HLA-A^*0201, -A^*2402, -B^*0702, and -B^*3501 monochain transgenic/H-2 null mice on the C57BL/6 background (34–36) were bred in the Institut Pasteur facility and used between 6 and 10 week of age. All animals were intraperitoneally inoculated with 2 mg of anti-ifnar1 antibody (clone MAR1-5A3, Interchim, France) (33) 24 h prior to DENV inoculation. For all experiments, mice were infected by retro-orbital injection of 10⁶ PFU DENV1 KDH0026A, in 200 μl PBS, then bled on days 1, 2, 3, and 6 after virus inoculation, for measurement of viremia and sacrificed on day 7 post-infection. All mouse experiments were performed following Institutional animal care and use committee-approved animal protocols.

Given the identification of CD8+ T cell epitopes from previously infected donors (24, 28, 37), 4 regions have been selected in the non-structural proteins NS3, NS4B, and NS5, for a total of 540 amino acids: 2 regions in NS3 (185 and 134 amino acids, respectively), 1 region in NS4B (86 amino acids) and 1 region in NS5 (135 amino acids) (Figure 1). Based on the sampled genetic diversity of the 4 serotypes of DENV, and with the idea that T cell epitopes are either conserved among different serotypes, or are serotype-specific with nevertheless the ability to induce cross-reactive CD8+ T cell responses, a prototype consensus sequence based on epidemic strains of DENV1 has been selected, the DENV1-NS T cell poly-epitope (patent WO2015/197565, initially filled on June 23, 2014) (Figure 1). A total of 2,033 full-length DENV genome sequences (865 for serotype 1, 1,663 for serotype 2, 427 for serotype 3 and 63 for serotype 4) were aligned using MAFFT (38), with manual adjustments according to the amino acid sequence. Mean pairwise sequence identity was evaluated at each position intra- and inter-serotypes at the nucleic and the amino-acid levels. Analysis of the concatenated regions of interest revealed strong intra-serotype conservation, and generally a higher degree of sequence identity compared to the genome as a whole, including for inter-subtypes comparisons. A serotype 1 consensus sequence was selected, as it presented the highest average sequence identity with the 4 serotypes (vs. serotype 1: 99.48%; serotype 2: 83.48%; serotype 3: 89.39%; serotype 4: 76.95%). For DNA immunization, the optimized nucleotide sequence encoding DENV1-NS poly-epitope was cloned in the pcDNA3.1 plasmid, under the CMV promoter. For mRNA vaccination, a lipid nanoparticle (LNP) encapsulated modified mRNA vaccine encoding DENV1-NS was prepared as described previously (39).

Chemically modified mRNA was synthesized in vitro by T7 RNA polymerase-mediated transcription from a linearized DNA template containing an open reading frame flanked by a 5′ and 3′UTR and a polyA tail. Uridine was fully replaced with N1-methylpseudouridine, and Cap1 was utilized to increase mRNA translation efficiency.

Lipid nanoparticle formulations were generated as described previously (40). Briefly, lipid stocks were dissolved in ethanol at a molar ratio of 50:10:38.5:1.5 (ionizable lipid: helper lipid: structural lipid: PEG-lipid) and mixed with mRNA at a ratio of 3:1 (mRNA:lipid). mRNA-loaded nanoparticles were exchanged into final storage buffer and had final particle sizes of 80–100 nm, >80% encapsulation, and <10 EU/mL endotoxin.

Before immunization, mice were first anesthetized with a mix solution of xylazine 2% (Rompun, Bayer santé, Loos, France) and ketamine 8% (Imalgen 1000, Merial, Lyon, France) in PBS through the intraperitoneal (I.P.) route according to individual animal weight and duration of anesthesia. DNA immunization was performed following a prime boost administration regimen at 3-week interval by intradermal (I.D.) injection of plasmid DNA pcDNA3.1 encoding the DENV1-NS sequence (each injection consists of 2 simultaneous I.D. injections of 50 μg plasmid DNA in the lower back followed by in vivo electroporation) and spleen cells were harvested 10 days after the boost. The electroporation settings, using the AgilePulse apparatus (BTX, Harvard apparatus, MA, USA) consist of 3 Voltages groups: 450 V, with a pulse length of 50 microseconds, a pulse interval of 0.2 microseconds and 1 pulse; 450 V, with a pulse length of 50 microseconds, a pulse interval of 50 milliseconds and 1 pulse; and 110 V, with a pulse length of 10 milliseconds, a pulse interval of 20 milliseconds and 8 pulses. The immunization with mRNA was performed following a prime boost regimen at 3-week interval or 4-week interval, by intramuscular (I.M.) injection of LNP-based mRNA encoding the DENV1-NS sequence or irrelevant mRNA as a negative control (each injection consists of one I.M. injection of 10 or 2 μg mRNA vaccine encoding the DENV1-NS sequence or 10 μg irrelevant mRNA). For quantification of the T cell responses following mRNA immunization, the prime boost regimen was performed at 3-week interval, and spleen cells were harvested 7 days after the boost. For the evaluation of the vaccine efficacy, the prime boost regimen was performed at 4-week interval, and mice were challenged with the virus 4 weeks after the boost.

The in vivo and the in vitro assays were conducted using the DENV1 KDH0026A strain (provided by Dr. L. Lambrechts, Institut Pasteur, Paris). The DENV1 virus was grown using the Aedes Albopictus mosquito cells line C6/36 cultured in Leibovitz's L-15 medium supplemented with 10% fetal bovine serum containing 0.1 mM non-essential amino acids and 1x tryptose phosphate broth. Vero-E6 cells were provided by Dr. M. Flamand (Institut Pasteur, Paris).

For assessment of viremia, blood was collected by retro-mandibular puncture in EDTA coated-microvette tubes (Fisher Scientific, Illkirch, France). After centrifugation, plasma samples from DENV-infected mice were extracted with the RNeasy Mini Kit (Qiagen). DENV RNA levels were determined by TaqMan one-step quantitative reverse transcriptase PCR (qRT-PCR) on a QuantStudio 12K Flex Real-time PCR system (Life Technologies) using standard cycling conditions. Viral burden is expressed on a log₁₀ scale as viral RNA equivalents per milliliter after comparison with a standard curve produced using serial 10-fold dilutions of DENV1 RNA from known quantities of infectious virus. The following primer sets were used:

Forward: 5′-GGAAGGAGAAGGACTCCACA-3′;

Reverse: 5′-ATCCTTGTATCCCATCCGGCT-3′;

Probe: 5′-(FAM) CTCAGAGACATATCAAAGATTCCAGGG-3′ (MGB).

All peptides were synthesized by Proimmune Ltd (Oxford, UK), as crude material, and were tested individually in Elispot assay and in intracellular cytokine staining for IFN-γ and TNF-α secretion. The peptides used in this study and corresponding to DENV1 sequences or to their DENV2, 3 or 4 variants are listed in Table 1.

Mouse spleen cells were collected and depleted of red blood cells and IFN-γ-producing splenocytes were quantified by ELISPOT assays after a 24 h period of stimulation with peptides, as already described (43). Briefly, 96-well nitrocellulose-backed plates (Multiscreen, Merck-Millipore, Molsheim, France) were coated with anti-mouse IFN-γ mAb at 5 μg/ml (BD Pharmingen, France) in 50 μl of 50 mM carbonate buffer (pH 9.6) overnight at 4°C. Wells were blocked with 200 μl of complete medium at room temperature for 2 h and then washed three times with serum-free medium. The coated wells were filled in quadruplicate with splenocytes from immunized mice in complete α-MEM medium with individual peptides at 2 μg/ml and incubated at 37°C for 20 h. After incubation, the wells were washed three times with PBS-0.05% Tween 20, and then incubated with 50 μl biotinylated anti-mouse IFN-γ (BD bioscience, France) at 1 μg/ml in PBS Tween for 1 h at room temperature. The spots were developed using streptavidin-alkaline phosphatase (Mabtech, Stockholm, Sweden) and BCIP/NTB substrate (Promega, Madison, MI, USA) and counted using an automated ELISPOT reader (Immunospot, Cellular Technology Limited, Cleveland, OH, USA). The number of IFN-γ-producing cells was expressed as spot-forming cells (SFC) relative to 1 × 10⁶ spleen cells. Values were calculated by subtracting the number of spots detected in non-stimulated control wells. Values were considered positive if they were equal to or >20 spots and at least three times above the means of the unstimulated control wells. As positive wells, cells were stimulated with ConA at 10 μg/ml (Sigma Aldrich, Lyon, France).

Spleen cells from immunized mice (100 μl/well in 96 U-bottom plates) were incubated with individual peptides (adding 50 μl at 4 μg/ml for 2 h, and 50 μl Golgi Stop at 4 μl/ml for 4 h), in complete medium containing 10% FCS. Cells were then harvested, membrane stained using anti-mouse CD4-V500 (BD Biosciences), CD3-alexa488 (BD Biosciences), and CD8-PerCP (eBioscience) antibodies, and intracellular staining was performed using the intracellular staining kit (BD Biosciences) and anti-mouse IFN-γ-APC and TNF-α-eFluor 450 antibodies (eBioscience).

The quantification of neutralizing activity of antibodies against DENV was determined using a flow cytometry-based assay, as described previously (44). Briefly, 2-fold serial dilutions of plasma samples were incubated at 37°C for 1 h with a 1/200 dilution of virus inducing 15–20% infection. Virus -antibody mixture was then added to 5 × 10⁴ Vero cells for 2 h at 37°C, after which cells were washed with fresh medium and incubated for 24 h at 37°C. Cells were then fixed with 4% paraformaldehyde, stained using the intracellular staining kit (BD Biosciences) with 4G2 antibody conjugated to Alexa Fluor^TM 488, and the percentage of infected cells was measured by flow cytometry. Plasma samples from mice immunized with mRNA CT vaccine and from ifnar^−/− mice injected intravenously with 10⁶ pfu DENV1 were used as negative and positive controls, respectively.

Differences between 2 groups were evaluated using non-parametric Mann-Whitney U-test in the assessment of the immunogenicity and the immune protection or unpaired t-test in the evaluation of the expansion and the contraction phases of the T cell response. All computer analyses were performed using GraphPad Prism 7 (GraphPad Software Inc. La Jolla, CA, USA). P < 0.05 were considered statistically significant.

Given the identification of CD8⁺ T cell epitopes from human donors and to minimize the dilution of the immune response observed following DENV infection, a minimal DENV antigen has been designed from conserved and highly antigenic T cell epitopes. Using compiled DENV T cell epitope distribution and strength (24, 28, 37, 41) and the consensus prediction method available through the IEDB Analysis Resource (available at: www.iedb.org), we have selected four regions from a DENV serotype 1 consensus that are enriched in CD8 epitopes: two regions in NS3, one in NS4B, and one in NS5 (Figure 1). The resulting DENV1-NS poly-epitope corresponds to the concatenation of amino acid positions 1650-1829, 1959-2092, 2262-2347, and 2766-2899 based on the reference sequence NP_059433. Within serotypes, pairwise identity along the genome ranged between 0.878 and 1 for DENV1, 0.816 and 1 for DENV2, 0.847 and 1 for DENV3 and 0.94 and 1 for DENV4.

To determine whether the DENV1-NS poly-epitope can be produced endogenously and then processed and presented in the context of human HLA class I molecules, we performed DNA-based immunization combined with electroporation (EP) in HLA-A^*0201, -A^*2402, -B^*0702, and B^*3501 monochain transgenic/H-2 null mice (34, 36), which are associated with low T cell responses (for the HLA-A^*0201 and -A^*2402 alleles) and high T cell responses (for the HLA-B^*0702 and -B^*3501 alleles) (24), and represent the most frequent alleles in Caucasoid, Oriental and Amerindian ethnic groups, from the allele frequency Net Database website (available at: www.allelefrequencies.net). The animals were immunized by intradermal (I.D.) injection of plasmid DNA expressing the DENV1-NS under a CMV promoter and the interferon (IFN)-γ response of spleen cells was quantified by Enzyme Linked Immunospot (ELISpot) assay against individual peptides. Among the potential epitopes expressed by DENV1-NS, and predicted from the Immune Epitope Database and Analysis Resource (www.iedb.org) to bind to HLA-A^*0201, or shown in literature to induce a significant T cell response in the context of HLA-A^*0201 molecules (41, 42, 45), 2 antigenic peptides were identified, which elicit a strong T cell response in the HLA-A^*0201 transgenic mice: p30 and p56 located in the NS3.1 and the NS5 regions of the DENV1-NS, respectively (Figure 2 and Table 1). In the immunized HLA-A^*2402 transgenic mice, 3 peptides were also identified: p17 and p32 located in the NS3.2 region and p33 located in the NS4B region (Figure 2). These results confirm previous identification of HLA-A^*0201- or HLA-A^*2402-restricted epitopes, as targets for T cell responses in humans (16, 28, 29, 37, 46), or predicted to bind strongly these alleles (www.iedb.org). Finally, 3 and 4 peptides have been identified in HLA-B^*0702 and HLA-B^*3501 transgenic mice, respectively, with a higher number of antigenic peptides inducing a stronger T cell response in the HLA-B^*3501 mice, in comparison with HLA-B^*0702 transgenic mice (Figure 2). These data clearly show that the DENV1-NS poly-epitope is expressed and processed correctly, resulting in the presentation of antigenic peptides, which elicit a strong T cell response in the context of different HLA molecules. They also show that all the antigenic peptides identified in DENV1-NS correspond to T cell epitopes previously identified from human donors or from HLA transgenic mice infected with DENV.

Intramuscular delivery of LNP with encapsulated modified mRNA revealed strong immunogenicity in vivo (39). Therefore, we inoculated HLA transgenic mice with the LNP-based mRNA vaccine encoding DENV1-NS. Eight to 10-week old HLA-B^*3501 transgenic mice were divided into 3 groups, which received an intramuscular inoculation of 10 or 2 μg of the mRNA encoding DENV1-NS or 10 μg of an irrelevant mRNA as a negative control. After a booster immunization at day 28, spleen cells from immunized animals were harvested 8 days later, at day 36, and the frequency of CD8 T cells secreting IFN-γ and TNF-α and spleen cells producing IFN-γ was evaluated by intracellular staining and by ELISpot assay, respectively. Both assays were used in parallel, as they provide complementary information, in particular by allowing the identification of responding cells with intracellular cytokine measurement by flow cytometry, and the detection of small number of IFN-γ-producing cells in response to poorly immunogenic peptides with the ELISpot assay which is more sensitive. Flow cytometry analyses of responding CD8 T cells after 6 h in vitro stimulation with the HLA-B^*3501-restricted peptides p49, p50, or p51 revealed different capacity of these peptides to stimulate CD8 T cells, with 26% of CD8 T cells producing IFN-γ and TNF-α- against p49, whereas only 0.85 and 0.06% of CD8 T cells produce both cytokines in response to p50 and p51, respectively (Figure 3A). As the amino acid sequence of the antigenic peptides may vary between the 4 DENV serotypes, with a few amino acid substitutions located in the HLA binding sites (anchor residues) (Table 1), we asked whether mice immunized with the DENV1-NS vaccine could recognize peptide variants, which are derived from heterotypic DENV serotypes. As shown in Figures 3A,B, both the p49 peptide (TPEGIIPAL for DENV1, 3) and its 2 serotype variants (TPEGIIPSM and TPEGIIPTL for DENV2 and DENV4, respectively) stimulate CD8 T cells to produce IFN-γ and TNF-α. We then compared the frequency of T cells responding to the DENV1-specific p50 and p51 peptides, with the frequency of cells responding to the serotype variants of these peptides, using the ELISpot assay, which is more sensitive than flow cytometry. Results show that while p50 and p51 induce a significant T cell response, none of the 2 DENV serotype variants of p50 (LPVWLAYRVA for DENV2 or LPVWLAHKVA for DENV3) stimulate T cells, and only 1 serotype variant of p51 (TLYAVATTF for DENV2) can stimulate T cells. Interestingly, the DENV2 variant of p51 is more potent in stimulating T cells than the DENV1-specific p51 peptide, both for the 10 μg and the 2 μg dose of DENV1-NS vaccine (Figure 3B). Altogether, these data show that immunization with mRNA vaccine encoding DENV1-NS induces a strong stimulation of CD8 T cells, which in some cases, could be highly cross-reactive against heterotypic peptides.

Most mouse models for dengue infection and disease make use of mice lacking type I or type I and type II IFN receptors, rendering them highly susceptible to DENV infection (47). However, most of these models are not relevant to study the role of memory and effector memory T cells in the immune protection against DENV infection. For instance, type I IFN plays a major role in controlling the CD8 T cell response to viral infection, notably by blocking the expansion of pre-existing non-specific memory T cells (out-of-sequence signaling), while promoting the proliferation of antigen-specific CD8 T cells at the beginning of the response (in-sequence signaling) (48). Importantly, this regulation depends essentially on the timing of type I IFN exposure relative to T cell receptor signaling. In this context, we developed a mouse model, which allows the clonal expansion of antigen-specific T cells, while retaining DENV susceptibility. This was achieved by injecting, 1 day before the challenge with DENV, the anti-IFNAR antibody to block transiently the type I IFN signaling, as already reported for WNV and ZIKV infection in wild type mice (32, 49). The benefit of this approach is that it ensures the normal development of memory T cell responses, because blockade of type I IFN signaling is performed just before the challenge, once the immune response is established.

Thus, to verify that the transient blockade of type I IFN signaling, performed at a late stage after the prime with the antigen (at day 27 after the immunization), does not prevent the establishment of a memory T cell response, we quantified T cell responses before and after the boost, with and without blocking IFN response (Figure 4A). Quantification of the T cell response in HLA-A^*0201 and HLA-A^*2402 transgenic mice revealed a 7-fold and a 50-fold increase in the frequency of p30- and p32-specific T cells, respectively, between days 27 and 36, confirming the expansion of antigen-specific T cells after the boost at day 28. The same T cell proliferation was observed in response to p30 (DENV4) or p32 (DENV2) peptides corresponding to serotype variants (Figure 4A). More importantly, the same proliferation was observed in these transgenic mice, with or without anti-IFNAR treatment, showing that depletion of type I IFN signaling at a late stage does not prevent the proliferation of antigen-specific T cells, which expand after the boost (Figure 4A). To assess also the effect of anti-IFNAR treatment on the contraction phase of effector T cells, the antibody was administered at a late stage after the boost (at day 55), and peptide-specific T cell responses were measured in HLA-A^*0201 and -A^*2402 treated or not with the antibody.

As shown in Figure 4B, in HLA-A^*0201 transgenic mice, the anti-IFNAR treatment does not prevent but rather increases the proliferation of p30-specific T cells, whereas it does not modify the proliferation of p32 or p32 (DENV2.1)-specific T cells in HLA-A^*2402 mice at day 56, after the contraction phase. This result shows that, while a transient blockade of type I IFN signals in HLA-A^*0201 can increase the proliferation of peptide-specific T cells, it does not prevent the contraction phase of peptide-specific T cells in HLA-A^*2402 mice, which is required for the development of memory T cell responses (Figure 4B).

Given the potential to induce strong T cell responses against several NS epitopes, we wanted to assess efficacy of the mRNA vaccine encoding DENV1-NS, in immunocompetent mice expressing different HLA class I molecules. The ability to elicit immunity and protection against DENV infection was tested in the HLA-A^*0201, -A^*2402, and B^*3501 transgenic mice, which developed a strong T cell response against different peptides located in the NS3, NS4B and NS5 proteins. The animals were vaccinated following a prime-boost immunization, with a boost at day 28 after the prime, followed by a treatment with anti-IFNAR antibody at day 55 and a challenge at day 56 with the DENV1 strain KDH0026A, derived from a clinical isolate.

In the HLA-A^*0201 mice, 5 out of 6 mice immunized with the control mRNA vaccine developed a significant viremia, at day 2 after the challenge, whereas 3 out of 6 mice vaccinated with the mRNA encoding DENV1-NS revealed a detectable viremia, of which 2 mice had viremia values close to the lower limit of quantitative detection (LLOQ) (Figure 5A). More strikingly, 5 out of 6 vaccinated HLA-A^*2402 mice revealed no viremia at days 1, 2, and 3 after the challenge, whereas 4 out of 5 HLA-A^*2402 mice immunized with the control mRNA developed a significant viremia until day 3 (Figure 5B). Finally, 8 out of 10 vaccinated HLA-B^*3501 mice revealed no detectable viremia at day 2 after the challenge, whereas 9 out of 9 HLA-B^*3501 mice immunized with the control mRNA vaccine developed a significant viremia at day 2 (Figure 5C). At day 3, no viremia was detected in vaccinated HLA-B^*3501 animals, whereas 5 out of 8 HLA-B^*3501 mice immunized with the control mRNA vaccine still had significant viremia, indicating that even in case an initial viremia is developed in vaccinated mice, it is resolved faster than in non-vaccinated animals (Figure 5C). Finally, it should be also noted that none of the vaccinated or control mice died after DENV infection.

To rule out the possibility that the immune protection induced by DENV1-NS vaccination could be mediated in part by neutralizing antibodies, plasma samples from vaccinated mice were tested for their neutralizing potential against DENV infection in vitro. Using a flow cytometry-based assay, as described previously (50), we thus analyzed the ability of immune serum from mice vaccinated with DENV1-NS to neutralize DENV1 infection of Vero cells. While serum from immunocompromised mice lacking type I IFN signaling (ifanr^−/− mice) and infected with DENV1 can efficiently neutralize DENV1 infection, plasma samples from HLA-B^*3501 mice immunized either with mRNA DENV1-NS vaccine or with CT mRNA did not reveal any neutralizing activity against DENV1 infection, even at a high concentration (Supplementary Table 1). These results confirm the absence of neutralizing antibodies induced after DENV1-NS vaccination, which could mediate protection against DENV1 infection. Taken together, these data demonstrate the efficacy of a T-cell based vaccine, which targets the immunodominant T cell epitopes from NS proteins, in the induction of a protective immunity against DENV infection and disease.

CR, ES-L, and AS are inventors on a patent filing (patent WO2015/197565, initially filled on June 23, 2014) related to this work. KB and GC are current or previous employees of Moderna, Inc. and receive salary and stock options as compensation for their employment. GC is employed by Beam Therapeutics, and JT and PL-D are current or previous employees of Invectys. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.