Dendritic cell targeting virus-like particle delivers mRNA for in vivo immunization

1 mRNA vaccine was approved clinically in 2020. Future development includes delivering 2 mRNA to dendritic cells (DCs) specifically to improve effectiveness and avoid off-target 3 cytotoxicity. Here, we developed virus-like particles (VLPs) as a DC tropic mRNA vaccine 4 vector and showed the prophylactic effects in both SARS-CoV-2 and HSV-1 infection models. 5 The VLP mRNA vaccine elicited strong cytotoxic T cell immunity and durable antibody 6 response with the spike-specific antibodies that lasted for more than 9 months. Importantly, 7 we were able to target mRNA to DCs by pseudotyping VLP with engineered Sindbis virus 8 glycoprotein and found the DC-targeting mRNA vaccine significantly enhanced the titer of 9 antigen-specific IgG, protecting the hACE-2 mice from SARS-CoV-2 infection. Additionally, 10 we showed DC-targeted mRNA vaccine also protected mice from HSV-1 infection when co- 11 delivering the gB and gD mRNA. Thus, the VLP may serve as an in situ DC vaccine and 12 accelerate the further development of mRNA vaccines.

The future of mRNA vaccine has been widely recognized since the breakout of severe acute 23 approximately 90-95% efficacy with minimal side effects [11][12][13][14] . While SARS-CoV-2 mRNA 27 vaccines have been widely dosed in developed countries and are effective in preventing severe 28 COVID-19 outcomes, the virus transmission is still not under full control 15 . Additionally, the 29 potential of mRNA vaccine beyond the SARS-CoV-2 infection awaits further exploration. 30 As mRNA are vulnerable to RNA nuclease and cannot enter cells by themselves, a variety of 1 carriers have been developed for mRNA transfer including lipid-nanoparticles (LNPs), polymers, 2 peptides, virus-like replicon particles (VRPs) and dendritic cells (DCs) 16 . LNP is now the first 3 runner due to the success in SARS-CoV-2 vaccines. The current LNP mRNA is unable to control 4 cell specificity and can be taken up by almost any cell type, near or far from the site of injection 17 . 5 DCs are the major antigen-presenting cells (APCs) and critical for vaccine function by 1) 6 instigating the T cell immune responses through antigen processing to T cells 18,19 , 2) processing 7 antigens to B-cells and inducing antibody responses 20,21 . 8 The DC-based vaccine has been approved by US FDA in use for the treatment of prostate cancer, 9 however, it was made ex vivo and labour intensive, weakening the availability to a broader 10 population. The DC-targeting strategy makes DC vaccine in situ, therefore, lowering the cost and 11 simplifying the manufacture. Moreover, it has been suspected that non-professional APCs with 12 antigen mRNA in translation may become a target of CD8+ T cell-mediated killing, which has been 13 linked to 'Covid-arm' that develops in some patients after receiving the COVID-19 mRNA 14 vaccine 22,23 . Furthermore, antibody-dependent cellular cytotoxicity (ADCC) may also destroy the 15 cells with antigen proteins inserted into, or secreted and associated with the plasma membrane 22, 24 . 16 Targeting DC in situ has been achieved by using pseudotyped lentiviral vectors (LVs) with Sindbis 17 virus glycoprotein as a ligand for DC-SIGN, however, due to the reverse transcription step, the DC-18 targeting LV vaccine has potential risks of insertional mutagenesis 25 . LNPs have also been 19 conjugated to specific antibodies or ligands to target DC for in vitro and in vivo evaluation, 20 although the evidences for effective LNP based DC-targeting mRNA vaccine are rare 22 . 21 Driven by the need for mRNA delivery vectors with DC specificity to improve antivirus responses 22 while limiting the possible off-target cytotoxicity, here, we developed a virus-like particle (VLP)-23 based mRNA vaccine technology and showed VLP delivered antigen mRNA elicited strong and 24 durable adaptive immune responses. The specific IgG was maintained at a high level for at least 9 25 months. Additionally, the VLP-mRNA was also compatible with the intranasal administration route 26 and induced significant mucosal immunity. We found the DC-targeting VLP elicited significantly 27 higher antigen-specific IgG response than the non-specific counterpart. Importantly, the DC-28 targeting VLP-mRNA vaccine efficiently protected mice from live virus infection in both SARS-29 CoV-2 and HSV-1 infection models. Together, the VLP is able to deliver mRNA specifically to 30 DCs and may accelerate the further development of mRNA vaccines including against infectious 31 diseases without vaccines and cancers. 32

Results 1
Design and characterization of VLP-based SARS-CoV-2 mRNA vaccine. To show the potential 2 of VLP as an mRNA vaccine carrier, we performed a proof-of-concept study by designing a 3 candidate SARS-CoV-2 mRNA vaccine which encodes the full-length spike mRNA composed of a 4 signal peptide from the human heavy chain of IgE and the codon-optimized sequence (Fig. 1a). To 5 increase the stability and expression of the spike, we also introduced two proline substation 6 mutations (K986P/V987P) in the S2 (Fig. 1a) 26 . We have previously shown delivery of Cas9 7 mRNA with lentivirus-derived VLP mediated efficient genome editing in vivo in the different 8 disease models 27,28 , however, it is unclear if VLP can serve as a vaccine platform. To package the 9 full-length spike mRNA into VLPs, we inserted MS2 stem-loop repeats in its 3' terminus between 10 the stop codon and the polyA signal. This design allows the spike mRNA to be internalized via its 11 interaction with the MS2 coat protein fused in the N-terminus of GagPol which can self-assemble 12 into VLP (Fig. 1b). As VSV-G coated lentiviruses are efficiently taken up by APCs and show the 13 high immunogenicity of antigens 29 , we therefore firstly pseudotyping VLP with VSV-G by 14 providing the pMD.2G plasmid in the production process. To analyse the morphology of the VLP, 15 we conducted electron microscopy which was in round shape with a size of approximately 100 nm 16 ( Fig. 1c). 17 To find out if spike mRNA has been indeed packaged into lentiviral particles, we performed RT-18 qPCR on VLP and normalized it to the traditional lentiviral vector which had two copies of RNA. 19 We found on average 3 copies for wildtype spike mRNA and 4 copies for the mutant one in each 20 VLP (Fig. 1d). As the spike is an envelope protein, we sought to examine if the protein could 21 automatically assemble into the membrane of VLP by Western blot analysis of the lysates of VLP 22 with an integration-defective lentiviral vector (IDLV) as the control (Fig. 1e). We found successful 23 decoration of both with or without proline mutations on the VLPs whereas more mutant spike 24 proteins could be loaded which was in accordance with the RT-qPCR analysis. As glycosylation 25 impacts the immunogenicity and immunodominance of a vaccine 30 , we set out to examine the 26 glycosylation status of the spike on the surface of VLPs. Notably, the S2 bands shifted downwards 27 after PNGase F treatment indicating that the spikes on VLPs were modified by N-linked 28 glycosylations mimicking the characteristics of SARS-CoV-2 revealed by the mass spectrometric 29 approach (Fig. 1e) 31 . 30 To examine whether the Spike mRNA in the VLP could be delivered intracellularly in a manner 1 mediated by VSV-G, we transduced 293T cells that were not infected by SARS-CoV-2 unless 2 supplemented with hACE2 32 , and then harvested the cells 36 h postinfection for Western blot (Fig.  3 1f). We found two major bands for spikes which were likely glycosylated full-length singlet spikes 4 and their dimeric/trimeric forms (Fig. 1f). Additionally, we confirmed efficient delivery of spike to 5 293T cells by confocal analysis (Fig. 1g). From Western blot and confocal microscopy, we 6 consistently observed more spike-mut antigens were delivered by VLP, we therefore chose the 7 mutant spike for in vivo evaluation. 8 As the mRNA transcribed in vitro for LNP delivery is recognized by intracellular RNA sensors 9 unless chemically modified 33,34 , we examined the innate immune property of the VLP-carrying 10 mRNA. Using THP-1 derived macrophages as a model of nucleic acids sensing, we found no 11 significant changes of the type I interferon (IFN) and IFN-stimulated genes ISG-15, and retinoic 12 acid-inducible gene I (RIG-I) ( Fig. 1h-j), suggesting that spike mRNA in the VLP was not 13 immunogenic which could be explained by the fact that it was produced intracellularly and shared 14 same modifications as any other endogenous mRNA. 15 VLP mRNA induces robust spike-specific and durable antibody responses. To evaluate the 16 potential of VLP-mRNA as a vaccine platform, we vaccinated the C57BL/6J mice (n=6 for each 17 group) with VLP carrying mutant spike mRNA (VLP S-mut) via footpad (Fig. 2a). Two weeks 18 later, we performed an enzyme-linked immunosorbent assay (ELISA) using the sera from mice to 19 get access to the spike-specific IgG. As shown in Fig. 2b, we observed significant elicitation of the 20 spike-specific IgG. To evaluate the level of neutralizing antibodies, we performed the neutralizing 21 assay using spike pseudotyped HIV which encodes firefly luciferase -a well-established 22 pseudovirus neutralization assay 35 and found a single injection of VLP S-mut was sufficient to 23 induce potent neutralizing immune responses (Fig. 2c). To confirm the neutralizing activity of 24 vaccinated sera, we adopted the spike pseudotyped lentiviral vector which encodes GFP to 25 transduce Huh-7 cells. We found pre-incubation with 1:40 diluted sera from vaccinated mice almost 26 completely abolished the fluorescence whereas the transduction for VSV-G pseudotyped lentivirus 27 was not apparently affected indicating the spike-specific neutralizing activity (Fig. 2d). Importantly, 28 induction of antibodies with high neutralization titers was demonstrated using live SARS-CoV-2 29 with an average EC50 titer of 1319 (Fig. 2e). Additionally, we analysed the neutralizing activity of 30 the VLP mRNA vaccine against the B.1.617.2 strain pseudovirus, which showed no significant 31 reduction in the EC50 titer compared to the wildtype strain (Fig. 2f). 32 To evaluate dynamic changes of VLP mRNA-induced spike-specific antibodies, we performed a 1 short-term follow-up starting from 1 day post-vaccination and a long-term follow-up up to 9 months 2 after vaccination. We found the spike-specific IgG was significantly enhanced on day 7, but was not 3 evident on day 1, 3, and 5 (Fig. 2g). In the long-term follow-up, we found a single dose vaccination 4 induced a durable spike-specific IgG response which was maintained at a high level up to 36 weeks 5 post-immunization (Fig. 2f). Notably, no weight loss was found during the course of vaccination 6 suggesting the safety of VLP-mRNA vaccination ( Supplementary Fig. 1). Interestingly, 7 administration of VLP-mRNA via intranasal route elicited spike-specific IgA in the lung, 8 suggesting that this vaccine platform may also be used as an intranasal vaccine to induce mucosal 9 immunity to block the SARS-CoV-2 infection at the first contact site ( Supplementary Fig. 2). 10 Linear epitope landscape in the VLP mRNA vaccinated mice. To dissect the linear epitope 11 profiles of the spike-specific antibodies in the VLP mRNA vaccinated mice, we used a peptide 12 microarray which contains short peptides covering the full-length of spike 36-38 . We found varying 13 intensities of signals corresponding to certain spike peptides for the vaccinated group while no 14 signal was observed for the placebo-treated mice (Fig. 3a). Next, we quantified the signal intensity 15 for antibodies against the S1 domain and receptor-binding domain (RBD), respectively, and found 16 the sera from vaccinated mice elicited significantly higher signals for both domains suggesting the 17 presence of high amounts of S1 and RBD specific IgG in vaccinated mice (Fig. 3b). 18 To access the panorama of epitopes, we made a heat map for all the 6 vaccinated mice ( Fig. 3c and  19 Supplementary Table 1). We found five linear epitopes (S1-55, 57, 60, 76 and 88) located on RBD 20 which is the domain responsible for the host receptor recognition, revealing the key linear motifs on 21 the RBD that are susceptible to specific antibodies (Fig. 3c). More specifically, linear epitope S1-76 22 located on the center of RBM (receptor binding motif) is the direct binding interface with hACE2. 23 Notably, although the identified epitopes were overall highly diverse, we also found three epitopes, 24 i.e. S2-22, S2-76, and S2-83, were shared by 66.7% of the vaccinated mice. Interestingly, the S2-22 25 epitope also appeared in the majority of the convalescents uncovered by the peptide microarray 36 . 26 Moreover, the S2-76 and S2-83 epitopes are conserved epitopes among different coronaviruses 27 ( Supplementary Fig. 3). Particularly, the S2-83 epitope from the heptad repeat 2 (HR2) region is 28 expected to undergo dramatic structural refolding upon receptor activation, leading to the formation 29 of a six-helix bundle structure that finally drives membrane fusion ( Fig. 3d and e). 30

DC-targeting VLP mRNA elicited a stronger immune response in vivo.
To target VLP 1 specifically to DC, we used an engineered Sindbis virus glycoprotein (designated SV-G) which 2 recognized the DC-SIGN, a surface protein of DC, to replace the broad tropic VSV-G (Fig. 3a). is preferably transducing DCs (Fig. 3b). Next, we assessed and compared the humoral and cellular 8 immune responses of SV-G and VSV-G pseudotyped VLP in vivo (Fig. 3c). Six-week-old 9 C57BL/6 mice (n=4) were immunized 2 µg p24 SV-G VLP-GFP or VSV-G VLP via footpad 10 injection. Humoral immune responses were evaluated at 14 days post-immunization by ELISA. We 11 found the DC-specific SV-G VLP significantly enhanced the level of spike-specific IgG by nearly 1 12 magnitude (Fig. 3d). Interestingly, the level of p24-specific IgG for SV-G VLP was lower 13 compared to VSV-G (Fig. 3e). Furthermore, we set out to evaluate the spike-specific T cell 14 responses for both the DC-targeting and the non-targeting VLP mRNA and found both elicited 15 strong T cell immune response as shown by the IFN-γ, TNF-α and IL-6 enzyme-linked 16 immunosorbent spot (ELISPOT) assays after stimulating splenocytes with a spike-peptide pool 17 ( Fig. 3f-h). Notably, SV-G VLP mRNA vaccination showed averagely more spot forming units 18 (SFUs), although the difference between the two pseudotypes was insignificant ( Fig. 3f-h). 19

DC-targeting VLP mRNA vaccine protected hACE2 mice from the SARS-CoV-2 challenge. 20
To evaluate whether the DC-targeting VLP mRNA vaccine is able to protect mice from live SARS-21 CoV-2, we challenged hACE2 transgenic mice with live SARS-CoV-2. To acquire optimal 22 efficacy, hACE2 transgenic mice (n = 6 per group) were dosed twice each with 1.5 µg p24 VLP 23 mRNA vaccine (Fig. 5a). The mice were then inoculated by intranasal infection of 10 5 p.f.u SARS-24 CoV-2 (BetaCoV/Wuhan/WIV04/2019) two weeks after boost vaccination. We detected strong 25 anti-SARS-CoV-2 neutralization antibodies at day 28 with a mean EC50 value of 2643 (Fig. 5b). 26 The weight of mice was monitored daily before euthanasia 3 days post-challenge. We found 27 vaccinated mice keep growing in contrast to the unvaccinated mice which lost 2% weight on 28 average (Fig. 5c). Next, we analyzed the viral RNA levels and found significantly reduction of viral 29 loads in the lung and trachea of vaccinated mice ( Fig. 5d and 5e). 30 To analyse the efficacy of VLP mRNA vaccination on lung protection, we conducted 31 immunofluorescence microscopy which showed the SARS-CoV-2 was hardly detectible in the lung 32 of vaccinated mice in contrast to the mock vaccinated mice (Fig. 5f). Moreover, we performed 1 hematoxylin and eosin (HE) staining to analyse the pathology SARS-CoV-2 infected mice, which 2 showed that the control mice had alveolar epithelial cell hyperplasia, local pulmonary alveoli 3 shrank and infiltration of inflammatory cells in lung interval (Fig. 5g). In contrast, vaccinated mice 4 showed attenuation of the inflammatory response with only mild perivascular and alveolar 5 infiltration of inflammatory cells observed in very few areas (Fig. 5g). Together, these results 6 indicate that the DC-targeting VLP mRNA vaccine mediated efficient protection against live 7 SARS-CoV-2 infection and prevented the inflammatory reaction. 8 DC-targeting VLP mRNA vaccine protected mice from the HSV-1 challenge. To evaluate the 9 flexibility of VLP mRNA as a vaccine platform, we designed an HSV-1 VLP mRNA vaccine by 10 incorporating HSV-1 gB and gD mRNA into the SV-G pseudotyped VLP (Fig. 6a). 14 days after 11 the prime-boost vaccination, the depilated mice were challenged with 10 7 p.f.u of HSV-1 17 syn+ in 12 the format of 10 L on the abraded skin (Fig. 6b). Prime vaccination significantly induced the 13 neutralizing IgG against HSV-1 while the second vaccination further boost the neutralizing 14 antibody titers by 4-fold (Fig. 6c). Interestingly, although the gB and gD antigens were derived 15 from HSV-1, we detected cross-neutralizing activity against HSV-2, suggesting the vaccine might 16 also be functional against HSV-2 infection (Fig. 6d). After challenging the skin with live HSV-1, 17 we found vaccinated mice did not show typical symptoms of disease progression (n = 4) in contrast 18 to the mock-treated mice which showed mild zosteriform lesions at 2 d.p.i and hunched posture, 19 abnormal gait and severe zosteriform lesions at 6 d.p.i (Fig 6e). 20 To evaluate whether this vaccine blocked the transmission of HSV-1 from the skin to the peripheral 21 nervous system, the skin and dorsal root ganglion (DRG) samples were collected at the time of 22 euthanasia and examined for the HSV-1 genome. The viral load was significantly reduced in the 23 skin tissues of the vaccinated group by plaque assay and viral genome analysis, respectively (Fig.  24 6f-g). Remarkably, we found an almost undetectable level of viral loads in the DRG of vaccinated 25 mice by both assays indicating the strong neuronal protection by VLP mRNA vaccination (Fig. 6h-26 i). To get access into the tissue structure after vaccination and virus challenge, we conducted HE 27 staining of the skin which was found well preserved in the vaccinated group while the unvaccinated 28 mice showed thickened epidermis and seriously damaged dermis (Fig. 6j). Next, we performed 29 immunohistochemistry (IHC) to compare the local immune response in the skin of infected mice 30 and found apparent CD4+ cells enrichment, but not CD8+ cells, in the skin of unvaccinated mice 31 after the HSV-1 challenge. Additionally, a large number of neutrophils infiltrated the dermis of the 32 unvaccinated mice, which was not evident for vaccinated mice and non-infected controls. Taken 1 together, the DC-targeting VLP mRNA vaccine effectively protected mice from live HSV-1 2 infection. 3

Discussion 4
Currently, there are still no effective vaccines available for several infectious diseases such as HSV-5 1, HSV-2 and HIV and non-infectious diseases including most cancers. Dendritic cells are the most 6 potent antigen-presenting cells and an important cell type to induce effective and durably protective 7 T cell immunity as well as the humoral immune response to block pathogens or attack cancer 8 cells 39 . The clinically approved mRNA vaccines are based on LNP which is internalized passively 9 by diverse somatic cells including muscle cells, B cells, CD4+ T cells and tissue-resident or 10 recruited APCs 17 . The alternative is to deliver mRNA vaccine specifically to DCs. In this study, we 11 developed a DC-targeting mRNA vaccine platform by incorporating mRNA into VLP and 12 decorating its surface with an engineered Sindbis virus envelope. We found the DC-targeting VLP 13 mRNA vaccine induced durable IgG response and strong T cell immunity. Moreover, the VLP whether DC-targeting vaccines will be superior to non-specific vaccines. Our study showed DC-1 targeting VLP mRNA induced nearly 1 magnitude higher spike-specific IgG than the broad tropic 2 VSV-G pseudotyped counterpart. Although DC-specific VLP mRNA induced averagely higher 3 level of spike-specific T cell response, the difference was insignificant compared to VSV-G, 4 possibly due to VSV-G entering into DC very efficiently or depending on antigens 42 . 5 The future applications of the VLP-based mRNA vaccine include as an in situ DC vaccine to cure 6 cancers in combination with immune checkpoint inhibitors or being used a therapeutic vaccine to 7 remove the established viral infections such as HBV and HPV. Furthermore, the potency of VLP 8 mRNA vaccine may further be improved by combining circular RNA or self-amplifying RNA 9 which may extend the persistence of antigen expression and lower the necessary dose for 10 vaccination, therefore, improving the efficacy while downregulating the cost.

ELISPOT assay 31
To find the involvement of cellular immunity, the cytokine production in the splenic cells upon 1 treatment with spike peptides in vitro was measured. Spleens were removed aseptically, placed in 2 the RPM 1640 medium, gently homogenized, and passed through the cell strainer (Jet Bio-3 Filtration) to generate a single-cell suspension. Erythrocytes were rapidly washed and lysed by the 4 RBC lysis buffer (Sangon Biotech), and the splenocytes were resuspended in 1 mL RPMI 1640 5 medium. 5×10 5 splenocytes were seeded in anti-mouse IFN-γ, IL-6 and TNF-α antibody precoated 6 ELISPOT plates (Mabtech). Cells were incubated with a pool of SARS-CoV-2 spike peptides (15-7 mer peptides with 11-amino acid overlap covering the entire spike protein; GenScript) of 0.2 8 µg/well for each peptide for 36 h, or 2 μg/mL concanavalin A (ConA) (Sigma) and culture medium 9 as controls. The detection procedure was conducted according to the manufacturer's instructions. 10 Spots were counted and analyzed by using Mabtech IRIS FluoroSpot/ELISpot reader. 11 12

Neutralization assay 13
To determine the serum neutralization activity against GFP-expressing spike pseudovirus. Lung tissues were fixed in 4% formaldehyde and embedded in paraffin. Lung or skin sections were 26 stained with hematoxylin and eosin and analyzed for tissue status. For immunohistochemistry, the 27 sections were treated with 3% hydrogen peroxide for 25 min to block endogenous peroxidase 28 activity. The sections were then blocked with 3% BSA at room temperature for 30 min and 29 incubated with anti-CD4 (1:100; Servicebio, gb13064) or anti-CD8 (1:1,000; Servicebio, gb11068) 30 at 4 °C overnight. Then sections were then incubated with an anti-rabbit secondary antibody (1:500; 31 Servicebio, gb23303), followed by incubation with freshly prepared DAB substrate solution to 32 detect the antibody. Sections were counterstained with hematoxylin, blued with ammonia water, 1 and then dehydrated and coverslipped. 2 3

Plaque assay 4
To quantify infectious SARS-CoV-2 particles in lung, endpoint titrations were performed on 5 confluent Vero E6 cells. Lung homogenates were serially diluted in DMEM supplemented with 2% 6 FBS and 1% penicillin-streptomycin and incubated on cells for 2 h at 37℃. Then, the supernatants 7 were replaced with 1% low melting-point agar in DMEM with 2% FBS and 1% penicillin-8 streptomycin. The plates were inverted and incubated at 37°C for 3 days. Plates were fixed with 4% 9 PFA for 10 min at room temperature and then stained with 1ml 1% crystal violet for 1.5 hours.

Figure 1. Design and characterization of VLP-based SARS-CoV-2 mRNA vaccine. a, 2
Construction of mRNA-encoding plasmid which transcribes a MS2 stem loop-containing spike 3 mRNA. The spike mRNA and protein will be packaged into VLP via the RNA-coat protein 4 interaction and self-assembly, respectively. NTD, N-terminal domain; RBD, receptor binding 5 domain; SD1 and SD2, subdomain 1 and 2; FP, fusion peptide; HR1 and HR2, heptad repeat 1 and 6 2; TM, transmembrane domain; CT, cytoplasmic tail. b, Schematic illustration of the production 7 process of the SARS-CoV-2 vaccine using VLP platform. c, Electron microscopy image of VLP. 8 Scale bar, 100 nm. d, Copy number of spike mRNA in each VLP particle. The copy number was 9 detected by absolute quantification RT-qPCR and normalized to IDLV S-mut (2 copies RNA per 10 virion). e, Western blot analysis of the spike protein in the virion treated with/without PNGase F. images of spike peptide microarray. S1 protein and RBD were included in the microarray as 5 controls. Highly frequent positive peptides were labeled. b. Antibody responses against S1 protein 6 or RBD in vaccinated mice. Signal intensity was averaged fluorescent intensity of tinplated spots 7 for each array. ***P< 0.0001. c, Heatmap of antibody responses against peptides. The gray grid 8 indicates a negative response. d and e, Analysing the epitopes of VLP induced spike-specific 9 antibodies on spike protein. 6 mice were used for each group, 1.5 µg VLP S-mut or 50 µL PBS 10 were injected via footpad into each mouse. Data and error bars represent mean ± s.e.m.; unpaired 11 two-tailed student's t-tests. immunization. Mice were challenged with 10 5 TCID50 of SARS-CoV-2 at 14 days post boost 1 immunization by intranasal administration. All mice were euthanized at 3 d.p.i. b, Neutralization 2 activity of vaccinated sera against live SARS-CoV-2 (USA-WA1/2020). c, The percentage of mice 3 weight change after infection. *P = 0.0253. d and e, Viral loads in lung and trachea detected by RT-4 qPCR. f, Confocal analysis of SARS-CoV-2 in the lung. **P = 0.0042 (d) and ***P < 0.0001 (c). 5 g, Lung histopathology analysis by hematoxylin and eosin (red arrow, inflammatory cell 6 infiltration; blue arrow, alveolar destruction). Each image is a representative of a group of 4 mice (f 7 and g). Data and error bars represent mean ± s.e.m.; unpaired two-tailed student's t-tests (c-e). Design and characterization of VLP-based SARS-CoV-2 mRNA vaccine. a, Construction of mRNAencoding plasmid which transcribes a MS2 stem loop-containing spike mRNA. The spike mRNA and protein will be packaged into VLP via the RNA-coat protein interaction and self-assembly, respectively. VLP mRNA induces robust and durable spike-speci c antibody responses. a, Schematic illustration of the working plan (n=5). The sera were collected 14 days after footpad VLP injection for further analysis. b, ELISA analysis of spike speci c IgG. ***P< 0.0001. c-f, Neutralization activity of vaccinated sera evaluated by luciferase assay (c and f), confocal microscopy (d) and plaque assay(e). A re y luciferaseencoding pseudovirus, GFP-expressing SARS-CoV-2 pseudovirus and live SARS-CoV-2 (USA-WA1/2020) was used, respectively, to transduce Huh-7 or Vero E6 cells. *P= 0.0260 (c). Images are representative of three independent biological replicates in one experiment (d). g and h, Antibody changes in short-term (g) and long-term (h) follow-up vaccination. Mice were immunized with 1.5 μg VLP S-mut via footpad injection, sera were collected at the indicated time for IgG ELISA. Data and error bars represent mean ± s.e.m.; unpaired two-tailed student's t-tests (b and c); two-tailed Wilcoxon matched-pairs 1 signed-rank test (f).

Figure 3
Linear epitope landscape in the VLP mRNA vaccinated mice. a, Representativ images of spike peptide microarray. S1 protein and RBD were included in the microarray as controls. Highly frequent positive peptides were labeled. b. Antibody responses against S1 protein or RBD in vaccinated mice. Signal intensity was averaged uorescent intensity of tinplated spots for each array. ***P< 0.0001. c, Heatmap of antibody responses against peptides. The gray grid indicates a negative response. d and e, Analysing the epitopes of VLP induced spike-speci c antibodies on spike protein. 6 mice were used for each group, 1.5 μg VLP S-mut or 50 μL PBS were injected via footpad into each mouse. Data and error bars represent mean ± s.e.m.; unpaired two-tailed student's t-tests.