Epitope(s) involving amino acids of the fusion loop of Japanese encephalitis virus envelope protein is(are) important to elicit protective immunity

ABSTRACT Dengue vaccine candidates have been shown to improve vaccine safety and efficacy by altering the residues or accessibility of the fusion loop on the virus envelope protein domain II (DIIFL) in an ex vivo animal study. The current study aimed to comprehensively investigate the impact of DIIFL mutations on the antigenicity, immunogenicity, and protective efficacy of Japanese encephalitis virus (JEV) virus-like particles (VLPs) in mice. We found the DIIFL G106K/L107D (KD) and W101G/G106K/L107D (GKD) mutations altered the binding activity of JEV VLP to cross-reactive monoclonal antibodies but had no effect on their ability to elicit total IgG antibodies in mice. However, JEV VLPs with KD or GKD mutations induced significantly less neutralizing antibodies against JEV. Only 46% and 31% of the KD and GKD VLPs-immunized mice survived compared to 100% of the wild-type (WT) VLP-immunized mice after a lethal JEV challenge. In passive protection experiments, naïve mice that received sera from WT VLP-immunized mice exhibited a significantly higher survival rate of 46.7% compared to those receiving sera from KD VLP- and GKD VLP-immunized mice (6.7% and 0%, respectively). This study demonstrated that JEV DIIFL is crucial for eliciting potently neutralizing antibodies and protective immunity against JEV. IMPORTANCE Introduction of mutations into the fusion loop is one potential strategy for generating safe dengue and Zika vaccines by reducing the risk of severe dengue following subsequent infections, and for constructing live-attenuated vaccine candidates against newly emerging Japanese encephalitis virus (JEV) or Japanese encephalitis (JE) serocomplex virus. The monoclonal antibody studies indicated the fusion loop of JE serocomplex viruses primarily comprised non-neutralizing epitopes. However, the present study demonstrates that the JEV fusion loop plays a critical role in eliciting protective immunity in mice. Modifications to the fusion loop of JE serocomplex viruses might negatively affect vaccine efficacy compared to dengue and zika serocomplex viruses. Further studies are required to assess the impact of mutant fusion loop encoded by commonly used JEV vaccine strains on vaccine efficacy or safety after subsequent dengue virus infection.

IMPORTANCE Introduction of mutations into the fusion loop is one potential strategy for generating safe dengue and Zika vaccines by reducing the risk of severe dengue following subsequent infections, and for constructing live-attenuated vaccine candidates against newly emerging Japanese encephalitis virus (JEV) or Japanese encephalitis (JE) serocomplex virus.The monoclonal antibody studies indicated the fusion loop of JE serocomplex viruses primarily comprised non-neutralizing epitopes.However, the present study demonstrates that the JEV fusion loop plays a critical role in eliciting protective immunity in mice.Modifications to the fusion loop of JE serocomplex viruses might negatively affect vaccine efficacy compared to dengue and zika serocomplex viruses.Further studies are required to assess the impact of mutant fusion loop encoded by commonly used JEV vaccine strains on vaccine efficacy or safety after subsequent dengue virus infection.
Flavivirus possesses a single-stranded, positive-sense RNA, which undergoes translation to yield three structural proteins [Capsid (C), precursor membrane protein (prM), and envelope (E) proteins] and seven nonstructural proteins (nonstructural protein 1, 2A, 2B, 3, 4A, 4B, and 5, abbreviated as NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5, respectively).Notably, CD4 + and CD8 + T-cell epitopes have predominantly been identified on structural and nonstructural proteins, respectively (9,10,12).The majority of antibody responses recognize the virus E and NS1 proteins (9,16).E protein is the primary target to elicit neutralizing antibodies correlated with vaccine potency (17)(18)(19).The E protein comprises three domains (DI, DII, and DIII) and plays a critical role in the binding of virus to cellular receptors and viral entry.Most potently neutralizing monoclonal antibodies (mAbs) recognized DIII and E dimer epitopes (EDE), whereas poor or non-neutralizing mAbs mainly targeted fusion loop (residues 98-110) on DII (DII FL ), which were involved in enhancing dengue disease severity (11,16).However, flavivirus infection predominantly elicited anti-DII FL antibodies rather than anti-DIII antibodies (16,20).Vaccine candidates using specific antigens or structurally modified antigens can be used to modulate host immunity against virus infection such as human immu nodeficiency virus, influenza viruses, and severe acute respiratory syndrome coronavi rus 2 (21)(22)(23).It has been suggested that flavivirus vaccine candidates should aim to refocus antibody response from targeting DII FL to DIII and EDE epitopes to enhance the neutralizing antibody response and reduce the antibody-dependent enhancement (ADE)-related antibody response, which is involved in developing severe dengue (11,16).
A DIII-based flavivirus vaccine candidate showed lower immunogenicity and induced less neutralizing antibodies than the soluble E monomer (24).Thus, the vaccine candidate should preserve the neutralizing epitopes on the other domains and/or inter-domains (24).E-homodimer antigens re-directed the antibody response from DII FL to non-DII FL and EDE epitopes and elicited higher protective potency as compared to E monomers (25,26).This modulation was related to different antigenicity in the quaternary epitopes rather than DII FL epitopes, the result suggested that reducing DII FL immunogenicity might indirectly improve the production of neutralizing antibodies (25,26).Flavivirus virus-like particles (VLPs) formed a similar structure to virions and were thought to induce more quaternary-dependent antibodies (27,28).Previous studies introduced mutations into DII FL to reduce its immunodominance and used VLPs or VLP-expressing DNA plasmids as immunogens (29)(30)(31)(32).Mutations in DENV and ZIKV DII FL were found to inhibit the occurrence of ADE in immunized mice.However, the impact of these mutations on the DII FL immunogenicity and its ability to elicit neutralizing antibodies remains unclear and inconsistent (31,32).This inconsistency may be due to different DII FL mutations and variations in the immune landscape of various serocomplex groups.It has been suggested that DII FL of JE serocomplex viruses is not necessary for inducing neutralizing antibodies, as the most potent neutralizing antibodies recognize DIII rather than DII FL (33)(34)(35)(36).Some studies have indicated that encoding DII FL mutations could influence the antigenicity of JE serocomplex virus-derived VLPs or the immuno genicity of JEV VLP-expressing plasmids (37)(38)(39).However, there is a need to conduct comprehensive research on the impact of DII FL mutations on the antigenicity, immuno genicity, and protective potency of JE serocomplex virus-derived immunogens.
This study aims to investigate the role of DII FL in eliciting B-cell-direct protective immunity against JE serocomplex viruses.The VLP-expressing DNA plasmid has been shown to elicit cytotoxic T-cell immunity (40); thus, we decided to utilize JEV VLP to examine the influence of DII FL mutations on the antigenicity, immunogenicity, and its ability to elicit neutralizing antibodies and evaluated the DII FL -mutated immunogen as the next-generation vaccine candidate in mice.

Identification of the DII FL epitopes on genotype I (GI) VLPs using flaviviral mAbs
JEV GI VLPs induced cross-neutralizing antibodies against four JEV genotypes and provided cross-protection against GI and GIII viruses in pigs (41).In this study, we used flaviviral mAbs to identify the DII FL epitopes on GI VLPs.Because the structure of VLP was unavailable, we analyzed the conformational location and antibody accessibility of DII FL on a simulated E-dimer structure with GI virus sequence or on an E-dimer structure of GIII JEV virion (PDB: 5WSN) (Fig. 1a and b).The 98 to 110 residues of DII FL were conformationally interfered by DIII and the N-linked glycan at residue 154 on another monomer (Fig. 1a).The 98, 102, 104, 106, 107, and 110 residues had higher accessibility of 35% for antibodies than the other residues on the simulated E-dimer of the GI virus and GIII JEV E dimer (Fig. 1b).The estimated accessibility of residues 100, 101, 105, and 108 varied between the two E-dimer structures.This discrepancy may be linked to the utilization of either recombinant E proteins (PDB: 3P54) or virus particles for structure resolution (PDB: 5WSN).Flavivirus DII FL epitopes were located at residues 101, 104, 106, 107, and 108 but the precise residues remained unknown for GI VLPs (37,(42)(43)(44).Flavivirus VLPs encoding W101G, G104H, G106K, L107D, or F108A were secretable and  (c) The fusion loop sequences of pVJGI WT and mutant VLP plasmids encoding W101G, G104H, G106K, L107D, or F108A mutations were presented.(D) Genotype I (GI) WT and mutant VLPs were collected from COS-1 cells transfected with the VLP-expressing plasmids.The endpoint titer of the secreted VLPs was measured by antigen-capture enzyme-linked immunosorbent assay (Ag-ELISA).The mean and standard deviation were presented.(e) Ag-ELISA was used to measure the interaction of GI WT and mutant VLPs with group cross-reactive, JE serocomplex cross-reactive, and type-specific mAbs.The mAb binding activity against mutant VLPs was compared to WT VLPs and calculated by [Log endpoint titer against (mutant VLP／ WT VLP) ] ×100%.The binding activity was presented as mean ± SD and color from deep blue to light blue based on the mean values of ˃100%, 10%-100%, ˂10%-1%, or ˂1%.All experiments were performed in triplicate.
interactive with flavivirus mAbs (43,44).Thus, we generated mutated GI VLPs with a single DII FL W101G, G104H, G106K, L107D, or F108A mutation (Fig. 1c).Antigen-capture (Ag)-ELISA confirmed the yield of the secreted wild-type (WT) and mutant VLPs were similar except for the G104H VLP, which could only be detected in concentrated culture fluid (Fig. 1d).As compared to WT VLPs, both W101G and F108A VLPs had less than 10% of binding activity against all group cross-reactive mAbs as well as 6B4A-10 mAb.By contrast, G106K and L107D VLPs increased the binding activity to the neutralizing T16 mAb and JEV type-specific 112 mAb (Fig. 1e).We selected G106K and L107D mutations to evaluate the role of DII FL in inducing neutralizing antibodies.We further included the W101G mutation to investigate the contribution of group cross-reactive antibodies to the neutralizing activity since the W101G mutation exhibited more reduction in binding activity with 23-1 and 6B4A-10 mAbs than the F108A mutation.We excluded G104H VLPs in future studies due to the low level of VLP secretion.

The impact of DII FL G106K/L107D and W101G/G106K/L107D mutations on the particle formation and antigenic activity of VLP
We then introduced DII FL G106K/L107D (KD) mutations or W101G/G106K/L107D (GKD) mutations onto WT VLP-expressing plasmid (pVJGI WT) to generate pVJGI KD and pVJGI GKD, respectively (Fig. 2a).Compared to pVJGI WT-transfected COS-1 cells, the pVJGI KDand GKD-transfected cells showed similar expression of intracellular staining patterns for JEV antigens and secreted a comparable yield of GI KD and GKD VLPs (Fig. 2b).We analyzed particle formation of the mutant VLPs by 5% to 25% of the sucrose density gradient.The secreted WT and mutant VLPs were sedimented in comparable layers and condensed at the 3rd fraction in the gradient (Fig. 2c).We further detected E and prM proteins incorporated onto GI WT, KD, and GKD VLPs as well as JEV particles in West ern blotting (Fig. 2d).These results suggested that KD and GKD mutations maintained the ability of VLP formation and secretion and preserved its VLP density or structure undifferentiated from WT VLP.
We subsequently analyzed the impact of KD and GKD mutations on the antigenic characteristic of respective VLP by the Ag-ELISA and Western blotting using flaviviral mAbs (Fig. 3).As expected, GI KD and GKD VLPs exhibited a similar pattern in reducing binding activity with all group cross-reactive mAbs and two serocomplex cross-reactive mAbs.However, increased binding activity of G106K or L107D VLPs with T16 and 112 mAbs was not detected against KD or GKD VLP (Fig. 3a and b).These results showed that GI KD and GKD VLP altered the group and supercomplex cross-reactive epitopes while preserving the T16 and type-specific epitopes.

G106K/L107D and W101G/G106K/L107D mutations did not influence the immunogenicity of VLPs
The antigenic characterization suggested that GI KD VLP and GKD VLP might induce variant antibody profiles targeting DII FL 106/107 or 101/106/107 residues as compared to WT VLPs.We compared the immunogenicity of these three types of VLPs in immunized mice and found that GI KD and GKD VLP induced similar IgG titers (10 5.07 ± 0.39 and 10 5.01 ± 0.78 ) against the homologous VLP as compared to the WT VLP (10 5.13 ± 0.35 ) (P ˃ 0.05) (Fig. 4a).We estimated the IgG antibody response targeting DII FL 106/107 and 101/106/107 residues in the immunized mice and found that KD VLP-and GKD VLP-immunized mice elicited a non-significantly different IgG antibody response to DII FL 106/107 and 101/106/107, respectively (P ˃ 0.05) (Fig. 4b and d).The Western blotting results also showed a similar staining pattern (Fig. 4c and e).These results demonstrated that GI WT VLP, KD VLP, and GKD VLP as well as their DII FL residues had similar immunogenicity even though the antigenic landscapes were different.

DII FL G106K/L107D and W101G/G106K/L107D mutations reduced the VLP ability to elicit neutralizing antibodies
The neutralizing antibody titer is considered as an immune correlate of protection against JEV infection but the role of DII FL -reactive antibody remains unclear in correlat ing the protective immunity (14).Reduction of DII FL immunogenicity on immunogens could induce a higher neutralizing antibody response by altering an antibody profile (25,26).We observed similar immunogenicity of the DII FL residues on the mutant and WT VLPs (Fig. 4).If DII FL -reactive antibodies exhibit non-or weak neutralizing activity, KD and GKD VLPs might elicit similar neutralizing antibody response with WT VLPs in mice.Therefore, we measured neutralizing activity of the immunized mice sera to investigate whether KD and GKD mutations influenced the VLP ability to induce neutralizing antibodies (Fig. 5a).Unexpectedly, the KD and GKD VLP-immunized sera showed significantly lower FRμNT 50 titers (59.6 and 32.4) than the titer of 146.6 for the WT VLP-immunized sera (P ˂ 0.05) (Fig. 5a).The quality of antibody response was evaluated by calculating the ratio of FRμNT 50 titer to the homologous IgG titer for each mouse (Fig. 5b).Overall, GI KD and GKD VLP-immunized mice elicited lower quality of the neutralizing IgG antibody response than WT VLP-immunized mice after third-dose immunization (P ˂ 0.05).These results suggested that DII FL 106/107 and 101/106/107 residues were critical for GI VLP to induce potently neutralizing antibodies against GI JEV.We observed the mutant VLPs elicited lower neutralizing antibodies than the WT VLP while inducing similar IgG antibody titers to the homologous VLPs and their DII FL region (Fig. 4).These results implied that W101, G106, and/or L107 in the WT VLPs might elicit antibodies that have a stronger neutralizing activity than the one induced by the mutant VLPs.Thus, we measured the contribution of anti-DII FL antibody response on neutralizing activity by comparing the difference in percent focus reduction of the homologous VLP-depleted sera to the heterologous VLP-depleted sera (Fig. 6).Most WT VLP-immu nized sera exhibited lower activity in focus reduction after the antibody depletion with the homologous VLP than with KD or GKD VLPs (Fig. 6a and d).We estimated that the WT 106/107-and 101/106/107-reactive antibodies contributed to 40.0% (95% CI in 2.9% to 77.1%) and 56.2% (95% CI in 44.5% to 67.9%) neutralizing activity in sera from mice immunized with WT VLP, respectively (Fig. 6c and f).By contrast, KD-and GKD-reactive antibodies only contributed 16.5% (95% CI in −7.0% to 39.9%) and 17.0% (95% CI in −10.6% to 48.7%) of the neutralizing activity in sera from mice immunized with KD VLP and GKD VLP, respectively (Fig. 6c and f).The small sample size used in this assay may have resulted in a nonsignificant difference observed in the contribution of DII FL -reactive antibodies between sera from WT VLP-and mutant VLP-immunized mice.However, these results did imply that the WT VLPs elicit a portion of neutralizing antibodies binding to WT 106/107 and 101/106/107 residues.To further dissect the neutralizing activity of anti-DII FL antibodies, we constructed infectious clones to generate recombinant GI viruses encoding single, double, and triple mutations at DII FL 101, 106, and 107 residues.Unfortunately, all mutated full-length clones were not viable, and we were unable to recover the infectious and mutant viruses after two passages (Fig. 7).

The involvement of DII FL 101/106/107 in eliciting cross-protective neutraliz ing antibodies
The KD and GKD mutations reduced the ability of VLP to induce neutralizing antibodies binding to WT 101/106/107 residues (Fig. 5 and 6).Since the amino acid sequences of DII FL 101/106/107 residues are conserved across flaviviruses (45), the DII FL mutations might affect the VLP-induced cross-protective neutralizing antibodies against all or some flaviviruses.We measured the FRμNT 50 titers against GI virus, GIII virus, WNV, and DENV2 for sera from mice immunized with KD and GKD VLPs (Fig. 5a and 8).The sera from mice

GI DII FL -mutant VLPs reduced the ability to elicit protective immunity against JEV infection
We observed that KD and GKD VLPs induced lower neutralizing antibody titers (Fig. 5) and that these titers were positively correlated with JEV vaccine potency (14).The VLP-immunized mice were challenged with GI YL2009-4 virus after the third dose of immunization.All the phosphate-buffered saline (PBS)-immunized mice were dead at 10 days post-challenge.All the GI WT VLP-immunized mice (12/12) survived for 21 days, while only 46% (6/13) and 31% (4/13) of the KD and GKD VLP-immunized mice survived, respectively, with an average survival time of 12 and 10 days (Fig. 9).These results demonstrated that KD and GKD mutations significantly reduced the ability of GI VLPs to protect mice from lethal JEV infection (P ˂0.05).The result was consistent with our observation that the neutralizing activity of mutant VLP-immunized mouse sera was 2.5-fold to 4.5-fold lower than that of WT VLP-immunized mouse sera (Fig. 5a) and suggested that JEV WT-DII FL 101/106/107 residues binding antibodies might play an important role in protective immunity against JEV infection.

Reduced induction of neutralizing antibodies results in a lower survival rate in GI DII FL -mutant VLP-immunized mice
We further investigated whether the reduced production of neutralizing antibody response contributed to the decreased survival rate in the KD VLP-and GKD VLP-immunized mice.We believe using the same intracranial route of the JEV challenge in the passive transfer experiment would provide a more comparable context to Fig. 9. Naive mice received pooled sera from PBS-immunized-, GI WT VLP-, GI KD VLP-, or GI GKD VLP-immunized mice and were subsequently challenged with GI JEV via the intracranial route (Table 1).We observed an average threefold to fourfold decline in neutralizing titers of injected mouse serum after circulating from the peritoneal cavity to the blood at 16 hours post-inoculation.All mice that received sera from PBS-immunized mice succumbed to the lethal virus challenge.Mice that received sera from WT VLP-immunized mice exhibited a GMT of 31.1 for FRμNT 50 against GI YL2009-4 virus before the virus challenge, with 46.7% (7/15) of them surviving after the challenge.By contrast, those that received sera from KD VLP-immunized mice and GKD VLP-immunized mice displayed GMTs of 10.5 and 5.9, respectively, and only 6.7% (1/15) and 0% (0/15) of them survived the challenge.This result suggests that DII FL KD and GKD mutations impair the ability of GI VLPs to induce neutralizing antibodies, resulting in a significantly lower protective rate (P ˂ 0.05) against homologous JEV infection in VLP-immunized mice.

DISCUSSION
This study elucidates the role of DII FL in eliciting protective immunity against JE serocomplex viruses by comparing the immunogenicity and protective potency of GI DII FL -mutated VLPs with GI WT VLPs in mice.We observed that the mutant DII FL reduced reactivity with the group and JE serocomplex cross-reactive antibodies but maintained similar immunogenicity compared to the WT VLP.The DII FL -mutated VLPs elicited less neutralizing antibodies against JEV and WNV compared to WT VLPs.Mice immunized with DII FL -mutated VLPs had a lower survival rate than those immunized with WT VLPs after lethal JEV challenge.That lower viral neutralizing activity of sera collec ted from DII FL -mutated VLP-immunized mice resulted in reduced protective efficacy against homologous virus infection.These results demonstrated and supported that JEV DII FL 101/106/107 residues were involved in inducing cross-neutralizing antibodies and protective immunity against virus infection.Our findings, along with previous studies, emphasized the pivotal role of neutralizing antibodies in protecting mice against lethal  a Fifteen naïve female BALB/c mice per group (PBS, GI WT VLP, GI KD VLP, or GI GKD VLP) were intraperitoneally injected with pooled sera collected from PBS-immunized mice, GI WT VLP-immunized mice, GI KD VLP-immunized mice, or GI GKD VLP-immunized mice after the third dose of immunization.
b The FRμNT 50 titers of pooled sera collected from PBS-immunized mice, GI WT VLP-immunized mice, GI KD VLP-immunized mice, or GI GKD VLP-immunized mice were measured against the GI YL2009-4 strain.FRμNT 50 titers were presented as geometric mean titers (GMT) ± standard deviation (SD).c Serum specimens were collected from mice at 16 hours post-administration of the pooled sera and subsequently measured for FRμNT 50 against the GI YL2009-4 strain.FRμNT 50 titers were presented as GMTs ± SD. d The mice were intracranially challenged with a 100-fold LD 50 dose of GI YL2009-4 virus at 18 hours post-administration of the pooled sera.Survival of the mice was recorded until 21 days post-challenge.Differences in mouse survival curves were compared to the GI WT VLP group using the Log-rank test.
e Statistical significance was considered as P < 0.05.
JEV infection (14,46).Nonetheless, we cannot rule out the potential impact of the DII FL mutation on the induction of T-cell responses.
A single DII FL mutation, G104H, significantly reduced the GI VLP yields in this study.This reduction was also observed for DENV-2 and GIII JEV VLPs (38,42).The structural location of G104 residue on an immature or mature virion suggests that G104H mutation might interfere with the interaction between DII FL and prM or glycan during VLP assembly or/and secretion (47) (Fig. 1).Introducing a mutation at W101 may potentially disrupt the formation of the E dimer due to its structural location.However, our findings, along with previous studies, have demonstrated the successful generation of flavivirus VLPs encoding W101G or W101A mutations, which retain the antigenic epitopes recognized by type-specific or DIII-reactive mAbs (43,44).The fusion loop epitope (FLE)-specific mAbs, which recognized the W101 residue alone or with other DII FL residues, showed low neutralizing activity against flavivirus (42,(48)(49)(50).The newly identified EDE-specific mAbs also recognized the DII FL 101 residue had potent cross-neu tralizing activity against the DENV serocomplex or DENV-ZIKV super serocomplex group (48,51).Some DII FL -reactive JE serocomplex mAbs that recognized residue(s) on the nearby DIII of the other E monomer might have higher neutralizing activity than the other serocomplex mAbs targeting DII FL alone (37,38,43,50,52).JEV type-specific mAbs recognizing DII FL remained unidentified from the virus-infected or vaccinated mice and humans (35).While observing a drastic reduction in the neutralizing activity against JEV than WNV between WT and DII FL -mutated VLP-immunized sera, we speculate that the impact on induction of serocomplex cross-reactive antibodies might have higher neutralizing activity against JEV than WNV.We were unable to exclude the potential impact of DII FL mutations on the induction of type-specific and neutralizing antibodies although two type-specific mAbs used here remained binding activity with the DII FLmutated VLPs.A human-derived mAb-targeted DII FL 101, 104, and 106 residues could prevent mice from WNV and DENV infections (53,54).E53 mAb, targeted DII FL 106 and 107 but not 101 residue, could still provide partial protection against WNV infection (50).These studies suggested that DII FL 101, 106, and 107 residues might be involved in inducing non-neutralizing as well as cross-neutralizing antibodies.
Here, we demonstrated DII FL 106/107 residues were involved in inducing neutralizing antibodies against JEV and these two residues together with the 101 residue were involved in eliciting neutralizing antibodies against JEV and WNV.However, a recent study showed that JEV GIII VLP-expressing plasmids encoding single or double DII FL G106V and L107F mutations performed comparably to the wild-type plasmid in eliciting neutralizing antibodies against JEV (39).This discrepancy could be due to the use of different amino acids at the DII FL 106 and 107 residues, the method to measure neutralizing antibody titers, or the type of immunogens (VLP vs DNA) used in the study (39).By contrast, DENV2 VLP-expressing plasmids encoding DII FL G106R/L107D mutations elicited comparable neutralizing antibodies to the wild-type plasmid by reducing DII FL immunogenicity and manipulating antibody profile (30).This result was consistent with evidence that DENV DII FL G106/L107 residues were not the part of highly neutralizing EDE epitope (48).Our study showed that JEV DII FL maintained its immuno genicity after introducing 101/106/107 mutations into the GI VLP, suggesting that DII FL 106 and 107 residues may play a different role in inducing antibodies and protective immunity against JE serocomplex virus compared to DENV serocomplex virus.However, DENV1 and ZIKV immunogens encoding DII FL 106/107 or 101/107 mutations elicited a lower neutralizing antibody response than its WT immunogens (29,55).A single DII FL F108A mutation altered particle maturity for ZIKV VLP and reduced its ability to induce protective immunity (32).However, these studies did not address the independent role of DENV1 and ZIKV DII FL in eliciting protective immunity.
The fusion loop is conserved in flavivirus and antibodies with footprint-covered fusion loop residues dominate the immune response.DENV infection induces most antibodies targeting the DII FL 101 residue, followed by the 106/107/108 residues in humans (56,57).By contrast, JEV-or WNV-infected humans produced less antibody responses targeting DII FL 106 and 107 residues (38).Our study also demonstrated low immunogenicity of DII FL on GI WT VLP, KD VLP, and GKD VLP.DENV or ZIKV infection induced nonneutralizing and anti-DII FL antibodies associated with developing ADE-related severe dengue upon subsequent infection (11,58,59).DENV and ZIKV immunogens with DII FL mutations reduced the occurrence of ADE in vitro and ex vivo by reducing DII FL immunogenicity and/or the binding activity of anti-DII FL antibodies to the virion (29)(30)(31)(32)55).The ADE of JEV and WNV infection was observed in vitro assay only; and this phenomenon requires the support of epidemiological evidence (4,13,(60)(61)(62).Further studies are required to evaluate the benefit of using DII FL -mutated vaccine candidates of JE serocomplex virus to reduce the risk of ADE for DENV or WNV infection after JEV vaccination.
Our results and other studies suggest a different role of DII FL in DENV/ZIKV and JEV to elicit protective and/or pathogenic immunities.DII FL -mutated immunogens could be a superior next-generation vaccine candidate against DENV and ZIKV but not for JEV in terms of vaccine efficacy and safety, particularly since DENV vaccine clinical trials showed the potential to enhance disease severity among seronegative children after vaccination (8,63).In addition to E antibody responses, the elicitation of T-cell responses and/or NS1-induced antibody responses proved to be crucial for flavivirus vaccines (64)(65)(66).The concurrent activation of CD4 + and CD8 + T-cell responses in mice demonstrated enhanced efficacy against JEV infection compared to individual CD4 + or CD8 + T-cell responses alone (67,68).Mice exclusively receiving vaccine-induced CD4 + T cells exhibited a higher survival rate than those receiving transferred CD8 + T cells following a lethal JEV challenge (68).A more robust interferon-gamma (IFN-γ)-dominated CD4 + T-cell response was correlated with improved recovery from JE clinical outcomes, with the potential capacity to recognize non-DII FL regions of the E proteins (69).Both CD4 + and CD8 + T-cell responses were essential for controlling and recovering from WNV infection in mice (70)(71)(72), and virus-specific CD8 + T cells were more abundant than CD4 + T cells in WNV-infected humans (73).CD4 + and CD8 + T cells induced by DENV and ZIKV contributed to protection against viral infection (10,(74)(75)(76)(77), with a particular emphasis on the critical role of CD8 + T-cell responses against heterologous serotype infections or flavivirus infections (10,(76)(77)(78).In contrast to WNV, ZIKV, and DENV, limited evidence supports the pivotal role of CD8 + T-cell responses against JEV infections.Further studies are required to determine whether DII FL mutations have an impact on vaccine-induced T-cell responses.
In conclusion, JEV vaccines should preserve the intact of DII FL to elicit neutraliz ing antibodies and provide protective immunity against JEV or JE serocomplex virus infections.The DII FL L107F mutation has been used to develop attenuated JE serocom plex virus vaccines based on the commonly used GIII vaccine strain (79)(80)(81)(82).Further research is needed to clarify the independent role of DII FL W101, G106, or L107 in inducing protective B cells and T-cell immunity against JE serocomplex virus infection (12).

Expression and purification of VLPs
We transferred 30 µg of pVJGI WT, W101G, G104H, G106K, L107D, F108A, KD, or GKD VLP plasmids into 10 7 COS-1 cells in 0.4-cm-electrode-gap cuvettes by electroporation at 250 V/ 975 µF with a Bio-Rad Gene Pulser II (Bio-Rad Laboratories, Hercules, CA).The transformed cells were recovered in a 37℃ incubator overnight and then transferred to a 28℃ incubator to enhance the secretion of VLPs in serum-free medium (SFM4Meg aVirTM, HyCloneTM) with the addition of 1× cholesterol (Gibco) (38).After 3-to 5-day incubation, WT and mutant VLPs were collected from the supernatant of the transfec ted COS-1 cells.We concentrated the VLPs by Amicon Ultra-15 Centrifugal Filter Unit (100,000 molecular weight cut-off, Merck, United States) and then proceeded to 20% sucrose density centrifugation with 19,000 rpm at 4℃ for 16 hours.The VLP pellets were suspended in 1× TNE buffer at 4℃ overnight.We layered 25% to 5% of sucrose into an ultracentrifuge tube (Beckman Coulter Inc., CA) from bottom to top and incubated it at 4℃ overnight to form the density gradient.Then, the concentrated VLPs were analyzed by 5% to 25% of sucrose density gradient centrifugation with 25,000 rpm at 4℃ for 3 hours.One milliliter of fractionating layers was collected from top to bottom.We detected VLP antigens in the fractionating layers by antigen-capture enzyme-linked immunosorbent assay (Ag-capture ELISA).The peak OD 450 values of GI WT, KD, or GKD VLPs in the gradient were purified using Amicon® Ultra-4 Centrifugal Filter Unit (100,000 molecular weight cut-off, Merck, United States) and recovered in 1× PBS.The purified VLPs were used to immunize mice.the pooled mouse sera obtained from PBS-, GI WT VLP-, GI KD VLP-, or GI GKD VLP-immu nized mice.At 16 hours post-serum transfer, mouse sera were collected and assayed for the neutralizing activity.Previous research estimates that 0.1% to 1% of circulating antibodies can traverse the intact blood-brain barrier (BBB) and enter the murine brain (86,87).Intracranial injection of JEV could disrupt the BBB, leading to heightened levels of passively transferred serum IgG in murine brains, as observed in other neurotropic flaviviruses such as WNV (88,89).Therefore, at 18 hours post-serum transfer, each mouse received an intracranial injection of a 100-fold LD 50 dose of the JEV GI YL2009-4 strain.Survival rates of the mice were recorded twice daily until 21 days post-challenge.

Mouse IgG antibodies-capture enzyme-linked immunosorbent assay
To detect IgG antibody responses in the immunized mice, we used a mouse IgG antibodies-capture enzyme-linked immunosorbent assay (GAC-ELISA) as described previously (30,57,90).Briefly, we coated 96-well immunoplates (Sigma-Aldrich, St. Louis, MO) with goat anti-mouse IgG (H + L) (KPL, Gaithersburg, MD) at 37°C for 1 hour and blocked them with StartBlock blocking buffer (Pierce, Rockford, Ill.).We then added serially diluted mouse serum specimens to the wells and incubated them at 37°C for 90 minutes.After washing the wells with 1× PBST to discard uncaptured antibodies, we added negative COS-1 cell antigens, GI WT VLPs, GI KD VLPs, or GI GKD VLPs to the wells, which interacted with the captured IgG antibodies at 4°C overnight.On the following day, VLPs were recognized by rabbit anti-JEV polyclonal antibodies at 37°C for one hour, and the VLP-captured antibodies reacted with peroxidase-conjugated goat anti-rabbit IgG (H + L) (Jackson ImmunoResearch, West Grove, PA) at 37°C for another 1 hour.We described the steps of coloring and OD detection in Ag-ELISA.We calculated the ELISA endpoint titers of mice sera as the reciprocal dilution reaching a P/N ratio equal to three in sigmoidal dose-response regression of GraphPad Prism version 5.01.We calculated the total IgG titer by the endpoint titer of sera against WT, KD, and GKD VLPs for GI WT VLP-, GI KD VLP-, or GKD VLP-immunized mice.The DII FL 106/107 IgG titer was calculated by subtracting the endpoint titer of sera against KD VLPs or WT VLPs from the total IgG titer for GI WT VLP-or KD VLP-immunized mice, respectively.The DII FL 101/106/107 IgG titer was calculated by subtracting the endpoint titer of sera against GKD VLPs or WT VLPs from the total IgG titer for GI WT VLP-or GKD VLP-immunized mice, respectively.We gave the endpoint titer below the cut-off dilution of 400 an arbitrary value of 200.

Focus-reduction micro-neutralization titer assay
We used the focus-reduction micro-neutralization titer (FRμNT) assay to measure the neutralizing activity of mice serum samples as described previously (91).Briefly, we seeded 2.25 × 10 4 Vero cells on 96-well plates and incubated them for 18-20 hours in a 37℃ incubator with 5% CO 2 .The mice sera were inactivated at 56℃ for 30 minutes and serially diluted.The 100 focus-forming units of the virus were incubated with the serial dilutions of mice sera at 37℃ for 1 hour.The serum-virus mixture infected the cells at 37℃ for another hour.The infected cells were then overlaid with 1% methylcellulose in DMEM with 2% FBS and incubated in a 37℃ incubator with 5% CO 2 for 30 hours (JEV and WNV) or 72 hours (DENV2).The wells were washed with 1× PBS to remove the overlaid reagents and then fixed with 75% acetone at room temperature for 20 minutes.Mouse anti-JEV, WNV, or DENV2 HIAF bound to the virus antigens in the fixed and dried cells at 37℃ for 40 minutes, and peroxidase-conjugated goat anti-mouse IgG (Jackson ImmunoResearch, West Grove, PA) captured the bound antibodies at 37℃ for another 40 minutes.The foci were observed after adding the Vector-VIP peroxidase substrate kit SK-4600 (Vector Laboratories, Burlingame, CA).The images of foci were acquired by IMMUNOSPOT S6 Universal Analyzer (CELLULAR TECHNOLOGY).We calculated the FRμNT 50 titer as the reciprocal dilution reaching a 50% reduction in fuci number compared to the virus control in sigmoidal dose-response regression of GraphPad Prism version 8. We assigned an arbitrary value of 5 to the FRμNT 50 titer below the cut-off value of 10 for calculating the geometric mean titer (GMT).

Depletion of VLP-reactive antibodies from the mouse serum specimen
Individual mouse serum used for the depletion study was calculated and diluted to have an eightfold higher end-point FRμNT 50 serum titer.The diluted mouse serum was incubated with 5 × 10 6 COS-1 cells transfected with VLP-expressing plasmids for an hour at 37℃.After incubation, the VLP-reactive antibodies were captured by the transfected cells, and the remaining antibodies, namely the VLP-reactive depleted serum specimen, were collected in the supernatant.We subsequently measured the neutralizing activity of the VLP-depleted sera in the FRμNT assay.The percentage of focus reduction was calculated for COS-1 cell control-depleted, WT VLP-depleted, KD VLP-depleted, and GKD VLP-depleted sera compared to 100% of focus reduction for the serum sample without depletion.

The construction and generation of recombinant and mutant GI JEVs
An infectious clone encoding the full genome of the GI YL2009-4 virus (pCMV GI WT) was previously established (92).The introduction of DII FL mutation(s) into pCMV GI WT was performed through PCRs (using KOD Hot Start DNA Polymerase from Merck, United States) with the primers listed in Table 2.The PCR treated with DpnI enzyme was then transferred into competent cells.Mutated clones were extracted from the transformed competent cells using a Mini-prep kit (Qiagen) and sequenced to confirm the complete viral genome insert.Recombinant GI JEV was generated as previously described (92).Briefly, BHK-21 cells were transfected with a mixture of 1 µg of the infectious clone, Opti-MEM (Life Technologies), and Lipofectamine 2000 (Life Technologies).After transfection for 5 hours at 37℃, the cell supernatant was replaced with the culture medium.Following a 3-to 4-day incubation, recombinant viruses secreted from transfected cells were harvested and subsequently amplified twice in BHK-21 cells.The production of rJEVs was detected in transfected cells using an immunofluorescence assay (IFA) with mouse anti-JEV HIAF.Virus plaques were identified by micro-plaque assay.Viral RNA was extracted from the supernatant of virus-infected BHK-21 cells using the RNeasy mini kit (Qiagen) and transcribed into cDNA with the JEV 3′UTR primer 5′-AGATCCTGTGTTCTTCCTCA-3′ using the Super script III transcription reaction (Thermo Fisher Scientific).The full length of the JEV E gene was amplified by PCR with a pair of GI JEV E primers (5′-CGTGTGGTaTTCAC TATTCTC-3′ and 5′-CATTCAGTTCGTCCCGCACA-3′).

Statistical analysis
We compared the antibody responses between GI WT VLP-, KD VLP-, and GKD VLP-immu nized mice by one-way ANOVA with Tukey's multiple comparisons test.The comparison between the two groups was analyzed by a two-tailed unpaired t-test.We analyzed the survival curve of PBS, GI WT VLP-, and GI GKD VLP-immunized mice using the Log-rank test.Statistical significance was recognized as P ˂0.05.All these statistical analyses were conducted in GraphPad Prism version 8.

FIG 1
FIG 1 Structural location and contribution of single amino acid residues in the fusion loop to the binding of flaviviral monoclonal antibodies (mAbs).
(a) SWISS-MODEL (https://swissmodel.expasy.org/) was used to model the E dimers of the JEV GI YL2009-4 virus with a PDB template of 3P54.The structural location of the fusion loop (on the simulated E dimers with 3P54) and the N-linked glycan (adapted from the PDB template of 5N0A) were presented in PyMOL (https://pymol.org/2/).The fusion loop and the mutated residues were colored in green and magenta, respectively.(b) Dezyme software (http:// www.dezyme.com/)estimated the solvent accessibility of residues on the simulated E-dimer with the 3P54 template and the E-dimer on JEV virion (PDB: 5WSN).

FIG 3 Full 6 FIG 4
FIG3 The antigenicity of GI WT, KD, and GKD VLPs.The reactivity of mouse anti-JEV HIAF (MHIAF), group cross-reactive (GCR), complex cross-reactive (CCR), and type-specific (TS) mAbs with KD and GKD VLPs was measured in Ag-ELISA (a) and WB analysis (b).The mAb binding activity against mutant VLPs was compared to WT VLPs and calculated by [Log endpoint titer against (mutant VLP／ WT VLP) ] ×100%.The binding activities were presented as mean ± SD and colored from deep blue to light blue based on their mean percentage of binding (˃100%, 10%-100%, ˂10%-1%, or ˂1%).All experiments were conducted in triplicate.The antigen samples, devoid of 2-mercaptoethanol treatment, were subjected to boiling and subsequently analyzed in a WB assay.The percentage of mAbs binding activity against E proteins was measured using ImageJ software (https://imagej.nih.gov/ij/) and was indicated below the bands.NA stands for not available; WT refers to GI WT VLP; KD refers to GI KD VLP; GKD refers to GI GKD VLP; PC stands for JEV culture media as a positive control; NC stands for cell culture supernatant as a negative control.

FIG 5
FIG 5 The neutralizing antibody titer against GI YL2009-4 virus and quality of the total IgG antibody response induced by GI WT, KD, and GKD VLPs.(a) The neutralizing antibody titers of the mice sera (WT VLP-immunized mice, n = 21; KD VLP-immunized mice, n = 13; GKD VLP-immunized mice, n = 21) were measured by FRμNT 50 against GI YL2009-4 virus.(b) The neutralizing index for each mouse was calculated by the ratio of Log FRμNT50 against GI YL2009-4 virus to Log ELISA IgG endpoint titer .The box and whisker with 2.5 and 97.5 percentiles were presented.A significant difference (P ˂ 0.05) was analyzed using one-way ANOVA with Tukey's multiple comparisons test and indicated by an asterisk.

FIG 7
FIG 7 Generation of recombinant GI viruses encoding DII FL mutations.BHK-21 cells were transfected with either pCMV GI WT or pCMV GI mutant clones containing DII FL mutations.The recombinant viruses released from the transfected cells underwent amplification for two passages (P1 and P2) in BHK-21 cells.Mouse anti-JEV hyperimmune ascitic fluid (HIAF) was employed to detect prM/M and E proteins in transfected BHK-21 cells using immunofluorescence assay (IFA).Cellular nuclei were stained with DAPI.The viral RNA and infectivity of the recombinant viruses were assessed through RT-PCR and micro-plaque assay, respectively.

TABLE 1
Passive protection of mice against lethal JEV challenge following pre-treatment with sera from GI WT-, KD-, or GKD-VLP-immunized mice

TABLE 2
Primers used for construction of the mutant VLP-expressing plasmids and infectious clones