CHO-S, HEK293T, HEK293F, BHK-21 cells and Vero cells were obtained from ATCC. The adherent cells were cultured in Dulbecco’s modified Eagle medium (DMEM; Sigma) supplemented with 10% fetal bovine serum (FBS). CHO-S cells were cultured in CHO S4 medium (Duoning Biology) and HEK293F cells in 293 medium (Union), with 8 mM glutamine at 37°C in 5% CO2.

Four strains of dengue viruses [DENV-1(D19044), DENV-2(16681), DENV-3 (D191267) and DENV-4(GZ14D4)] were kindly provided by Professor Yiping Li (Sun Yat-sen University). DENV-1 was propagated in BHK-21 cells; DENV-2, DENV-3 and DENV-4 were propagated in Vero cells. Virus titers were quantified as described before and expressed as focus-forming units FFU/ml (25).

Based on our recent works regarding ferritin-based nanoparticle vaccines, a similar dengue vaccine was constructed (20, 23, 24). Briefly, four sequences of E protein (1-397aa) from different dengue viruses (Dengue virus 1 Thailand/AHF 82-80/1980, Dengue virus 2 Thailand/16681/84, Dengue virus 3 Philippines/H87/1956, Dengue virus 4 Philippines/H241/1956) were chosen. Amino acid sequences of DII (53-132 aa and 191-280 aa) in E protein were respectively replaced by sequences of ‘GGGGS’ and ‘S’, to form the new E13 protein (DENV-1: E113, DENV-2: E213, DENV-3: E313, DENV-4: E413). Then the N-termini were fused by the sequences of secretory signal peptide, while the C-termini were followed by the sequences of GvTagOpti from GvTagOpti (Gv)/SdCatcher(sd) connection system described previously (26). Four reconstructed sequences of E13-Gv (DENV-1: E113-Gv, DENV-2: E213-Gv, DENV-3: E313-Gv, DENV-4: E413-Gv) were codon-optimized for mammalian cells and cloned into the vector pLVX plasmid. HEK293T cells were co-transfected with the constructed pLVX plasmids, the psPAX2 (Addgene) plasmid and PLP/VSVG by utilizing polyethyleneimine (PEI, Sigma). The recombinant viruses in the supernatant were collected after 48 h and then infected CHO-S cells for another 48 h ( Figure S1C ). After expending cultivation of the infected cells, the supernatant was harvested by centrifugation at the day 7.

For the expression of Sd-ferritin, the Sd gene was first fused to the N-terminus of the ferritin (HPF) gene, and the DNA sequence of Sd-ferritin was cloned into the pET28a vector. The constructed plasmid was transformed into BL21 (Takara) prokaryotic expression bacteria. Sd-ferritin was expressed from E. coli.

Both E13-Gv and Sd-ferritin were purified via Size Exclusion Chromatography (GE Healthcare) on a Superose 6 column. The purified protein was quantified by a BCA protein assay kit and the western blotting was performed to confirm the purification of the protein. Four purified E13-Gv in separate were irreversibly covalently conjugated to the Sd-ferritin to generate the E13-HPF nanoparticle which was further purified by Size Exclusion Chromatography and confirmed by the Coomassie blue staining. The tetravalent DENV nanoparticle vaccine (E13-HPF vaccine) consisted of a 1:1:1:1 mix of four E13-HPF nanoparticles.

For ELISpot assay, N-terminus of DENV-2 DIII protein was fused with 6-his-tag, and then the protein was expressed in HEK293F cells and purified with a nickel column.

Transmission electron microscopy (TEM) grids of HPF and E13-HPF vaccine were proceeded to negative-stain electron microscopy. Briefly, each sample (0.05 µg/µl) was applied to glow-discharged electron microscopy grids covered by a thin layer of continuous film and stained with 2% uranyl acetate. Then images were recorded at a magnification of 120,000 × on Talos L120C microscope (ThermoFisher) operating at an acceleration voltage of 120 kV. Scale bars represented 100 nm.

AG6 mice were a gift from Professor Zhiwei Wu (Nanjing University) and six-week-old female BALB/c mice were purchased from Gempharmatech Co., Ltd. These mice were raised in SPF facilities at the Laboratory Animal Center of Sun Yat-sen University. AG6 mice and BALB/c mice were immunized subcutaneously (s.c.) with 10 μg E13-HPF vaccine and 10 μg HPF proteins as control. Unless noted, antigens were mixed with an equal volume of Alum (InvivoGen) adjuvant. Sera were collected and heat inactivated at 56°C for 30 min.

For immunized AG6 mice challenge, 6×10⁵ FFU of DENV-2 16881 strain or 10⁵ FFU of DENV-3 D191267 strain were injected intravenously (i.v.) into each mouse. Survival analyses of the challenged AG6 mice were recorded every 2 days, including survival rate and clinical scoring. In order to detect the viremia 3 days post infection (dpi), AG6 mice serum was collected to exact RNA through the viral RNA kit (Magen).

For immunization strategy research, each immunized BALB/c mouse in different groups was conducted an intravenous injection of DENV-2 16881 strain (10⁷ FFU). Then BALB/c mice were sacrificed 7 dpi. Livers were harvested and fixed in 4% paraformaldehyde buffer (Servicebio) for 48 h, followed by embedding with paraffin. Longitudinal sections were performed on these tissues and sections (3-4 mm) were stained with hematoxylin and eosin (H&E).

Four types of 5 μg/ml E13 proteins were respectively coated on high-binding 96-well plates (Greiner Bio-one), overnight at 4°C. Then the plates were blocked with 100 μl 5% non-fat milk/PBS at room temperature for 1 h. After washing with PBS, immunized animal serum was serially diluted and added into each well at 37°C for 1 h. After washing with PBS/T (containing 0.1% Tween-20) three times, 100 μl HRP-conjugated goat anti-mouse secondary antibody (Invitrogen) at a dilution of 1:4000 was added at 37°C for another 1 h. The plates were washed four times before 50 μl TMB solution (eBioscience) was added to each well. After 5 min quenching reaction by adding 50 μl stop solution (Solarbio), measure absorption at 450 nm. GraphPad Prism 8.0 software for non-linear regression was used to calculate endpoint titers.

This method was described previously in many studies (27, 28). Briefly, Vero cells or BHK-21 cells were seeded at 4×10⁴ cells/well in 96-well plates and incubated overnight at 37°C in 5% CO₂. Mice serum was diluted at 1:10, then serially diluted 5-fold down a 96-well plate in serum-free OPTI-MEM. After adding the equal volume of dengue virus stocks at 70 FFU/well, the virus/serum mixture was incubated for 1h at 37°C. The confluent Vero cells were washed by PBS before the virus/serum mixture was added. Following another 1 h incubation at 37°C, the mixture was removed, and the cells were overlaid with 200 µl of 1.6% CMC in DMEM supplemented with 2% FBS. The plates were incubated for 2 days at 37°C.

After the supernatant was removed, the cells were fixed with 50 µl ice-cold methanol for 30 min at RT. After washing with 200 µl PBS, cells were incubated with 50 µl of rabbit anti-NS3 polyclonal antibody (GeneTex) which was 1:3000 diluted with PBS containing 1% BSA at 4°C overnight. Following three washes with PBS/T (0.1% Tween-20), 1:2000 diluted in PBS containing 1% BSA, FITC-conjugated secondary antibody (Abcam) was added 50µl to per well and the plates incubated for 1 h at 37°C. After washing 4 times with PBS/T, the plates were allowed to dry and imaged with ImmunoSpot microanalyzer. The 50% inhibition of infection of cells was represented by the decrease of number of DENV-infected cells in the sample wells compared to virus control wells. Reduction rates of the serial dilution assay were analyzed by Graphpad Prism 8.0 using non-linear regression to measure the NT50 titer.

Primary cells were isolated from bone marrow of BALBc/c mice as described (29). To generate bone marrow derived macrophage (BMDM), the isolated cells (1×10⁶/well) were cultured in a 24-well plate for 7 days in RPMI 1640 (10% FBS, penicillin, streptomycin) containing 20 ng/ml murine macrophage-stimulating factor (M-CSF, PeproTech). On day 8, 200 µl of mice serum was diluted at 1:25 in RPMI 1640 and were incubated with the equal volume of diluted virus stocks (MOI=0.1) for 1 h at 37°C in 5% CO₂, the virus-only well as mock control. After removing the supernatant, BMDM were incubated with the serum/virus mixture for another 1 h at 37°C. 400 µl of the RPMI 1640 (4% inactivated FBS, penicillin, streptomycin) was added to each well, and then the plates were allowed to maintain for 24 h at 37°C in 5% CO₂. On the next day, the cells were collected, and total RNA was extracted with a RNA extraction reagent kit (Magen). The RNA was stored in -20°C to be analyzed.

To obtain cDNA for qRT-RCR analysis, for RNA from all the samples, reverse transcription was carried out using HiScript II First-strand cDNA Synthesis Kit. RNA was quantified by SYBR Green real-time PCR (Universal SYBR RT-RCR Master Kit) with strain-specific primers ( Table S2 ). For ADE assays, viral RNA levels were normalized by housekeeping gene β-actin levels (β-actin: forward primer: 5’- GGCTGTATTCCCCTCCATCG-3’, reverse primer: 5’- CCAGTTGGTAACAATGCCATGT-3’). For viremia detection, the 3’-end specific gene of DENV-2(16881) or DENV-3(D191267) was cloned into a pMD-18T vector for standards. Then 10-fold serially diluted standards were subjected to SYBR Green real-time PCR to generate a standard curve for the quantification of viral RNA represented by the copy number. All the reactions were carried out on a QuantStudio 7 Flex System (Applied Biosystems) under the following reaction conditions: 95°C for 3 min, 40 cycles of 94°C for 15 s and 60°C for 30 s.

The spleen of each mouse was collected in PBS and homogenized, following incubation in ACKlysis buffer to remove red blood cells. Then splenocytes were stored in cell saving medium (KeyGEN) at -80°C until analysis. For analysis, cells were rapidly recovered at 37°C and balanced in RPMI 1640. After centrifugation, the cells were re-suspended with culture medium and remained at 37°C incubator for 0.5 h for complete cells recovery. Then the cells were washed with PBS and stained with the diluted fluorochrome-conjugated monoclonal antibodies for 30 min within PBS containing 0.5% BSA on ice. The following antibodies were used: eBioscience Fixable Viability Dye eFluor 780, anti-mouse CD3-AF647 (Biolegend), anti-mouse CD4-R718 (BD Biosciences), anti-mouse CD8-BV510 (Biolegend), anti-mouse CD62L-FITC (eBioscience), anti-mouse CD44-Percp/Cy5.5 (BD Biosciences), anti-mouse CD19-BV510 (Biolegend), anti-mouse B220-Percp/Cy5.5 (Biolegend) and anti-mouse CD38-R718 (BD Biosciences), anti-mouse IgG1-FITC (Biolegend) and anti-mouse IgG2a-AF647 (Biolegend). All flow cytometry data were acquired on FACS Aria II flow cytometer (BD Biosciences) and analyzed with FlowJo software.

The splenocytes from each group (n=5) was collected at week-26. The splenocytes were cultured with DENV-2 DIII at a concentration of 10 µg/ml and co-stimulated with 2 mg/mL anti-CD28 (Biolegend) at 37°C with 5% CO₂ for 20-24 h. Antigen specific splenocytes of BALB/c mice were detected by mouse IFN-γ ELISPOT kit and mouse IL-4 ELISPOT kit (Dayou) according to the manufacturer’s protocol. Antigen-specific spots were counted using ImmunoSpot microanalyzer. The number of spots was converted into the number of spots per 10⁵ cells.

Experiments were conducted independently in triplicate. Statistical analyses were conducted utilizing Graphpad Prism 8.0. The statistical details of specific experiments had been described in the figure legends and methods. All data are shown as the mean ± SEM. P values <0.05 were considered significant. ns: p>0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

The animal experiments were approved by the Ethics Committee of Zhongshan School of Medicine (ZSSOM) of Sun Yat-sen University on Laboratory Animal Care (Assurance Number: 2021002109).

To construct a DENV nanoparticle vaccine, we modified the E protein of the DENV and displayed it on a ferritin-based nanoparticle. E protein is the main target of nAbs to block DENV infection. The DII domain of E protein is the conservative domain that induces antibodies to mediate ADE.

We replaced the DII domain with a linker in the full-length E protein to ensure the natural structure of the reconstructed E protein (E13 protein) ( Figures 1A , S1A ). The Sd-coding sequence was fused genetically to the N-termini of ferritin (HPF) ( Figure S1B ). The Gv-coding sequence was fused to the C-termini of the E13 sequences of DENV 1–4 (E113-Gv, E213-Gv, E313-Gv, E413-Gv) ( Figure 1B ). Four stable CHO-S cells, expressing the four serotypes of DENV E13-Gv, respectively, were constructed. The purity of proteins in the cell supernatant was high ( Figure S1C ). Size-exclusion chromatography (SEC) was conducted to purify all of these proteins. Protein expression was analyzed in western blotting using a homemade anti-Gv antibody ( Figure 1C ). The purified DENV 1–4 E13-Gv was irreversibly and covalently conjugated to Sd-ferritin to generate DENV 1–4 E13-HPF (E113-HPF, E213-HPF, E313-HPF, E413-HPF), followed by purification by SEC ( Figure 1D ). Due to covalent conjugation between GvTagOpti and SdCatcher [this system had been described in detail in our work (26)], E13-Gv was incubated with Sd-ferritin to form E13-HPF, which could self-assemble into a 24-mer nanoparticle ( Figure 1B ). The molecular weight of E13-HPF was ~70 kDa, whereas that of E13-Gv and Sd-ferritin was ~40 kDa and ~30 kDa, respectively ( Figure 1C ). Besides, the yield of E13-Gv was ~30 mg/L as measured with a BCA protein assay kit. SEC revealed a single and smooth peak but also showed the high purity of the four types of nanoparticles ( Figure 1D ). To generate the tetravalent DENV nanoparticle vaccine (E13-HPF vaccine), four E13-HPF nanoparticles were mixed in equimolar proportions. The morphology and homogeneity of E13-HPF vaccine were also verified by transmission electron microscopy (TEM) ( Figure 1E ).

To evaluate the immunogenicity of the E13-HPF vaccine, BALB/c mice from the vaccine group were subcutaneously immunized with the E13-HPF vaccine (10 μg) formulated with an equal volume of alum adjuvant. The control group was given the same dose of HPF nanoparticles. All groups were given a booster dose at week-4 and sera were collected at week-6 ( Figure 2A ). Figure 2B indicates a significant difference in the serum IgG titer specifically against E13 between the vaccine group and HPF group. The titers of specific IgG against E13 of the four serotypes were ≥10⁴ until week-10 ( Figure S2A ). Hence, the E13-HPF vaccine could induce balanced and strong immunogenicity among all four serotypes of the DENV. To ascertain if DENV infection could be blocked by the specific IgG, the nAb level was measured. FRNT showed that the NT50 of four-serotype DENV was in the range 10² to 10³ ( Figures 2C , S2B ). These data indicated that sera from the vaccine group protected Vero cells or BHK21 cells from DENV infection effectively, although NT50 of DENV-1 and DENV-4 was slightly lower than that of the other serotypes of the DENV.

Except for pre-existing nAbs, rapidly reactivated memory B cells (MBCs) are important for antibody generation in the antiviral immune response (30, 31). Therefore, we isolated lymphocytes from the spleen of each immunized mouse at week-6. The percentages of total IgG1⁺ MBCs (CD19⁺ B220⁺ CD38⁺) and IgG2a⁺ MBCs were determined by flow cytometry ( Figure S2C ). Vaccine-immunized mice induced more IgG1⁺ MBCs and IgG2a⁺ MBCs ( Figure 2D ) than those of the control group.

To determine the ADE induced by vaccines in vitro, we applied a detection system based on bone-marrow-derived macrophages (BMDMs) from BALB/c mice. Studies have shown that ADE is an obstacle to developing a DENV vaccine because non-nAb or sub-nAb concentrations of DENV-specific antibodies can enhance viral replication in phagocytes expressing Fc receptors (13, 32, 33). In our study, diluted serum was pre-incubated with the DENV to form immune complexes and then the mixture was added to BMDMs. Cells were harvested after 24 h and RNA viral loads were detected by qRT-PCR. At a final serum dilution of 1:50, when different serotypes of the DENV infected BMDMs, the relative viral RNA load in BMDMs presented different fold increases with DENV-2-infected serum from wild-type mice ( Figure 2E ). However, significant enhancement of the relative viral RNA load in BMDMs in the presence of mice serum from the vaccine group was not observed. Hence, the E13-HPF vaccine that enriched E13 antigens in ferritin nanoparticles evoked a robust humoral immune response against the four serotypes of the DENV without the risk of ADE.

To evaluate the T-cell immune responses induced by the E13-HPF vaccine, we investigated central memory T (T_CM) cells and effector memory T (T_EM) cells in the spleen of mice 2 weeks after boost vaccination. Although clear understanding of cellular immunity in preventing DENV infection is lacking, studies have shown that antiviral CD8⁺ and CD4⁺ T cells play a key part in this process (34, 35). Recent studies have revealed that CD4⁺ T_EM cells and CD8⁺ T_EM cells function efficiently upon DENV challenge (35, 36). Exposure to pathogens can aid the corresponding self-renewal and differentiation of T_CM cells into T_EM cells to support further protection. We used flow cytometry to monitor these processes. The gating strategies of flow cytometry are shown in Figure S2D . Populations of CD4⁺ T cells and CD8⁺ T cells were defined further as T_CM cells (CD62L⁺ CD44⁺) and T_EM cells (CD62L^- CD44⁺). The E13-HPF vaccine increased the numbers of total CD4⁺ T_CM cells and CD8⁺ T_CM cells in the spleen significantly ( Figures 3A, B ). A higher percentage of CD4⁺ T_EM cells and CD8⁺ T_EM cells were activated in the nanoparticle-vaccinated group ( Figures 3C, D ) than in the control group, which indicated that the E13-HPF vaccine induced strong and long-term immune memory. Further, we detected DENV-specific T cells responses using DENV-2 DIII as a stimulus by ELISpot assay. Compared with other groups, we found that the vaccine group elicited more DIII-specific IFN-γ-secreting (Th1-biased cells) and IL-4-secreting (Th2-biased cells) splenocytes ( Figures 3E, F ). In the vaccine group, the number of IFN-γ-secreting splenocytes was much higher than that of IL-4-secreting splenocytes, indicating that E13-HPF vaccination drived Th1-biased cell responses in BALB/c mice.

To further investigate the protective efficiency of the E13-HPF vaccine, the same immune strategy was conducted in AG6 mice ( Figure 4A ). AG6 mice (which are deficient in IFN-α/β and IFN-γ receptors) have been reported to be susceptible to DENV infection and are used frequently in assessment of DENV vaccines (37, 38). Two weeks after the boost, two groups of vaccinated mice were challenged separately with a lethal intravenous dose of DENV-2 (16881 strain) and DENV-3 (isolated clinical strain D191267). Viral RNAs in mice sera were detected by qRT-PCR at 3 dpi. The mean viral RNA load in sera reached 10⁶ copies/mL and 10⁵ copies/mL in the groups of HPF-immunized mice infected with DENV-2 and DENV-3, respectively ( Figure 4B ). In contrast, for most mice vaccinated with tetravalent E13-HPF nanoparticle, the viral RNA load was below the limit of detection. In addition, the clinical score of each mouse was recorded every 2 days. If mice exhibited overwhelming signs of severe disease, they were euthanized. DENV-2-infected mice in the control group started to show lethargy, ruffled fur, and a hunched posture on day-3, and the clinical score of each mouse was ≥4 on day-5 ( Table S1 ). The mice in the vaccine group had a clinical score ≤3, and all the mice survived at the end of the experiment ( Figures 4C, D ). Compared with mice challenged with DENV-2, the signs of severe disease in DENV-3-infected mice appeared later. Two mice in HPF-immunized group displayed reduced mobility on day-13 and the number of survival mice declined on day-15. However, mice in the vaccine group were essentially healthy. Taken together, these results demonstrated that the E13-HPF vaccine provided AG6 mice with full protection against DENV-2 and DENV-3.

To determine the protective ability with different doses of the E13-HPF vaccine, single-dose, prime-boost, and third-dose immunization strategies were employed in BALB/c mice ( Figure 5A ). The immune interval was 3 weeks. For single-dose and prime-boost strategies, mouse sera were harvested to measure the titer of specific IgG and nAbs on day-42, whereas sera in the third-dose group were obtained 2 weeks after the final dose. All mice generated a high level of four serotypes of E13-specific IgG at approximate titers of 10⁴, but the prime-boost and third-dose strategies evoked a higher titer than the single-dose strategy ( Figure 5B ). Moreover, anti-DENV-2 nAb titers seemed to be increased in groups with a boost or booster dose, though none of these differences were significant ( Figure 5C ). Significant differences in IgG titers and nAb titers were not found between prime-boost and third-dose groups. ADE assays with BMDMs were also carried out to confirm if different immune strategies influenced ADE. qRT-PCR data revealed that the sera in vaccine groups did not increase the relative viral RNA load (DENV-2) in BMDMs, whereas sera from the DENV-2-infected group resulted in a 5-fold increase ( Figure 5D ). Studies have illustrated that immunocompetent mice (such as BALB/c mice) showed impairment of liver function when they underwent intravenous inoculation of a high dose of an adapted DENV-2 strain (39, 40). Therefore, immunized BALB/c mice were injected (i.v.) with 10⁷ FFU of the 16881 strain of DENV-2 on the day after blood collection. Histopathology of the liver in HPF-immunized mice revealed infiltration of inflammatory cells, signs of necrosis, and cytopathic effects (e.g., intracellular vacuole formation and particulate degeneration) at 7 dpi ( Figures 5E , S3 ). In contrast, H&E staining of the liver from the single-dose group showed normal hepatocytes even though there was slight widening of sinusoidal spaces. For prime-boost and third-dose groups, almost no impairment was seen in liver sections. Taken together, these results indicated that different dosing strategies of the E13-HPF vaccine protected mice from DENV-2 infection efficiently.

To investigate which adjuvants could facilitate E13-HPF vaccine immunogenicity, we compared immune responses in vaccinated BALB/c mice with several types of adjuvants: four TLR agonists (Pam2CSK4, Poly I: C, Vesatolimod, RS09), a cyclic dinucleotide (c-di-GMP), and alum ( Table S3 ). Among these adjuvants, RS09 and c-di-GMP could induce a high nAb titer ( Figure S4 ). We found that c-di-GMP could induce a higher titer of E13-specific IgG ( Figure 6A ). The nAb titer in the c-di-GMP-treated group reached 10⁴ in two of three samples, which was 10-times higher than that in the alum-treated group, whereas RS09 stimulated the same level of nAbs as alum ( Figure 6B ). To further evaluate ADE under different levels of antibodies, mouse sera from different groups were diluted at 1:50, 1:500, and 1:5000. The relative viral RNA load in BMDMs was enhanced in the DENV-2-infected group with an increase of sera dilution, whereas obvious enhancement was not seen in adjuvant groups ( Figure 6C ). At a sera dilution of 1:5000, the relative viral RNA load in the c-di-GMP group was 17.42-times lower than that in the DENV-2 infected group. At a sera dilution of 1:500, the relative viral RNA load in the c-di-GMP group was 11.14-times lower than that in the DENV-2-infected group, though a significant difference was not observed among groups. These data indicated that the c-di-GMP group showed a stronger ability to avoid ADE than the alum group.

A high IgG2a/IgG1 ratio has a close correlation with the Th1 cell-biased immune response, which enhances protection against virus infection (41). Hence, we examined whether c-di-GMP could induce a stronger Th1-biased immune response. ELISA revealed that RS09 and c-di-GMP stimulated higher E13-specific IgG2a titers than that stimulated by alum ( Figure 6D ). These results indicated that c-di-GMP evoked the highest level of IgG2a, though there were no significant differences in the level of total IgG between alum- and c-di-GMP-treated groups. Taken together, these data implied that c-di-GMP was a more promising adjuvant for the E13-HPF vaccine.