A Dengue Virus Serotype 1 mRNA-LNP Vaccine Elicits Protective Immune Responses

ABSTRACT

Dengue virus (DENV) is the most common vector-borne viral disease, with nearly 400 million worldwide infections each year concentrated in the tropical and subtropical regions of the world. Severe dengue complications are often associated with a secondary heterotypic infection of one of the four circulating serotypes. In this scenario, humoral immune responses targeting cross-reactive, poorly neutralizing epitopes can lead to increased infectivity of susceptible cells via antibody-dependent enhancement (ADE). In this way, antibodies produced in response to infection or vaccination are capable of contributing to enhanced disease in subsequent infections. Currently, there are no available therapeutics to combat DENV disease, and there is an urgent need for a safe and efficacious vaccine. Here, we developed a nucleotide-modified mRNA vaccine encoding the membrane and envelope structural proteins from DENV serotype 1 encapsulated in lipid nanoparticles (prM/E mRNA-LNP). Vaccination of mice elicited robust antiviral immune responses comparable to viral infection, with high levels of neutralizing antibody titers and antiviral CD4⁺ and CD8⁺ T cells. Immunocompromised AG129 mice vaccinated with the prM/E mRNA-LNP vaccine were protected from a lethal DENV challenge. Vaccination with either a wild-type vaccine or a vaccine with mutations in the immunodominant fusion loop epitope elicited equivalent humoral and cell-mediated immune responses. Neutralizing antibodies elicited by the vaccine were sufficient to protect against a lethal challenge. Both vaccine constructs demonstrated serotype-specific immunity with minimal serum cross-reactivity and reduced ADE in comparison to a live DENV1 viral infection.

IMPORTANCE With 400 million worldwide infections each year, dengue is the most common vector-borne viral disease. Forty percent of the world's population is at risk, with dengue experiencing consistent geographic spread over the years. With no therapeutics available and vaccines performing suboptimally, the need for an effective dengue vaccine is urgent. Here, we develop and characterize a novel mRNA vaccine encoding the dengue serotype 1 envelope and premembrane structural proteins that is delivered via a lipid nanoparticle. Our DENV1 prM/E mRNA-LNP vaccine induces neutralizing antibody and cellular immune responses in immunocompetent mice and protects an immunocompromised mouse from a lethal DENV challenge. Existing antibodies against dengue can enhance subsequent infections via antibody-dependent enhancement (ADE). Importantly our vaccine induced only serotype-specific immune responses and did not induce ADE.

INTRODUCTION

Dengue virus (DENV) is the most common vector-borne viral disease affecting humans (1–3). Its region of endemicity now includes 100 countries in Asia, the Pacific, the Americas, and the Middle East (3), with 40% of the world’s population at risk. Disease states during dengue infection manifest as a range of severities, from a self-limiting, febrile illness to more severe cases with life-threatening vascular leakage that can lead to multiorgan failure associated with a virus-driven cytokine storm (4, 5).

DENV is a member of the family Flaviviridae of which Zika virus, West Nile virus, yellow fever virus, and Japanese encephalitis virus are also members. It is spread by the arthropod vector Aedes aegypti and, to a much lesser extent, Aedes albopictus (2, 3). The virus contains a single-stranded, positive-sense RNA genome which codes for a single polypeptide containing three structural proteins, premembrane (prM), envelope (E), and capsid (C), as well as seven nonstructural proteins (6). Dengue virus is categorized into four distinct serotypes, dengue serotypes 1 to 4 (DENV1 to DENV4), with amino acid sequence variations of 30 to 35% across serotypes.

Most countries where dengue is endemic are affected by all four serotypes (1). Infection with a single serotype of DENV does not protect against a secondary infection of a heterologous serotype. Instead, primary infection increases an individual’s probability of developing severe clinical symptoms, including shock and death, upon a secondary heterotypic challenge. In this scenario, humoral immune responses after a primary infection produce cross-reactive, nonneutralizing antibodies. These antibodies can bind to infectious virus particles from a secondary, heterotypic challenge and lead to increased infection of cells possessing Fcγ receptors via antibody-dependent enhancement (ADE). This poses a challenge for vaccination, as a successful vaccine must elicit a neutralizing, long-lasting immune response balanced equally against all four serotypes of DENV.

DENV vaccines that have progressed the furthest in clinical evaluation include CYD-TDV (Dengvaxia; Sanofi-Pasteur), TAK-003 (DENVax; Takeda), and TV003 (NIAID/NIH) (7–11). All three of these vaccines are tetravalent, live attenuated vaccines that encode the membrane-embedded DENV viral proteins prM and E, in different viral backbones. Other vaccine strategies are in various preclinical stages, including recombinant E and subunit vaccines (12–15), purified inactive viruses (16), DNA encoding prM and E (17, 18), and purified virus-like particles (VLPs) (19–21). VLPs, like an infectious viral particle, are comprised of ENV dimers on the surface, resulting in production of particles that have many of the same three-dimensional epitopes as an infectious virus particle (21, 22).

Previously, we developed an mRNA vaccine against the related Zika virus encoding the viral prM and E proteins (23, 24). This vaccine elicited a robust neutralizing antibody response that protected mice from a lethal Zika viral challenge and prevented vertical transmission of the virus to the fetus. mRNA vaccines have also been shown to provide protective immunity against viral pathogens in nonhuman primates (25, 26). In this study, we have developed an mRNA vaccine against DENV serotype 1. A construct coding for prM and E proteins was in vitro transcribed using the modified nucleotide pseudouridine, and the resulting mRNA was packaged into lipid nanoparticles (LNPs). Following intramuscular injection, mRNA-LNPs are taken up into the muscle cells at the site of injection, as well as antigen-presenting cells in the draining lymph node (27, 28). Once the cells endocytose the mRNA-LNP, the LNP degrades in the acidified endosome, releasing the mRNA into the cytoplasm. The mRNA is then translated into the viral prM/E proteins. The prM/E polyprotein is embedded in the membrane of the endoplasmic reticulum (ER) and cleaved by host protease into the individual viral proteins. The prM and E self-assemble into VLPs on the surface of the ER membrane, and then the VLPs are trafficked through the trans-Golgi network and secreted from the cell. Administration of the DENV1 prM/E mRNA-LNP vaccine elicited neutralizing antibody titers and antivirus-specific T cells in wild-type (WT) C57BL/6J mice and conferred protection in DENV permissive immunocompromised AG129 mice. Importantly the mRNA-LNP vaccine induced serotype-specific immunity with low levels of ADE.

RESULTS

Design of DENV1 prM/E construct and viral protein expression.

We designed a construct comprising the wild-type nucleotide sequence encoding prM and E proteins from dengue serotype 1 (DENV1) strain 16007 downstream of a Japanese encephalitis virus (JEV) signal peptide. The coding sequence was flanked by a 5′ untranslated region (UTR) previously utilized in other mRNA vaccines (23) and the 3′ UTR from the human hemoglobin subunit alpha 1 mRNA (HBA1) (Fig. 1A). The 5′ and 3′ UTRs contribute to translation regulation and mRNA stability essential for optimum protein expression. We in vitro transcribed mRNA from a T7 RNA polymerase promoter site upstream of the 5′ UTR. A 5′ cap-1 structure and a 3′ poly(A) tail were enzymatically added to produce fully mature mRNA that resembles host mRNA. We also generated a separate construct (ΔFL) containing the amino acid substitutions G106R, L107D, and F108A to remove the fusion loop epitope of the envelope protein. These mutations have been previously characterized and shown to ablate both fusion loop activity and production of fusion loop-specific antibodies responsible for ADE (29–31).

FIG 1.

DENV prM/E vaccine design and viral protein expression. (A) Schematic of the DENV genome and engineered mRNA construct. An mRNA encoding the prM and ENV viral proteins was engineered with N-terminal signal peptide sequence, 5′ and 3′ untranslated regions (UTR) flanking the coding sequence, a 3′ poly(A) tail, and a 5′ cap-1 structure. In vitro-synthesized mRNA is encapsulated in a lipid nanoparticle for use in in vitro and in vivo experiments. (B) 293T cells were transfected with the in vitro-transcribed mRNA encoding the wild-type sequence (WT) or a mutant version with amino acid substitutions in the fusion loop epitope (ΔFL). Lysate was analyzed by Western blotting with the domain III-specific 1A1D-2 monoclonal antibody and the fusion loop-specific 4G2 monoclonal antibody. (C) Supernatant from transfected cells was purified and concentrated through ultracentrifugation and analyzed for VLPs by Western blotting with the 1A1D-2 monoclonal antibody or anti-GAPDH. Unpurified cell lysate from WT mRNA-transfected cells is included as a control. Shown are representative blots. (D) Electron microscopy image of VLPs from purified supernatant of transfected 293T cells showing homogenous shape and size of approximately 30 nm.

In vitro-synthesized mRNA was transfected into 293T cells, followed by collection of cell lysate and supernatant. We performed immunoblot analysis with the monoclonal antibodies (MAb) 1A1D-2 and 4G2. 1A1D-2 is specific for domain III of the E protein (32–34), and 4G2 binds to the fusion loop epitope. Western blotting with the monoclonal antibody 1A1D-2 identified a band representing DENV1 E after transfection with both WT and ΔFL constructs (Fig. 1B), demonstrating successful viral protein expression. Western blotting with 4G2 resulted in a band only in the lysate from wild-type-transfected cells, thus revealing successful ablation of the fusion loop epitope in the ΔFL construct (Fig. 1B). Expression of prM and E alone is sufficient to induce the formation and secretion of VLPs (23, 35, 36). To detect secreted VLPs, we purified the supernatant from the transfected cells via ultracentrifugation and analyzed it on immunoblots. We detected E protein bands with the 1A1D-2 antibody in the purified supernatant of WT and ΔFL construct-transfected supernatants (Fig. 1C), demonstrating that fusion loop ablation did not affect secretion of VLPs from transfected cells. We did not detect any GAPDH (glyceraldehyde-3-phosphate dehydrogenase) in the purified supernatants, verifying that ultracentrifugation removed any cytoplasmic contamination. Particles secreted from transfected cells had properties similar to those of VLPs, with relatively uniform semismooth surfaces and diameters of approximately 30 nm, as confirmed by electron microscopy (Fig. 1D). Together, these results show that in vitro-synthesized mRNAs can induce viral structural protein expression and secretion of VLPs. Further, mutation of key amino acids within the fusion loop successfully ablates the antigenic epitope while maintaining protein expression and VLP excretion.

Optimization of protein expression and LNP delivery.

Signal peptides are short N-terminal peptides that traffic proteins through the appropriate processing and secretory pathways within the trans-Golgi network. We compared different signal peptides on the DENV1 ΔFL construct to optimize protein expression. We generated five new ΔFL mRNA constructs, with the original JEV signal peptide exchanged for signal peptides from either interleukin-2 (IL-2), tissue plasminogen activator (tPA), or Gaussia luciferase (GLUC). Additionally, we synthesized two constructs with theoretical signal peptides computationally predicted to elicit robust protein secretion in skeletal muscle cells (SP1 and SP2) (37). Mice are administered the mRNA-LNP vaccine intramuscularly, so characterization and optimization of protein expression in muscle cells is key. We transfected differentiated skeletal muscle myoblast C2C12 cells with the different constructs and blotted for E protein expression with the 1A1D-2 antibody. The tPA signal peptide resulted in the most robust E protein expression (Fig. 2A). To ensure that signal peptide modification did not alter VLP secretion and directly compare VLP secretion across the different mRNA constructs, we also analyzed the supernatant of transfected cells via dot blotting with the 1A1D-2 MAb. The tPA signal peptide also resulted in the highest levels of VLP secretion (Fig. 2B).

FIG 2.

Optimization of signal peptide and LNP delivery. (A) Constructs were engineered with alternative signal peptides, and in vitro-transcribed mRNA was transfected into differentiated murine muscle myoblast C2C12 cells. Cell lysate was analyzed by Western blotting with 1A1D-2 monoclonal antibody or anti-GAPDH antibodies. (B) Supernatant of transfected C2C12 cells was analyzed by dot blotting with 1A1D-2. (C) In vitro-synthesized WT or ΔFL mRNA was encapsulated in a lipid nanoparticle and administered to C2C12 cells. Lysate was analyzed by Western blotting with 1A1D-2 antibody. Shown are representative blots.

For in vivo administration, mRNA is synthesized with the modified nucleotide, pseudouridine, in place of uridine. This replacement dampens innate immune stimulation and interferon activation which inhibits protein translation (38). In vitro-synthesized mRNA is further purified and encapsulated in a lipid nanoparticle (LNP). Encapsulation within an LNP shields the mRNA from extracellular RNases and ensures efficient delivery into cells (39). LNPs are composed of pH-sensitive lipids that bind to endogenous apolipoprotein E, which facilitates entry. When the mRNA-LNP is endocytosed, the acidic environment of the late endosome initiates degradation of the LNP, leading to release of the mRNA to the cytoplasm. We encapsulated mRNA containing the original JEV signal peptide that has been utilized in previous flavivirus mRNA vaccines. We achieved >90% encapsulation efficiency, as determined by a Ribogreen RNA quantification assay, and stored encapsulated mRNA at 4°C for extended periods of time to accommodate a two- or three-shot vaccination schedule. Delivery of nucleotide-modified WT and ΔFL prM/E mRNA-LNPs to C2C12 cells resulted in E protein expression in cell lysate (Fig. 2C).

DENV1 prM/E mRNA vaccines elicit adaptive immune responses.

Initially, wild-type C57BL/6 mice were vaccinated according to a three-shot vaccination schedule with 10 μg of mRNA per dose and serum collections at day 0 (prevaccination), day 28 (post primary), day 42 (post secondary), and day 56 (post tertiary), as shown in Fig. 3A. We quantified neutralizing antiviral antibody titers in serial dilutions of serum with a focus reduction neutralization test (FRNT) against the homologous DENV1 strain 16007. All mice within each cohort of WT and ΔFL vaccine groups seroconverted, with EC₅₀ neutralizing titers (serum concentration at which 50% of the virus is neutralized) reaching a maximum of ∼1/200. WT and ΔFL prM/E mRNA-LNP vaccines elicited neutralizing antibody responses after a single dose, with secondary and tertiary doses boosting titers (Fig. 3B). A third vaccine dose did not significantly enhance the neutralizing antibody titers from that of a second dose (P value = 0.20; WT). As such, a two-dose, prime-boost vaccination schedule was used in future studies. These data reveal that in vivo delivery of an mRNA-LNP vaccine induces a humoral immune response against the exogenous viral protein.

FIG 3.

DENV1 prM/E mRNA vaccines induce neutralizing antibody responses. DENV1 prM/E mRNA-LNP vaccines were administered to 10-week-old C57BL/6 mice. (A) Mice were administered 10 μg of mRNA vaccine in a three-shot schedule, and serum was collected at the indicated time points. (B) Serial dilutions of serum from vaccinated mice were analyzed for neutralization activity by an FRNT against DENV1 strain 16007. Neutralization curves at each time point are shown for WT vaccine recipients (left) and ΔFL vaccine recipients (right). The average values ± standard errors of the mean (SEM) of results for five vaccinated mice are shown. (C) Mice were administered a high (10 μg) or low (3 μg) dose of the mRNA vaccines or vaccine encoding GFP. A separate group of mice were infected with wild-type DENV1 by following the same schedule. (D) Antiviral IgG titers were determined by ELISAs, and the endpoint dilution titer was calculated. (E) Serum was analyzed by FRNTs, and the normalized percentage of infection of each group is plotted as the mean ± SEM for each serum dilution. n = 5 mice per group of mice infected with virus or receiving 3-μg vaccine doses. n = 10 mice per group in mice receiving 10-μg doses of the ΔFL and GFP vaccines. n = 15 for mice receiving 10-μg doses of the WT vaccine. (F) EC₅₀ values of the neutralization curves for individual mice are shown. The statistical significance of results for each group in comparison to that for the GFP control was determined via unpaired t test. **, P < 0.01; ***, P < 0.001. Statistical comparisons with P values of >0.05 are not shown in this figure.

A separate cohort of mice were administered high or low doses (10 μg or 3 μg per injection) of WT or ΔFL prM/E mRNA-LNP vaccine in a prime-boost schedule (Fig. 3C). We also included mice infected with live DENV1 virus (10⁵ focus-forming units [FFU] DENV1) as a positive control and an mRNA vaccine encoding green fluorescent protein (GFP) (10 μg) as a negative control. We quantified the levels of antiviral IgG in the sera isolated from the different vaccine groups via an ELISA against purified DENV1 strain 16007. All mice receiving the infectious DENV1, WT mRNA vaccine, or ΔFL mRNA vaccine had significantly higher titers than mice receiving the GFP mRNA control vaccine (Fig. 3D). No statistical differences were observed between results for the WT and ΔFL vaccines. Virus-infected mice and mice receiving the high dose of the WT or ΔFL vaccines all had antibody endpoint dilution titers of approximately 1 × 10⁵. The 3-μg low dose of the vaccine induced antibody titers slightly lower than the titers induced by the higher dose of 10 μg. Serum neutralization titers were determined via FRNTs (Fig. 3E and F) against infectious DENV1 strain 16007. High and low doses of the WT prM/E mRNA vaccine elicited EC₅₀ values of 1/420 and 1/263, respectively, revealing little to no dose-dependent response (Fig. 3F) (P value = 0.36). Additionally, high and low doses of the ΔFL vaccine resulted in similar EC₅₀ values of 1/329 and 1/175, respectively. These differences were not statistically significant (Fig. 3F) (P value = 0.29). The mice vaccinated with the WT and ΔFL vaccines had lower neutralizing titers than the DENV1 virus-infected mice (EC₅₀ = 1/729), although these differences were not statistically significant. All vaccines or infections resulted in higher neutralizing values than those of the GFP-vaccinated mice (Fig. 3E) (P value < 0.001). Neutralizing titers of WT and ΔFL construct-vaccinated mice were very similar, indicating that the fusion loop mutation did not alter humoral immune responses.

To quantify antiviral T cells, splenocytes were harvested from vaccinated mice at day 56 after a tertiary vaccination schedule and stimulated with a pooled 15mer overlapping peptide array for the ENV protein from DENV1 or DENV2 as well as the NS1 protein from DENV1. Stimulated cells were analyzed for intracellular gamma interferon (IFN-γ) by flow cytometry, and antiviral IFN-γ⁺ T cells were quantified. prM/E mRNA vaccines elicited modest yet significant antiviral CD4⁺ and CD8⁺ T cell responses specific for the DENV1 ENV protein, with equivalent levels between the WT and ΔFL vaccines (Fig. 4). No T cell responses were detected against the homologous DENV2 E protein or the irrelevant DENV1 NS1 protein. Thus, prM/E mRNA-LNP vaccines elicit both humoral and cellular immune responses against DENV1 E protein.

FIG 4.

DENV1 prM/E mRNA vaccines induce antiviral CD8⁺ and CD4⁺ T cells. DENV1 prM/E mRNA-LNP vaccines were administered to 10-week-old C57BL/6 mice in a three-shot vaccination schedule. Spleens were harvested after the final vaccination dose (day 56) and stimulated with an overlapping peptide array of DENV1 E protein, DENV2 E protein, or DENV1 NS1 protein. Stimulated cells were stained for the intracellular cytokine IFN-γ and analyzed by flow cytometry. Plotted are the IFN-γ⁺ T cells as a percentage of total CD8⁺ T cells (A) or CD4⁺ T cells (B). n = 5 mice in the WT and ΔFL construct-vaccinated groups. (C) Representative flow cytometry plots and gates from a single mouse.

DENV1 prM/E mRNA vaccines protect against a lethal challenge.

AG129 mice lack the type I interferon α/β receptor and the type II interferon γ receptor, and they are permissive to a lethal DENV challenge (40–42). All serotypes of DENV are capable of replication in AG129 mice with quantifiable viremia, vascular leakage, and increased cytokine levels. Some strains can induce more severe disease states, indicative of severe disease, in humans, such as DENV2 D2S20 or DENV1 Western Pacific (40, 42). AG129 mice were vaccinated according to the previously described schedule (Fig. 3C) with GFP mRNA-LNP or DENV1 wild-type prM/E mRNA-LNP. Serum was collected from vaccinated mice and analyzed for neutralization titers as previously described. DENV1 prM/E mRNA-LNP vaccination induced EC₅₀ values of greater than 1/3,000 (Fig. 5A and B). Vaccinated AG129 mice were challenged with 10⁶ FFU DENV1 Western Pacific strain and monitored for 40 days postinfection. Mice receiving the GFP mRNA-LNP vaccine lost weight beginning at day 6, and all mice succumbed to viral infection by day 32 postinfection. DENV1 prM/E mRNA-LNP-vaccinated mice did not show any signs of morbidity or mortality, with weight remaining stable postinfection and 100% of the mice surviving (Fig. 5C and D). These data demonstrate that a DENV1 prM/E mRNA-LNP vaccine protects against a lethal DENV1 challenge in an immunocompromised mouse model.

FIG 5.

DENV1 prM/E mRNA vaccines protect against a lethal challenge. Ten micrograms of DENV1 prM/E or GFP mRNA-LNP vaccines was administered to AG129 mice in a prime-boost schedule 4 weeks apart. n = 5 mice per group. (A) Serum from vaccinated mice was isolated 2 weeks after the boost and analyzed for neutralization by FRNT of serially diluted serum samples. Plotted are the means ± SEM of results from five mice for each dilution. (B) EC₅₀ values for each mouse are plotted. The vaccinated mice were then challenged with a lethal dose of DENV1 strain Western Pacific. Mice were monitored for weight (C) and survival (D) postchallenge. **, P < 0.01; ***, P < 0.00.

The DENV1 prM/E mRNA vaccine elicited both antiviral antibodies and an antiviral T cell response. We hypothesized that the antiviral antibodies are sufficient to protect against a lethal challenge. To address this hypothesis, we adoptively transferred pooled serum from WT construct-vaccinated mice into AG129 mice. As controls, a second group of mice received pooled serum from naive mice and a third group of mice received phosphate-buffered saline (PBS). One day after adoptive transfer, mice were challenged with 10⁶ FFU DENV1 Western Pacific strain and monitored for 40 days postinfection (Fig. 6). Seven of 8 mice that received serum from the vaccinated mice were protected against lethality. Six of seven mice that received naive serum lost weight and succumbed to viral lethality postchallenge. Thus, antibodies elicited by the DENV1 prM/E mRNA-LNP vaccine are sufficient for protection.

FIG 6.

Passive transfer of immune sera protects against a lethal challenge. Serum from naive or WT prM/E mRNA-vaccinated mice was passively transferred into AG129 mice. One day after transfer, mice were challenged with a lethal dose of DENV1 strain Western Pacific. Mice were monitored for weight (A) and survival (B) postchallenge. Survival curves comparing vaccinated and naive serum recipients were analyzed by log rank test. **, P < 0.01.

DENV1 prM/E mRNA vaccination induces serotype-specific humoral immunity.

Infection with DENV will lead to antibodies that cross-react with heterotypic DENV serotypes with the potential to cause ADE. We characterized the cross-reactive immune response in serum of the prM/E mRNA-vaccinated mice. We quantified ADE by incubating DENV2 with serial dilutions of serum from vaccinated mice before infecting Fcγ receptor-positive K562 cells. Infection was determined via flow cytometry with the monoclonal antibody 1A1D-2 against the viral E protein. The percentage of infected cells was compared to that of a DENV2 infection in the absence of immune sera (Fig. 7A). Serum from DENV1 virus-infected mice significantly enhanced DENV2 infections, even with dilutions as high as 1/6,000. At a serum dilution of 1/100, an 8-fold enhancement was observed. Conversely, serum from mRNA-vaccinated mice induced very low levels of DENV2 enhancement (Fig. 7A), with only a 1.2-fold enhancement at a 1/100 serum dilution (Fig. 7B). The amino acid sequence of the WT prM/E mRNA-LNP vaccine was identical to the sequence of the infecting virus. Surprisingly, WT and ΔFL mRNA vaccines enhanced heterotypic DENV2 to nearly identical values. Similar results were seen with DENV4 (data not shown). As a negative control, serum from naive mice showed no enhancement at any dilution. To assess the role of ADE on viral replication and egress, we quantified the levels of infectious virus in the supernatant of K562 cells infected with immunocomplexed virus. Serum from WT or ΔFL mRNA-vaccinated mice enhanced viral replication relative to serum from GFP-vaccinated mice; however, enhancement was significantly lower than that of serum from virus-infected mice, in agreement with the flow cytometry data (Fig. 7C). Importantly, DENV1 vaccines did not elicit neutralizing antibodies against DENV2 (strain New Guinea C) in an FRNT (Fig. 7D). These data demonstrate that DENV1 prM/E mRNA vaccines do not induce cross-reactive antibodies which elicit heterotypic enhancement, in contrast to a viral infection.

FIG 7.

DENV1 prM/E mRNA vaccination results in reduced ADE levels. Serum from naive mice, WT prM/E mRNA-vaccinated mice, ΔFL prM/E mRNA-vaccinated mice, or mice infected with DENV1 2 weeks after boost were analyzed for enhancement of DENV2 infection. Serial dilutions of serum were incubated with DENV2 and added to Fcγ receptor-positive K562 cells. Fifteen hours later, infected cells were stained for intracellular ENV and quantified by flow cytometry. The percentage of infected cells was normalized to that of infection in the absence of serum. (A) The fold change in percentage of cells infected is shown compared to that of infections in the absence of serum. The average fold enhancement ± SEM for five mice per group is graphed. (B) A representative flow cytometry histogram of the ENV signal for each different treatment at a 1/100 serum dilution is shown. (C) A 1/100 dilution of serum was incubated with DENV2 for 1 h and added to K562 cells at an MOI of 1. At 48 h later, viral titers in the supernatant were quantified via a focus-forming assay. Viral titers were normalized to virus-infected serum within each independent experiment, and the results of four independent experiments are shown. *, P < 0.05; **, P < 0.01. (D) Serial dilutions of serum from vaccinated mice were analyzed for neutralization activity against DENV2 (strain New Guinea C) with an FRNT. The average values ± SEM of results for five vaccinated mice are shown.

DISCUSSION

Despite a longstanding effort in the field, there still remains an unmet need for a DENV vaccine that elicits robust, balanced immune response against all four serotypes. Here, we developed a vaccine against DENV1 with a modified mRNA encoding the prM and ENV viral proteins encapsulated in a lipid nanoparticle (LNP). The mRNA-LNP vaccine platform has now been developed for several viruses, including rabies virus (26), influenza virus (26), and HIV (43). More recently, mRNA vaccines have been rapidly developed against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Moderna’s mRNA-1273 and BioNTech’s mRNA BNT162b1 were the first vaccine candidates to show safety and efficacy in human trials, demonstrating the speed of the mRNA platform and its role in emerging infectious diseases (44–46). In the flavivirus field, mRNA vaccines have been developed against Zika virus (23, 24) and Powassan virus (47). These flavivirus vaccines encode the viral structural proteins which are expressed and lead to the development of neutralizing antibodies against the viral structural proteins. Recently, another group published the results of a mRNA vaccine against DENV2 (48). Zhang et al. developed mRNA vaccines encoding full-length prM-ENV, the soluble portion of ENV, and NS1. Vaccination with the mRNA encoding the soluble portion of DENV2 ENV (E80) elicited humoral and cell-mediated immune responses that protected against a lethal challenge with a homologous serotype of DENV2, similar to our findings with a DENV1 serotype mRNA vaccine. However, the DENV2 E80 mRNA vaccine induced serotype cross-reactive immune responses and high levels of heterologous ADE (48). The significant reduction of DENV2 ADE seen with the DENV1 mRNA vaccine we outline here shows promise for future efforts that aim to produce a safer dengue vaccine that offers broad protection while successfully avoiding ADE, a major hurdle in the development of protective flavivirus vaccines.

A high level of antigen expression is key for the success of mRNA vaccines. The signal peptide plays a critical role in directing the translated protein into the appropriate locations for processing and secretion. Previous flaviviral mRNA vaccines have included an N-terminal JEV or IgE signal peptide (23, 24, 48). In our study, the tPA signal peptide led to far greater ENV expression and VLP secretion in C2C12 cells than other signal peptides, including the JEV signal peptide. All in vivo studies here were performed with the original vaccine construct encoding the JEV signal peptide, but we predict that future vaccine formulations with the tPA signal peptide will lead to greater antigen expression and higher antiviral antibody titers.

The DENV1 mRNA-LNP vaccine elicited humoral and cell-mediated immunity following a two-dose vaccination regimen, with antibody titers of 1/120,000 and neutralizing titers of 1/420 (WT, 10 μg). The antiviral antibodies were sufficient for protection. The neutralizing antibody titers reported here are similar to those of other DENV1 vaccination strategies. Neutralization EC₅₀ titers of approximately 1/100 were achieved by administration of a DNA vaccine encoding the modified viral structural proteins (29). DENV1 purified-VLP vaccine generated with fusion loop mutants resulted in neutralization EC₅₀ titers of approximately 1/1,000 (31). In phase III human clinical trials, the CYD-TDV (Dengvaxia) vaccine elicited EC₅₀ neutralization titers of approximately 1/60 in seronegative individuals (49), and TAK-003 neutralization titers reached 1/184 (50). In phase I human studies, TV003 elicited EC₅₀ neutralization titers of 1/63 against DENV1 (51).

The neutralizing antibody titers in the vaccinated AG129 mice (EC₅₀ of 1/3,125) were significantly higher than those in the C57BL/6 mice (EC₅₀ of 1/420) following equivalent vaccination schedules (P value < 0.001). Likely, the lower neutralization titers in C57BL/6 mice are due to decreased antigen expression in the presence of an intact type I interferon response. Indeed, previous studies have demonstrated that mRNA vaccines engage RNA-sensing pattern recognition receptors and activate the type I IFN pathway, leading to eIF2α phosphorylation and blunted translation of the exogenous transcript (39, 52). Increasing vaccine efficacy could be achieved through lowering the RNA-sensing and IFN response. We have included the pseudouridine modification in our DENV1 prM/E mRNA-LNP vaccine, but further modifications such as 5-methylcytosine could further lower innate immune stimulation and increase antigen expression and associated antibody titers (39).

In humans, CD4⁺ and CD8⁺ T cells predominantly target capsid and NS3, respectively, following a DENV infection (53). Although our vaccine does not encode these immunodominant T cell epitopes, we detected antiviral CD4⁺ and CD8⁺ T cell responses against the E protein in the vaccinated mice. Intriguingly, the CD4⁺ and CD8⁺ T cell responses were not cross-reactive with other DENV serotypes, likely due to the high variability across the different DENV serotypes in the E protein. The overall magnitude of the T cell response from our vaccine was lower than that of a recently described mRNA vaccine against DENV1 which encoded the immunodominant HLA epitopes from the nonstructural proteins of DENV (54). Our vaccine was designed to elicit antibodies against the structural proteins to neutralize infectious virus particles as opposed to robust T cell responses. Although we cannot rule out a role for antiviral T cells in a vaccinated host, neutralizing antibodies in serum were sufficient to protect against a lethal homotypic challenge in a passive transfer model. Together, these studies demonstrate that mRNA vaccines can be developed to induce both protective T and antibody-dependent immunity against DENV.

Interestingly, the prM/E mRNA vaccines elicited serotype-specific antibody responses. Here, we demonstrate vaccine protection against two DENV1 strains, but further studies will be required to demonstrate broad serotype specificity. Serum from DENV1 virus-infected mice enhanced a DENV2 in vitro infection, whereas heterotypic ADE was largely absent with serum from the mRNA-vaccinated mice. These observations are surprising, given that neutralization titers were similar between the virus-infected and vaccinated mice and the identical amino acid sequences shared between the WT mRNA vaccine and the infecting DENV1 16007 strain. Further, this suggests that the polyclonal antibody repertoire induced by the mRNA vaccine is inherently different than the polyclonal repertoire induced during a viral infection. In our previous study, mutation of the fusion loop epitope in the Zika virus mRNA vaccine led to ablation of cross-reactive DENV enhancement through ADE (23). Similarly, in previous studies with VLP- and DNA-based vaccines, mutation of the fusion loop epitope lowered the prevalence of ADE (29, 55). Unexpectedly, mutation of the fusion loop epitope in the mRNA vaccine did not alter ADE. These findings suggest that antibodies against the fusion loop epitope are not dominant in the polyclonal response to our mRNA vaccine. Future efforts will focus on identification of the structural epitopes within the VLP secreted from a viral infection and an mRNA-LNP vaccine.

In this study, we have demonstrated that an mRNA vaccine encoding the prM and E proteins from DENV1 can elicit robust adaptive immune responses and protect against a lethal viral challenge. This study paves the way for future development of mRNA vaccines against the remaining DENV serotypes, with the ultimate goal of developing a tetravalent vaccine that will elicit a balanced, protective immune response against all four DENV serotypes. Current leading vaccination efforts rely on live attenuated virus, yet these vaccines fall short in either their ability to induce a broadly neutralizing antibody response or their ability to avoid ADE. In contrast to live attenuated vaccines in which differential replication of the attenuated viruses dictates antigen dosing in vivo, the antigen dose can be carefully modulated with mRNA vaccines to elicit a balanced immune response. Additionally, mRNA vaccines will allow the modification of epitopes which elicit ADE yet are impossible to incorporate into a live attenuated vaccine due to their critical role in viral replication.

MATERIALS AND METHODS

Viruses and cells.

DENV serotype 1 strain 16007 and DENV serotype 2 strain New Guinea C were provided by Michael Diamond at Washington University in St. Louis. DENV serotype 4 strain UIS 497 was obtained through BEI Resources (NR-49724), NIAID, NIH, as part of the WRCEVA program. Viral stocks were propagated in C6/36 cells, and titers were determined by a focus-forming assay (FFA). All propagated viral stocks were deep sequenced to confirm the viral strain. FFAs were performed to titer viral stocks with monoclonal antibody clone 9.F.10 obtained through Santa Cruz Biotechnology (catalog no. SC-70959). Experiments with DENV were conducted under biosafety level 2 (BSL2) containment at the University of Illinois College of Medicine or St. Louis University College of Medicine with institutional Biosafety Committee approval. Vero-E6 cells (catalog no. CRL 1586) and K562 cells (catalog no. CCL-243) were obtained from American Type Culture Collection (ATCC) and maintained for low passage number in accordance with ATCC guidelines. C6/36 cells were provided by the Diamond lab at Washington University in St. Louis, C2C12 cells were obtained from Ahke Heydemann, University of Illinois at Chicago (UIC), and 293T cells were obtained from Donna MacDuff at the University of Illinois at Chicago.

Generation of mRNA and mRNA-LNP.

Wild-type constructs encoding dengue serotype 1 strain 16007 prM and Env viral proteins were synthesized by Integrated DNA Technologies (IDT). Constructs contained a T7 promoter site for in vitro transcription of mRNA, 5′ and 3′ UTRs, and a Japanese encephalitis virus signal peptide. The sequences of the 5′ and 3′ UTRs were identical to those of previous publications with a Zika virus (ZIKV) mRNA vaccine (23, 24). mRNA was synthesized from linearized DNA with T7 in vitro transcription kits from CellScript and in accordance with the manufacturer’s protocol. Standard mRNA was produced with unmodified nucleotides (catalog no. C-MSC11610). RNA to be encapsulated in lipid nanoparticles was generated with pseudouridine in place of uridine with the Incognito mRNA synthesis kit (catalog no. C-ICTY110510). The 5′ cap-1 structure and 3′ poly(A) tail were enzymatically added. mRNA was encapsulated in lipid nanoparticles using the PNI Nanosystems NanoAssemblr benchtop system. mRNA was dissolved in PNI formulation buffer (catalog no. NWW0043) and run through a laminar flow cartridge with GenVoy ILM (catalog no. NWW0041) encapsulation lipids at a flow ratio of 3:1 (RNA in PNI buffer; Genvoy ILM) at a total flow rate of 12 ml/min to produce mRNA-LNPs. These mRNA-LNPs were characterized for encapsulation efficiency and mRNA concentration via RiboGreen assay using Invitrogen's Quant-iT Ribogreen RNA assay kit (catalog no. R11490).

Mouse experiments.

C57BL/6J mice were purchased from Jackson Laboratory and housed in the pathogen-free Biomedical Resources Laboratory at the University of Illinois College of Medicine. AG129 mice were bred in the animal facilities at St. Louis University. For vaccinations, mice were injected intramuscularly in the thigh with 50 μl of the indicated amount and type of mRNA-LNP suspended in PBS. Vaccinated C57BL/6J mice were challenged with 1 × 10⁵ FFU of DENV1 strain 16007, retro-orbitally. Vaccinated AG129 mice were challenged with 10⁶ FFU of DENV1 strain West Pac, intravenously (i.v.). For serum adoptive transfer studies, sera from vaccinated or naive mice were pooled and then 200 μl was administered i.v. into naive AG129 mice 1 day prior to challenge with DENV1. The vaccination and viral challenge protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Illinois College of Medicine (protocol no. 18-114) and St. Louis University (assurance no. D16-00141).

In vitro transfections.

293T and C2C12 cells were transfected with mRNA using the Mirus TransIT RNA transfection kit (catalog no. MIR 2225) according to the manufacturer’s protocol. 293T cells were 60 to 70% confluent at the time of transfection, with C2C12 cells being 100% confluent at the time of transfection to achieve differentiation into muscle tissue. Supernatant was collected 24 h posttransfection. To collect lysate, cells were washed with PBS and lysed with radioimmunoprecipitation (RIPA) buffer (Millipore-Sigma; catalog no. R0278). Lysate and supernatant were centrifuged at 16,000 × g for 10 min at 4°C to remove cell debris. Supernatant from transfected cells was purified using a 20% sucrose cushion and ultracentrifugation at 141,000 × g overnight (16 h) at 4°C. Purified protein complexes were resuspended in 50 μl of 1% bovine serum albumin (BSA) in PBS for subsequent storage and analysis.

Viral protein detection.

For Western blot analysis, 10 μl of lysate or purified supernatant samples was run on a 4-to-12% Bis-Tris SDS-PAGE gel (Invitrogen; catalog no. NW04120BOX) with subsequent transfer to a 0.45-μm-pore-size polyvinylidene difluoride (PVDF) membrane. Membranes were blocked in TBST (10 nM Tris-HCl, pH 7.5, 150 nM NaCl, and 1% Tween 20) buffer with 5% skim milk. Membranes were blotted with envelope domain III-specific 1A1D-2 (1:600) monoclonal antibody (CDC Arbovirus Reference Collection) or envelope fusion loop-specific 4G2 (3.33 mg/ml) (BEI catalog no. NR-50327, Novus Biologicals catalog no. NBP2-52709FR). Secondary antibody goat anti-mouse–horseradish peroxidase (HRP) (200 ng/ml) (Invitrogen catalog no. A16072) in blocking buffer allowed for detection of dengue viral envelope proteins. Western blots were imaged on a ChemiDoc Image Lab system (Bio-Rad).

For dot blot analysis, clarified transfection supernatant was diluted 1/4 in a 20-μl volume of transfer buffer (Life Technologies catalog no. NP0006-1) and applied dropwise to a presoaked 0.45-μm-pore-size PVDF membrane. Sample was allowed to infiltrate the membrane through capillary action for no more than 1 h (before the blot starts to dry). Blots were stained and imaged in the same manner as described for Western blots above.

Electron microscopy.

One T-75 cell culture flask was seeded with 293T cells at 70 to 80% confluence the day of transfection. The flask was transfected with 20 μg of mRNA encoding WT DENV1 prM and envelope protein using the Mirus TransIT RNA transfection kit (catalog no. MIR 2225) according to the manufacturer’s protocol. Supernatant was collected 48 h posttransfection. Supernatant was centrifuged at 16,000 × g for 10 min at 4°C to remove cell debris. Six milliliters of supernatant was then dialyzed overnight at 4°C in 20,000-molecular-weight-cutoff (MWCO) Slide-A-Lyzer dialysis cassettes (catalog no. 66003) submerged and spinning in PBS. Dialyzed sample was provided to the University of Illinois at Chicago electron microscopy core for imaging by using the following parameters. Ten to 15 μl of sample was loaded dropwise onto a 300-mesh, Formvar/carbon-coated copper electron microscopy (EM) grid with the excess removed by filter paper via capillary action. One drop of 2% uranyl acetate solution was deposited onto the EM grid with the excess removed by filter paper via capillary action. Once the grid was allowed to dry further, the sample was examined via transmission electron microscopy using a JEOL JEM-1400F transmission electron microscope, operating at 80 kV. Digital micrographs were acquired using an AMT BioSprint 12M-B charge-coupled device (CCD) camera and AMT software (version 701).

ELISA.

Four T-150 cell culture flasks of C6/36 cells were infected with WT DENV1 at a multiplicity of infection (MOI) of 0.1. Seven days after infection, 60 ml of supernatant was collected and clarified via centrifugation at 3,200 × g for 10 min at 4°C. Supernatant was further purified via 20% sucrose cushion ultracentrifugation at 141,000 × g for 2 h at 4°C to pellet virus. Virus pellets were resuspended in PBS for a total volume of 5 ml. ELISA plates were coated overnight at 4°C with 50 μl/well of a 1:25 dilution of concentrated viral stock (1E3 FFU/well) in coating buffer (0.1 M sodium carbonate, 0.1 sodium bicarbonate, 0.02% sodium azide, at pH 9.6). After the plates were coated overnight, they were incubated with blocking buffer (PBS with Tween 20 [PBST], 2% BSA, 0.025% sodium azide) for 1 h at 37°C. The plates were then incubated with 50 μl of serial dilutions of vaccine and virus-enhanced mouse serum at 4°C overnight. The plates were subsequently incubated with goat anti-mouse–HRP secondary antibody (200 ng/ml) (Invitrogen catalog no. A16072) in blocking buffer for 1 h at room temperature. ELISA plates were developed using 100 μl of TMB (3,3′,5,5′-tetramethylbenzidine) substrate (Thermo Fisher catalog no. 34029). The optical density at 450 nm (OD₄₅₀) reading was measured with a BioTek ELISA microplate reader.

Serum neutralization assay.

Focus reduction neutralization tests (FRNT) were performed as described previously (23). Briefly, serial dilutions of heat-inactivated serum from vaccinated mice were incubated with 50 to 70 FFU of DENV for 1 h at 37°C before infection of a monolayer of Vero cells in a 96-well plate. One hour after infection, cells were overlaid with 1% (wt/vol) methylcellulose in 2% fetal bovine serum (FBS), 1× minimal essential medium (MEM). Plates were fixed for 30 min with 4% paraformaldehyde (PFA) 48 h after infection. Staining involved primary antibody 9.F.10 (500 ng/ml) and secondary antibody goat anti-mouse–HRP (200 ng/ml) in PermWash buffer (0.1% saponin, 0.1% BSA, in PBS). Treatment with TrueBlue peroxidase substrate (KPL) produced focus-forming units that were quantified on an ImmunoSpot ELISpot plate scanner (Cellular Technology Limited).

Antiviral T cell quantification.

Spleens were collected from vaccinated mice, and splenocytes were collected. An overlapping 15mer peptide library from DENV2 ENV, DENV1 ENV, and DENV NS1 was obtained from BEI Resources, NIAID, NIH (catalog no. NR-507, NR-9241, and NR-2751). Individual peptides were pooled for ex vivo T cell stimulation. Spleens were ground over a 40-μm-pore-size cell strainer and brought up in RPMI 1640 medium with 10% FBS, HEPES, and 0.05 mM β-mercaptoethanol. Then, 2 × 10⁶ cells were plated per well in a round-bottom 96-well plate and stimulated for 6 h at 37°C, 5% CO₂, in the presence of 10 μg/ml brefeldin A and 10 μg of pooled peptide in 90% dimethyl sulfoxide (DMSO). Following peptide stimulation, cells were washed once with PBS and stained for the following surface markers: anti-CD8-PerCP-Cy 5.5 (clone 53-6.7), anti-CD3-AF700 (clone 500A2), and anti-CD19-BV605 (clone 1D3). Cells were then fixed, permeabilized, and stained for the following intracellular marker: anti-IFN-γ-APC (clone B27). The cells were analyzed by flow cytometry using an Attune-NXT.

ADE flow assay.

Serial dilutions of heat-inactivated serum from naive, vaccinated, or virus-infected mice were mixed with DENV2 and incubated for 1 h at 37°C. Fcγ receptor (CD32A)-positive K562 cells were infected with immunocomplexed virus at an MOI of 1 in a 96-well plate. After a 15-h incubation, cells were fixed with 4% PFA for 30 min and stained for intracellular ENV with 1A1D-2 (1/500) monoclonal antibody and anti-mouse 647-conjugated antibody (2 μg/ml; Invitrogen catalog no. A21235).

ADE viral replication assay.

Serum samples from virus-infected mice and mice receiving the WT, ΔFL, or GFP mRNA vaccine were separately pooled and heat inactivated. Serum was mixed with DENV2 and incubated for 1 h at 37°C at a 1/100 dilution. A total of 10,000 Fcγ receptor (CD32A)-positive K562 cells were infected with immunocomplexed virus at an MOI of 1 in a 200-μl volume in a 96-well plate. After a 48-h incubation to allow viral replication and egress, cells were centrifuged to separate cells from supernatant. Viral titers in supernatant were determined via FFA as described above.

Data analysis.

All data were analyzed with GraphPad Prism software. Statistical significance was determined by unpaired t tests for comparison of antibody titers and by log rank tests for comparisons of survival curves. Flow cytometry data were analyzed using FlowJo software (BD Biosciences).