Immune-Focusing Properties of Virus-like Particles Improve Protective IgA Responses

Human noroviruses (HuNoVs) are the leading cause of acute epidemic gastroenteritis worldwide. Globally, noroviruses (NoVs) infect an estimated 700 million patients, resulting in up to 200,000 deaths and are responsible for economic losses of over $60 billion every year (1–3). NoVs are positive-sense, ssRNA viruses of the Caliciviridae family, with at least six genogroups (GI-GVI) and 30 genotypes (4). NoV genotyping is based primarily on the ORF2 sequence encoding the major capsid protein (VP1) (5). NoV strains in genogroups GI, GII, and GIV infect humans, and those in the GI and GII genogroups are responsible for the majority of such human infections (4). GI.1 represents the dominant strains circulating prior to the 1980s; however, since the 1990s, GII.4 strains have been most prevalent, and are associated with ∼70% of all HuNoV infections. In addition, continual antigenic drift generates escape mutants, which overcame herd immunity (6).

No licensed vaccines are currently available for HuNoVs; however, the introduction of recombinant technology in this field established recombinant virus-like particles (VLPs) as a first generation of vaccine candidates (7). HuNoV-VLP vaccines are produced by self-assembly of VP1 protein, which bears morphological and antigenic similarity to live HuNoVs (7–10). The highly repetitive presentation of antigenic epitopes in this vaccine has been speculated to allow the cross-linking of BCRs and complement activation through IgM trapping (11, 12). Moreover, pattern recognition receptor ligands that are often packaged in VLPs exhibit immunostimulatory effects (13), including enhanced germinal center responses, durable IgG responses, and rapid IgG responses through the bypassing of T cell dependency (11, 12, 14). Indeed, previous clinical evidence has demonstrated that i.m. administration of NoV-VLP vaccines elicits anti-VP1 IgG and IgA Abs, which are able to inhibit virus binding to host histo-blood group Ags (HBGA), the surrogate for protection against HuNoV gastroenteritis (15–17). However, it is still not clear how VLP structure regulates the Ab responses and what its impacts on mucosal IgA responses are, despite the significant correlation between virus-specific IgA titers and a reduction in the risk of HuNoV infection (18).

In this study, two approaches were introduced for dissecting human memory responses against NoVs: identification of NoV-specific human memory B cells via flow cytometry in PBMCs and reconstitution of human memory responses in a human PBMC–transplanted mouse model. We demonstrated that the highly repetitive epitopes of NoV-VLPs crucially regulate NoV-specific IgA responses in both quantitative as well as qualitative manners, whereas IgG responses are impacted in a less pronounced manner. Thus, our results illustrate the immune-focusing properties of VLPs, which could be relevant to mucosal vaccine efficacy.

NoV-VLPs were prepared as previously described (19). In brief, ORF2 in the genome end regions of Saga (GII.4) (AB447456, https://www.ncbi.nlm.nih.gov/nuccore/AB447456), 124 (GI.1) (accession no. AB031013, https://www.ncbi.nlm.nih.gov/nuccore/AB031013), and mouse NoV (MNV)-S7 (accession no. AB435514, https://www.ncbi.nlm.nih.gov/nuccore/AB435514) strains were cloned and used to produce a recombinant baculovirus in a BAC-to-BAC system (Thermo Fisher Scientific), according to the manufacturer’s protocol. Recombinant NoV-VP1 capsid proteins were expressed in an insect cell line (High Five cells; Thermo Fisher Scientific) prior to VLP concentration by ultracentrifugation at 32,000 rpm in an SW32 rotor (Beckman Coulter, Palo Alto, CA). VLPs of native virion size (38-nm diameter) were purified by CsCl ultracentrifugation. Similarly, histidine-tagged recombinant P- and S-domains of MNV-S7-VP1 protein were expressed in Sf21 cells (Thermo Fisher Scientific) and purified using a TALON column (Clontech). Disrupted VLPs were prepared by treatment with 0.1 M Tris (pH 9.4) for 3 h at room temperature. Disrupted VLPs were diluted with water and mixed with mouse serum to avoid the reassembly. For the generation of NoV-VLP fluorescent probes, NoV-VLPs were conjugated to PE (Dojindo), allophycocyanin (Dojindo), or Alexa Fluor 594 (Thermo Fisher Scientific).

Anti-CD2 (RPA-2.10)–biotin, anti-CD4 (161A1)–biotin, anti-CD14 (63D3)–biotin, anti- CD19 (HIB19)–BV785, anti-CD27 (O323)–Alexa Fluor 700, anti-CD3 (145-2C11)–biotin, anti-B220/CD45R (RA3-6B2)–biotin/Alexa Fluor 700, anti-F4/80 (BM8)–biotin, anti-CD11b (M1/70)–biotin, anti-CD11c (N418)–biotin, anti-CD117 (2B8)–biotin, anti-TER-119 (TER-119)–biotin, and anti-CD38 (90)–Pacific Blue were purchased from BioLegend. Anti-CD10 (ebioCB-CALLA)–biotin, anti-IgM (II/41)–biotin, anti-IgD (11-26c)–biotin, anti-CD93 (AA4.1)–biotin, anti-Thy1.2/CD90.2 (53-2.1)–biotin, anti-Gr-1 (RB6-8C5)–biotin, anti-CD8α (53-6.7)–biotin, anti-CD5 (53-7.3)–biotin, and anti-CD4 (GK1.5)–biotin were purchased from eBioscience. Anti-IgD (IA6-2)–biotin, anti-IgG (G18-145)–FITC, anti-CD43 (S7)–biotin, anti-CD138 (281-2)–biotin, anti-CD19 (1D3)–biotin, and anti-IgG1 (A85-1)–FITC were purchased from BD Biosciences. Polyclonal rabbit anti-human IgA-FITC was purchased from DAKO. Anti-human IgA1 (RM124; Abcam) and anti-human IgA2 (IS11-21E11; Miltenyi Biotec) were labeled with Pacific Blue (Thermo Fisher Scientific) and PerCP-Cy5.5 (Innova Bioscience), respectively. Goat anti-human IgG-alkaline phosphatase (AP), goat anti-human IgG-HRP, goat anti-human IgA-HRP, goat anti-human IgA1-HRP, goat anti-human IgA2-HRP/AP, goat anti-mouse IgG-HRP, and goat anti-mouse IgA-HRP were purchased from Southern Biotech. Streptavidin (SA)-HRP and Live/Dead Aqua were purchased from Thermo Fisher Scientific. Anti-FcγRII/III (2.4G2) mAbs were purified in our laboratory. Anti-GII.4 mAbs G2F3L, G6C6k, and H2C2k were established by V_H/V_L cloning into human IgG1 H chain and κ/λ L chain expression vectors from single cell–sorted GII.4-VLP–binding human memory B cells, as previously described (20). Anti-GI.1– and GII.4–polyclonal IgG were purified from sera of mice vaccinated with GI.1- or GII.4-VLP using Protein G (Thermo Fisher Scientific) and were biotinylated with EZ-Link Sulfo-NHS-Biotin (Thermo Fisher Scientific).

Heparinized peripheral blood was obtained from healthy donors after obtaining written informed consent, and PBMCs were isolated by density centrifugation using Ficoll-Paque PLUS (GE Healthcare). PBMCs were incubated with biotinylated mAbs against CD2, CD4, CD10, CD14, and IgD for human memory B cells. This was followed by staining with GI.1-VLP-allophycocyanin, GI.1-VLP–Alexa Fluor 594, GII.4-VLP-PE, SA-BV510, Live/Dead Aqua, anti-CD19-BV785, anti-CD27–Alexa Fluor 700, anti-human IgA1–Pacific Blue, anti-human IgA2-PerCP-Cy5.5, and anti-human IgA- or IgG-FITC. Murine splenocytes were pretreated with anti-FcγRII/III mAb and then incubated with biotinylated mAbs against IgM, IgD, CD43, CD138, CD5, CD93, CD90, CD3, F4/80, CD11b, CD117, CD11c, Gr-1, and TER-119 for mouse memory B cells. This was followed by staining with GI.1-VLP-allophycocyanin, GI.1-VLP–Alexa Fluor 594, GII.4-VLP-PE, SA-BV510, Live/Dead Aqua, anti-B220–Alexa Fluor 700, anti-CD38–Pacific Blue, and anti-IgG1–FITC. Stained cells were analyzed or sorted using a FACSAria instrument (BD Bioscience). The studies using human samples were approved by the Institutional Ethics Committee of Human Experimentation and performed in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects of Japan.

Human memory B cells were cultured and expanded as previously described (21). In brief, GI.1- or GII.4-VLP–binding human memory B cells were directly sorted into 96-well plates as single cell and cultured with MS40L–low feeder cells in RPMI 1640 medium (Fujifilm) supplemented with 10% FBS (HyClone; Thermo Fisher Scientific), penicillin-streptomycin, 10 mM HEPES, 1 mM sodium pyruvate, MEM nonessential amino acid, 2-ME (Thermo Fisher scientific), recombinant human IL-2 (50 ng/ml), IL-4 (10 ng/ml), IL-21 (10 ng/ml), and BAFF (10 ng/ml; all obtained from PeproTech). Cultures were maintained under 5% CO₂ at 37°C for 4 wk, and the culture supernatants were harvested for screening the production of VLP-binding human IgG or IgA Abs by ELISA.

C57BL/6J mice were purchased from Japan SLC, and C57BL/6-scid (B6-SCID) mice were purchased from The Jackson Laboratory. NOD/SCID/Jak3^−/− (NOJ) mice were kindly provided by Dr. S. Okada (Kumamoto University). Mice were maintained under specific pathogen-free conditions at our institute. For the development of a humanized mouse model, 2 × 10⁷ human PBMCs from each donor were i.v. transferred into NOJ mice. The following day, recipient mice were i.v. or i.m. vaccinated with GI.1- or GII.4-NoV-VLPs (20 μg/mouse). Sera and splenocytes were recovered 10 or 14 d postvaccination for further analysis. All murine work was performed in accordance with the guidelines of the Institutional Animal Care and Use Committee of the National Institute of Infectious Diseases, Japan.

C57BL/6J mice were orally inoculated with MNV-S7 at 1 × 10⁶ TCID₅₀/mouse. Two months after the infection, primed B cells and CD4⁺ T cells were purified from pooled spleens and lymph nodes of infected mice using a MACS system. The cells were incubated with biotinylated mAbs (CD3, CD90, CD4, CD8α, F4/80, Gr-1, CD11b, CD117, and CD138 for B cells; CD19, B220, IgM, IgD, F4/80, Gr-1, CD11b, and CD8α for CD4⁺ T cells), followed by the incubation with SA-microbeads (Miltenyi Biotec). Purified B cells (4 × 10⁷ cells per mouse) and CD4⁺ T cells (2 × 10⁷ cells per mouse) were i.v. transferred into B6-SCID mice. One day posttransfer, mice were i.v. vaccinated with MNV1-S7 VLP at 20 μg/mouse. Sera and feces were collected 14 d postvaccination.

For detection of anti-GI.1 and -GII.4 plasma cells by ELISPOT assay, nitrocellulose membranes were coated with 10 μg/ml VLP, and the cells were incubated on the membranes for 3 h at 37°C. After the cells were washed off, membranes were treated with anti-human IgA-HRP/anti-human IgG-AP and anti-human IgA1-HRP/anti-human IgA2-AP. AP and HRP activities were evaluated as previously described (22). Anti-GI.1, -GII.4, and -MNV-S7 Ab titers were detected by ELISA, using VLPs or histidine-tagged recombinant P- or S-domains as coating Ag, and anti-human IgG-HRP, anti-human IgA-HRP, anti-human IgA1-HRP, anti-human IgA2-HRP, anti-mouse IgG-HRP, and anti-mouse IgA-HRP as secondary Abs, as previously described (22). IgG and IgA forms of G3B9λ mAbs were used as standard. The avidity index was calculated by dividing titers of anti-VLP IgG or IgA resistant to 7 M urea treatment (15 min) using titers of anti-VLP IgG or IgA Ab.

The ability of vaccinated sera to block binding of GI.1- and GII.4-VLPs to HBGA was assayed as previously described (23). In brief, ELISA plates were coated with 10 μg/ml pig gastric mucin (PGM) type III (Sigma Chemicals). GI.1- (0.1 μg/ml) or GII.4- (0.2 μg/ml) VLPs were incubated with serially diluted sera or purified Abs for 6 h at 37°C. These mixtures were then added to PGM-coated plates and incubated for 30 min. Binding of VLPs to coated plates was detected using biotinylated mouse polyclonal anti-GI.1 or -GII.4 Abs followed by staining with SA-HRP.

Virus-neutralizing Ab titers were determined via a microneutralization assay using RAW267.4 cells. In brief, serially diluted sera, supernatants of fecal suspensions, or purified IgA or IgG were incubated with MNV-S7 (50 TCID₅₀) for 4 h at 37°C and then added to RAW267.4 cells. After 3 d of incubation, the end point Ab titers that protected the cells from infection were determined.

Sera were collected from VLP-boosted B6-SCID mice after reconstitution with primed B and T cells from MNV-infected C57BL/6J mice. IgG and IgA Abs were purified by Protein G (GE Healthcare) and CNBr-activated Sepharose 4B (GE Healthcare) coupled with goat anti-mouse IgA (Southern Biotech). Concentrations of VLP-binding Abs in purified IgG and IgA fractions were determined by ELISA using standard G3B9λ IgG and IgA mAbs. C57BL/6J mice were fasted for 2 h followed by oral administration of 100 μl of 7.5% NaHCO₃ to neutralize stomach acid, followed by oral infection with MNV-S7 at 1 × 10⁵ TCID₅₀/mouse. Purified IgG was i.v. injected 2 h before infection. For IgA administration, purified IgA was premixed with MNV-S7, incubated for 4 h at 37°C, and orally administered. Feces were collected daily, and supernatants of fecal suspensions were added to RAW264.7 cells for determining virus titers as TCID₅₀.

Mouse splenocytes (1 × 10⁵ cells) were cultured with either 0.1–1 μg of GII.4-VLP or inactivated influenza virion for 24 h. Then, the culture supernatants were subjected to ELISA for detection of mouse IFN-α.

The endotoxin levels were examined using Endospecy ES-50M kit (Seikagakukougyou, Japan) as per the manufacturer’s protocol. The amount of endotoxin included in 1 mg/ml GII.4 VLP aliquot was determined using endotoxin standard (JRPS Lab, Japan), and 1 ng/ml of LPS derived from Escherichia coli (O111:B4; Sigma) was used as positive control.

Statistical analysis was performed using the PRISM v7.03 software (GraphPad, La Jolla, CA). The nonparametric two-tailed Mann–Whitney U test and Wilcoxon matched-pairs signed-rank test were used to determine statistical significance. Asterisks represent significant difference (*p < 0.05, **p < 0.01, ***p < 0.001).

We sought to identify and quantify NoV-specific memory B cells in the peripheral blood of healthy adults by flow cytometry. We generated GI.1 (124 strain)- and GII.4 (Saga strain)-VLP–based fluorescent probes as previously reported for influenza virus (22, 24, 25). To confirm the specificity of NoV-VLP probes to NoV-specific memory B cells, we first analyzed splenocytes from mice immunized with either GI.1- or GII.4-VLP. Both GI.1- and GII.4-VLP probes specifically detected GI.1- or GII.4-binding B cells among IgG1⁺CD38⁺ memory B cells in accordance with the priming VLPs (Supplemental Fig. 1A, 1B). We then employed a similar method for detection of NoV-specific human memory B cells in PBMCs. To minimize contamination by fluorochrome binders, we defined GI.1-VP1–binding B cells as the fraction that simultaneously binds to GI.1-VP1 labeled with two different fluorochromes, but not to GII.4-VP1. Similarly, GII.4-VP1–binding B cells were defined as the fraction that binds to GII.4-VP1, but not to GI.1-VP1. Both populations were clearly visualized as VLP-binding B cells among class-switched memory B cells, and all healthy donors exhibited detectable levels of IgG⁺ and IgA⁺ memory B cells against both genotypes (GI.1⁺IgG⁺: 63 ± 37 cells per 10⁷ PBMCs, GI.1⁺IgA⁺: 64 ± 83 cells per 10⁷ PBMCs, GII.4⁺IgG⁺: 383 ± 231 cells per 10⁷ PBMCs, GII.4⁺IgA⁺: 348 ± 319 cells per 10⁷ PBMCs, n = 10), supporting that these healthy donors have experienced a previous NoV infection and continue to maintain long-term humoral memory responses against this pathogen (Fig. 1A, 1B). We detected comparable abundance of IgG⁺ and IgA⁺ memory B cells specific for both genotypes, a similar finding being observed in prevaccinated donors from a previous clinical study (26). The currently predominant circulation of GII.4 strains in the human population likely accounts for the abundance of GII.4-specific memory B cells of both IgG and IgA isotypes (Fig. 1B) (27–29).

To validate Ag specificity as identified by the flow-cytometric strategy, we employed a single-cell culture approach. Sorted single cells were individually cultured on feeder cells supplemented with a human cytokine mixture for 4 wk, after which VLP-specific Ab production in the culture supernatants was evaluated (21). Approximately 60–70% of IgG and IgA were detectable among single-sorted GI.1- or GII.4-binding memory B cells (Fig. 1C). In addition, we found that the majority of sorted GI.1-VLP–binding memory B cells produced GI.1-binding Abs (85 and 100% among IgG⁺ or IgA⁺ wells, respectively). Similarly, 68 and 85% of sorted GII.4-VLP–binding memory B cells produced GII.4-binding IgG⁺ and IgA⁺ Abs, respectively. We speculate that the VLP-nonbinding fraction includes low-affinity B cells, because low-affinity Abs in a soluble form become undetectable by ELISA, although the same Abs in membrane-bound form are detectable by flow cytometry (30). Thus, the single-cell culture approach confirmed the specificity of our flow-cytometric detection of NoV-specific human memory B cells.

Human IgA comprises two subclasses: IgA1 is abundantly present in serum, and IgA2 localizes to mucosal sites (31). We next assessed the frequency of IgA1⁺ versus IgA2⁺ B cells among NoV-specific IgA⁺ memory B cells. In accordance with the results for serum IgA, ∼70% of CD27⁺IgA⁺ memory B cells were IgA1⁺. However, we found that the IgA1 dominance was more pronounced among NoV-specific IgA⁺ memory B cells, such that >90% of cells were IgA1⁺, regardless of NoV genotypes (Fig. 1D, 1E).

Recall responses of pre-existing human memory B cells to HuNoV-VLP were examined. For this purpose, we used a humanized mouse model wherein human PBMCs from eight donors (shown in Fig. 1) were adoptively transferred into three groups of NOJ mice (hu-PBL-SCID mice). The PBMCs used for adoptive transfer were confirmed to include comparable numbers of IgG and IgA memory B cells binding to GI.1- and GII.4-VLPs (Fig. 1B). Two groups of hu-PBL-SCID mice then received either GI.1- or GII.4- VLP to assess the boosting ability for human recall responses (Fig. 2A). The remaining one group of mice received PBS only to assess the background of pre-existing plasma cells in the transplanted PBMCs. i.v. boosting with GI.1- and GII.4-VLPs abundantly elicited VLP-binding plasma cells of IgA and IgG isotypes at day 10 postboosting with specificities relevant to the type of boosted VLPs. These results indicated that human memory IgG and IgA responses against NoVs were successfully reconstituted in our humanized mouse model. The recall responses were not specific to i.v. boosting but were comparably elicited by i.m. boosting (Supplemental Fig. 2B, 2C). Moreover, we did not see a significant increase or decline in the Ab responses even at a later time point (day 14 after boosting) (Supplemental Fig. 2D, 2E). Therefore, the following analysis was performed at day 10 after i.v. boosting.

The previous clinical trial for NoV-VLP vaccines demonstrated that i.m. administration of NoV-VLP vaccines with alum and monophosphoryl lipid A adjuvant preferentially induces IgA⁺ plasma cells in peripheral blood (15). Of note, we observed similar results wherein the numbers of IgA⁺ plasma cells were significantly higher than those of IgG⁺ plasma cells following vaccination without adjuvant and through an i.v. route (IgA/IgG ratios: 6-fold for GI.1 and 7-fold for GII.4) (Fig. 2B). Thus, our data advanced the previous findings from clinical trial and strongly suggested that IgA-focused recall responses are irrelevant to the adjuvant and vaccination route. Rather, they appear to depend on VLP Ag itself. Moreover, given comparable abundance of IgG⁺ and IgA⁺ memory B cells in transplanted PBMCs (Fig. 1B), VLP vaccines induced a more robust reactivation of IgA⁺ memory B cells and/or sequential IgA switching from IgM/G⁺ memory B cells for supplying IgA⁺ plasma cells.

The distribution of IgA1 and IgA2 subclasses among recalled IgA⁺ plasma cells was examined in humanized mice. Similar to the results for IgA⁺ memory B cells (Fig. 1E), majority of recalled IgA⁺ plasma cells were of the IgA1 subclass (Fig. 2C, 2D). The IgA1 dominance among VLP-binding IgA⁺ memory B cells account for similar bias in plasma cells; however, the rapid homing of IgA2⁺ plasma cells to mucosal sites might additionally contribute to IgA1 dominance. Thus, HuNoV-VLP vaccines were able to elicit IgA1⁺ recall responses from pre-existing memory B cells in the serum of humanized mice.

We quantified the levels of NoV-VLP–specific IgG and IgA Abs in the sera of VLP-boosted hu-PBL-SCID mice using low-affinity chimeric IgG and IgA mAbs sharing the same Ag binding sites. The usage of low-affinity mAbs as a reference generally overestimates Ab titers and, indeed, we observed a serum IgG concentration of >100 μg/ml (Fig. 3A). However, in contrast to IgA-biased plasma cells as seen in Fig. 2, serum IgA levels were slightly lower than IgG levels in both GI.1- or GII.4-VLP–boosted mice, possibly reflecting differential t_1/2 of Ig subclasses in sera (32). Indeed, human serum IgA, i.v. transferred into the mice, exhibited a shorter t_1/2 compared with that of human serum IgG (Supplemental Fig. 3).

PGM is a surrogate of HBGAs (33). We assayed HBGA-binding blockade by VLP-boosted sera using PGM. Each boosted serum was preincubated with VLPs and then subjected to an assay assessing the inhibitory activity of PGM binding. Compared with control serum, VLP-boosted serum significantly blocked VLP-PGM binding, regardless of the VLP genotype used for boosting (Fig. 3B, 3C). These results indicated that NoV-VLPs recall Abs that block VLP binding to HBGA, the well-known surrogate for protection against HuNoV gastroenteritis.

The VLPs often include TLR ligands and present highly repetitive epitopes on the particle structures, which significantly contribute to shaping Ab responses in several viruses and vaccine studies (12). Highly repetitive epitopes presented on viral particles efficiently activate complement and cross-link BCRs, both of which enhance mouse IgG responses (11, 14). We then assessed the impact of the highly organized VLP structure on the quality and magnitude of NoV-specific human recall responses. Treatment of GII.4-VLP with pH 9.4 disrupted its highly organized structure into dissociated VP1 proteins (Fig. 4A); thus, we addressed whether the antigenic structure remained intact in disrupted VLP. Comparable levels of NoV-binding polyclonal IgG in human serum bound to intact and disrupted VLP (Supplemental Fig. 4A). Similar findings were reproducible using three different anti-GII.4 mAbs generated from human memory B cells (Supplemental Fig. 4B). Thus, the disrupted VLP retains its antigenic structure.

Intact and disrupted VLPs were then used as boosting Ags to elucidate the impact of VLP structures on IgA-focused recall Ab responses. As expected, human IgG plasma responses were significantly attenuated after boosting with disrupted VLP (Fig. 4B); however, a more prominent reduction occurred in human IgA responses. In agreement with the attenuation of plasma cell responses, serum anti-GII.4-VLP Ab levels were also lowered in recipients boosted with disrupted VLP, with a more pronounced effect being noted for IgA (Fig. 4C). Such attenuated recall responses from disrupted VLP were also reflected in a reduction of HBGA-blocking Ab titers (Fig. 4D). These data indicated that IgA-focused recall responses are more dependent on VLP structure.

VLPs obtained via expression in a baculovirus system sometimes package baculoviral DNA, which potentially exhibit adjuvant effects through TLR9 engagement (34). We, therefore, assessed whether HuNoV-VLP stimulates type I IFN response, a representative outcome of TLR9 engagement by viral DNA. Even under a condition where inactivated influenza virus stimulated a type I IFN response in mouse splenocytes, GII.4-VLP did not trigger detectable levels of type I IFN response (Supplemental Fig. 4C). In addition, the contaminated endotoxin was below the detection limit (<0.0195 endotoxin unit/ml) (Supplemental Fig. 4D). These results reduced the likelihood of viral DNA packaging and contamination of endotoxin within HuNoV-VLP.

Although HuNoVs do not replicate in the mouse intestine, several strains of MNVs belonging to genogroup V do replicate in this location (35, 36), allowing us to estimate the functional outcomes of IgA-focused recall responses. We first examined whether the IgA enhancement is similarly recapitulated in the MNV system (Fig. 5A). Before analysis, we prescreened MNV-binding IgG and IgA titers in serum and selected the mice that were confirmed to be negative for these tests. Murine recall responses against MNV-VLP vaccines were evaluated in the transfer system through which mouse B cells and CD4⁺ T cells from infected C57BL/6J mice were transferred into the B6-SCID mice (Fig. 5A). Consistent to results observed for the human recall responses, anti-NoV IgA titers were significantly higher in both sera (systemic site) and feces (mucosal site) of mice boosted with intact MNV-VLP. In contrast, serum and fecal IgG titers were not significantly different between intact and disrupted VLP-boosted groups (Fig. 5B). Thus, IgA-focused recall responses are reproducible in MNV-driven, murine B cell responses.

To assess the functional impact of IgA enhancement, we first compared the binding avidity and virus-neutralizing activity of IgG/IgA to MNV. Although we found no difference in the binding avidity between IgG and IgA (Fig. 5C), the virus-neutralizing activity in both serum and fecal Abs was attenuated after boosting by disrupted VLPs (Fig. 5D). More prominent reduction in fecal Abs implied the differential neutralizing activity between IgG and IgA. Indeed, purified serum IgA neutralized MNV at 4-fold lower concentrations than IgG (Fig. 5E), demonstrating the superior virus-neutralizing activity.

VP1 proteins are composed of membrane-distal P-domain and membrane-proximal S-domain. Because the P-domain includes the receptor-binding sites, the Abs targeting P-domain exhibit better neutralizing activity. To assess the Ab epitope distribution among P- and S-domains, ELISA was performed using truncated forms of VP1 P-domain- and S-domain–only proteins as detecting Ags. Intriguingly, in contrast to IgG Abs, which equivalently bound to P- and S-domain, a higher percentage of IgA Abs directed toward P-domain, including neutralizing epitopes (Fig. 5F). Therefore, the epitope preference for receptor-binding domain accounts for, at least partly, the superior neutralization activity of IgA Abs. Finally, protective functions were evaluated in vivo by transferring purified, virus-binding IgA or IgG into recipient C57BL/6J mice, which were subsequently challenged with MNVs. Booster vaccination by MNV-S7-VLP elicited IgG-dominant responses in serum (IgG/IgA ratio = 86.8) and IgA-dominant responses in feces (IgA/IgG ratio = 32.4) (Fig. 5B). Therefore, purified, virus-binding IgA was transferred into recipient mice from oral route to recapitulate the mucosal distribution of IgA in the donor mice. The purified, virus-binding IgG was transferred from i.v. route to reproduce its dominance in the serum. Importantly, a more prominent reduction in viral load was observed in mice receiving IgA compared with those receiving an equivalent quantity of IgG (Fig. 5G). The mucosal localization of orally transferred IgA was required for in vivo protective activity, because i.v. transfer of the same IgA failed to confer protection later than day 2 postinfection (Fig. 5G). The defective protection by serum IgA in the late time points probably reflects their rapid clearance from circulation, as shown in Supplemental Fig. 3. Taken together, these data suggest that VLP-driven mucosal IgA responses are superior to systemic IgG responses in terms of virus-neutralizing activity in vitro and protective function in vivo.

Given the superior immunogenic properties of VLPs compared with those of soluble protein, the VLP platform is now being employed for many vaccines (37). At least two structural characteristics of VLPs are associated with their immunogenic properties: highly repetitive epitopes on the particle surface and packaging of PRR ligands. Such structural characteristics have been shown to enhance protective IgG responses via activation of both innate immune cells and B cells. However, their potential impact on mucosal IgA responses remains unknown. In this study, we demonstrated that the IgA-focusing properties of VLPs are particle-structure dependent. Particle-dependent IgA enhancement was conserved in both human and murine lymphocytes; this had relevance to clinical results in which a similar bias in the IgA recall response was noted upon administration of the HuNoV-VLP vaccine (15). Data presented in this study further revealed the correlation of VLP-dependent IgA enhancement to protection against mucosally infecting NoV. Thus, this study significantly advanced our understanding of virus IgA recall responses and provided a rational basis for using VLP platform as NoV vaccine.

There are no licensed NoV vaccines; therefore, it is highly likely that NoV-binding memory B cells in humans are elicited by natural infection in the past. The initial B cell priming by mucosal infection probably accounts for the abundance of IgA⁺ memory B cells as well as IgG⁺ memory B cells in peripheral blood (Fig. 1). It is also conceivable that the mucosal B cell priming could at least partially contribute to the IgA-biased recall responses by predisposing memory B cells to the enhanced IgA responses upon restimulation. However, comparative studies between intact and disrupted VLPs for the boosting ability of mouse and human memory B cells clearly demonstrated that the restimulation by intact VLP is required for maximizing the IgA-biased recall responses (Figs. 4, 5).

VLPs and nanoparticles often package ligand molecules for PRRs (e.g., TLRs) that are expressed in dendritic cells and B cells (22, 38, 39). As mentioned above, baculovirally expressed VLPs sometimes contain viral DNA, which has the potential to engage DNA sensors (i.e., TLR9) in B cells and dendritic cells (34). Thus, we initially speculated that TLR signals might play IgA-enhancing roles during restimulation of memory B cells in our experimental systems. However, stimulation of mouse splenocytes with HuNoV-VLPs did not elicit detectable levels of IFN-α under these conditions, where high level of IFN-α production occurs upon stimulation with inactivated influenza virus. In addition, the endotoxin levels were below the detection limit. Therefore, we speculate that the contribution of TLR-mediated stimulation is less prominent for stimulating IgA responses.

Alternatively, BCR cross-linking and complement activation by highly repetitive epitopes, present on VLPs, might contribute to IgA enhancement. VLP-driven IgA responses exhibited superior virus-neutralizing activity with preferential targeting to P-domain epitopes (Fig. 5E, 5F). Because the P-domain is located in the membrane-distal region of VP1 with possession of receptor-binding sites, we speculate that the repetitive P-domain presentation on the VLP outer surface could be optimally recognized by P-domain–directed B cells. The effective cross-linking of BCR on memory B cells may itself reactivate P-binding memory B cells more strongly relative to S-binding memory B cells, or it might lead to the enhanced presentation of VP1 Ags to memory T cells and receipt of helper signals, both of which are expected to induce greater levels of memory B cell activation. Therefore, we prefer the idea that VLP directly acts on P-binding memory B cells; they eventually receive robust BCR signals or increased T cell help as a result of strong BCR cross-linking. IgA is an isotype that is located downstream of IgG, and thus, it may be possible that the strong stimulation triggered by VLP accelerates the production of IgA⁺ plasma cells not only through IgA⁺ memory B cells, but also through upstream IgM or IgG⁺ memory B cells following de novo IgM/IgG to IgA switching, both of which together contribute to the IgA-focusing effects of VLP.

P-directed IgA responses were superior to IgG responses in terms of virus-neutralizing activity. In addition, the IgA in mucosa obtained remarkable protective function toward orally administrated virus in vivo. These data suggest that IgAs are more protective than IgGs and are more correlated with protection, although the mucosal localization is prerequisite for maximizing the protective function in vivo. In addition, the IgA-enhancing activity of the VLP structure is advantageous for increasing vaccine efficacy. Indeed, i.m. VLP vaccine in clinical trials significantly reduced the incidence of acute gastroenteritis following experimental challenge in volunteers (40). Although its correlation to protection still remains to be determined in this study, it is intriguing to assess the amounts of mucosal IgA and serum IgG responses targeting the receptor-binding domain of the virus particles in relevance to the levels of protection conferred by the vaccine. Furthermore, the improvement of IgA-inducing effects, possibly via the addition of adjuvants or the route of vaccination, might be required for increasing vaccine efficacy under clinical conditions.

We thank Akira Dosaka, Takahito Fukushima, Kouji Ichiki, Kouki Tsuchiya, Kei Takada, Eriko Izumiyama, and Wakako Yachi for technical assistance.

M.M. is an employee of Denka Co., Ltd. The author acknowledges a potential conflict of interest and attests that the work contained in this report is free of any bias that might be associated with the commercial goals of the company. The other authors have no financial conflicts of interest.