either virion or sE were boosted with DIII, and (ii) mice were primed with DIII and boosted with virion or sE (compare the immunization schedules in Fig. 1F). The resulting sera were pooled and analyzed in virion- and DIII-specific ELISAs as well as in virus neutralization assays. As shown in Fig. 6, DIII proved to be an excellent booster antigen, despite its low immunogenicity when it was used as a single immunogen (Fig. 2A and B). In combination with inactivated virus as well as with sE, it strongly boosted DIII- and virus-reactive antibodies in ELISA (Fig. 6A, B, D, and E) in association with a strong increase of neutralizing activity (Fig. 6C and F). This effect was more pronounced with virion-primed mice than with sE-primed mice (Fig. 6A to C and D to F, respectively) and resulted in significantly higher neutralization titers compared to those for the controls, which received the inactivated virus for priming as well as for boosting (Fig. 6C).

FIG. 6.

ELISA and NT assay analysis of postimmunization sera using virion or sE for priming and DIII for boosting. (A to C) Priming (P.) with virion and boosting (B.) with DIII; (D to F) priming (P.) with sE and boosting (B.) with DIII. Postimmunization sera were analyzed in virion ELISA (A and D), DIII ELISA (B and E), and NT assay (C and F). The error bars represent the standard errors of the means obtained in at least three independent determinations.

The alternative schedule, i.e., priming with DIII and boosting with virion or sE, had a quite unexpected outcome. In both instances, preimmunization with DIII resulted in an impairment of the antibody response, measured in ELISA and neutralization assay, compared to that for the unprimed controls (Fig. 7).

FIG. 7.

ELISA and NT analysis of postimmunization sera using DIII for priming and virion or sE for boosting. (A to C) Priming (P.) with DIII and boosting (B.) with virion; (D to F) priming (P.) with DIII and boosting (B.) with sE. Postimmunization sera were analyzed in virion ELISA (A and D), DIII ELISA (B and E), and NT assay (C and F). The error bars represent the standard errors of the means obtained in at least three independent determinations.

DISCUSSION

When novel vaccines are designed, it would be desirable to use immunogens and immunization schedules with an optimized immunological performance by focusing the response to those sites that are expected to induce the most potent neutralizing antibodies. This may be achieved by the selection or specific design of the immunogen (15, 55, 61) and/or by designing specific prime-boost regimens that direct and strengthen the immune response to the most desirable functional parts of the antigen. In the case of flaviviruses, there is evidence (mostly from the analysis of mouse monoclonal antibodies [16, 42, 59]) that DIII of the envelope protein E can induce potent neutralizing antibodies which are directed to the upper lateral ridge of this domain (12, 47).

In our study, we demonstrate that the functional quality of polyclonal DIII-specific antibody responses was strongly dependent on the structural context of this domain in the immunogen. The differences found were substantial. DIII-specific antibodies induced by the inactivated virus had an almost 4-fold higher neutralizing activity (relative to the DIII ELISA titer; Fig. 3D) than those induced by soluble E. The discrepancy between ELISA reactivity and functional activity was even more pronounced with antibodies induced by the isolated DIII, which had a more than 15-fold lower specific neutralizing activity than those obtained after virion immunization and which failed to completely neutralize the virus. These differences were also reflected by the overall avidities of the antibodies to the virion, which were shown to be the highest after immunization with inactivated virus, intermediate with sE, and the lowest with DIII. These findings are in agreement with those of a dengue virus type 2 immunization study in rhesus macaques (53) using inactivated virion, a DNA vaccine expressing prM and E, and a recombinant DIII-maltose-binding fusion protein. Also in this case, the antibodies induced by the inactivated virus had the highest avidity toward the virus, and this property correlated with a reduction in the total number of days of viremia after challenge (53).

It is likely that the phenomena observed in our study are related to differences in the exposure of antigenic sites in DIII and sE that are buried or less accessible in the virus. Antibodies to such sites may not reach an occupancy at the virion surface that is required for efficient neutralization (46). Inspection of the virion structure shows that DIII forms the highest protrusion in the mature virus but that only a fraction of the total surface of DIII is accessible to the binding of antibodies. Substantial parts of potentially antigenic sites are shielded in the context of the mature virion, specifically, those at the bottom of the domain, directed toward the viral membrane, and those at its lateral sides because of interactions between E dimers in the tightly packed icosahedral envelope of the virus (Fig. 1A). The location of DIII epitopes as a decisive factor in virus neutralization is best exemplified by the WN virus-specific E16 MAb, which has been characterized by X-ray crystallography as well as cryo-EM (26, 41), and MAbs 5A2-7, 13D4-1, and E111 (raised against dengue virus), which have been mapped by yeast display combined with mutagenesis (58). E16 and related MAbs are strongly neutralizing and bind to sites at the upper lateral ridge of DIII (Fig. 1C) that are highly accessible in the context of whole virions (16, 26, 41, 42). In contrast, the epitopes defined by MAbs 5A2-7, 13D4-1, and E111 include the conserved AB loop at the bottom of DIII. This site is exposed in isolated or yeast-displayed DIII but has limited exposure in the context of the virus, because it is located inside the shell of E-protein dimers and faces the viral membrane of the mature virion (58). The corresponding MAbs are therefore poorly or not neutralizing. It has to be kept in mind, however, that the E proteins in flaviviruses can undergo dynamic movements at the virion surface at 37°C (breathing), thus also allowing virus neutralization by antibodies to partially occluded sites (34). Furthermore, the accessibility of antigenic sites may also be influenced by the presence of partially immature virions in the virus neutralization assay (39). Whether and the extent to which such phenomena can also have an influence on the dominance of antibody responses to certain antigenic sites remain to be elucidated.

The relative dominance of DIII-specific antibodies in our study (32% and even 76% of the total antibody response after immunization with virion and sE, respectively) is consistent with other reports on antibody responses to flavivirus infections and immunizations in mice (43). In contrast, several studies with flavivirus-infected humans (11, 31, 43, 59, 62) and horses (52) do not provide evidence for a dominance of DIII antibodies but suggest that the antibody response is skewed to potentially less potent epitopes in DI and DII, including the broadly cross-reactive fusion loop (11, 31, 43, 59). In the monomeric WN virus sE, the FP loop is highly exposed at the tip of DII (Fig. 1B) (25, 40). It was therefore surprising that, like with whole virions, only low titers of broadly cross-reactive antibodies were induced by this antigen. In both instances, this response was minimal relative to the total antivirion response. Given the strong dominance of FP loop-specific antibodies reported in studies using human postinfection sera (11, 31, 43, 59), this is an additional indication for significant differences between mice and humans in the immunodominance of individual epitopes. The factors responsible for directing the antibody response to selected sites at the surface of the same antigen in different species are unknown and will have to be elucidated for generating a more rational basis of engineering antigens for future vaccine design.

Consistent with published data, the immunizing potency of DIII alone was low in our study, and different attempts to increase its immunogenicity have been described in the literature (reviewed in reference 18). This includes the use of fusion proteins (1, 22, 35, 64), combinations with strong adjuvants such as Freund's adjuvant (1, 2, 22, 35), strings of DIII (7, 27), expression by adenovirus vectors (28), or coupling to large carriers such as a bacteriophage (54) to mimic its presentation at the surface of the virus. However, despite its low immunogenicity when it is used alone, DIII proved to be an excellent boosting antigen and strongly enhanced the neutralizing antibody response of mice preimmunized with inactivated virion or sE. In this respect, DIII was superior to the more complex immunogens used for priming. It is justified to assume that DIII (in the prime-boost regimen applied) primarily induced an anamnestic response to those sites of DIII presented at the virion surface that are relevant for virus neutralization. Considering the lack of immunodominance of DIII in humans, novel vaccination strategies could be based on the exploitation of the superior specific immunogenicity and priming capacity of inactivated virions combined with the excellent boosting capacity of DIII. Such an immunization regimen could foster the antibody response to the most relevant neutralizing site in the virus and thus potentially lead to a strengthening of the immunity induced by vaccination. Similar boosting approaches would also be possible in combination with live flavivirus vaccines (60).

As an alternative for focusing the antibody response toward DIII, we also investigated a regimen of priming with recombinant DIII followed by boosting with the more complex antigens, i.e., virion and sE. Surprisingly, this schedule resulted in a significant negative effect on the total antibody response against the booster antigens and, most importantly, in an apparent impairment of the neutralizing response. Factors proposed to be involved in such negative effects include T-cell-mediated suppression, masking preexisting antibody, or enhanced clearance of immune complexes (13, 23, 45). The phenomenon observed may also be influenced by the specific combinations of doses and the immunization schedule used in our study, since opposite effects, i.e., enhanced neutralizing antibody responses after priming with recombinant DIII followed by live virus infection, were described in a dengue virus study (5). Possible effects of using different combinations of antigens, doses, and live virus immunizations on the immunodominance of specific sites may be a topic of future investigations. In any case, our data point out that the use of the isolated DIII for priming can lead (at least under certain circumstances) to an unwanted negative effect, and some caution should therefore be applied in such immunization regimens.

Overall, our study emphasizes that the functionality of antibodies induced by WN virus antigens not only is dependent on their native conformation but also is strongly influenced by their quaternary organization. The flavivirus particle displays the sites most relevant for virus neutralization at its surface, whereas other potentially antigenic faces of the envelope proteins are occluded due to their tight packing in an icosahedral lattice. Such sites, however, are exposed in isolated recombinant proteins, which are therefore potentially prone to induce a higher proportion of antibodies that are irrelevant for neutralization. Despite its low specific immunogenicity when it is used alone, DIII could be superior as a booster vaccine compared to virion or sE, because it optimizes the development of neutralizing antibodies and leads to an increase of the immunodominance of the most critical neutralizing determinants in the virion.