A COVID-19 mRNA vaccine encoding SARS-CoV-2 virus-like particles induces a strong antiviral-like immune response in mice

Dear Editor, Since the beginning of this century, humanity has been struck three times by the coronavirus outbreak. The most recent one is caused by the SARS-CoV-2 virus, which was first reported in January 2020 and spread rapidly worldwide, developing into a global coronavirus disease pandemic coded COVID-19. By July 28, 2020, SARS-CoV-2 has caused over sixteen million COVID-19 cases worldwide and 650,805 deaths. Such a grave situation has made the development of a COVID-19 vaccine imperative and urgent. In this study, we designed three mRNA vaccine candidates for COVID-19, and they encode various forms of antigens in vaccinated hosts (Fig. 1a). RQ3011-RBD encodes the receptorbinding domain of the S (spike) glycoprotein (residues 331–524) of SARS-CoV-2 with an N-terminal signal peptide and a C-terminal membrane-anchoring helix. Vaccine RQ3012-Spike encodes the full-length wild-type S, while RQ3013-VLP is formulated from a cocktail of mRNAs encoding three structural proteins: S, M (membrane), and E (envelope) to form SARS-CoV2 virus-like particles (VLPs). To increase the expression capacity of mRNA vaccines, all mRNAs were subjected to an in-depth sequence optimization procedure of two parameters: codons in the DNA template and modified nucleotides incorporated into mRNA. We designed ten coding sequences of the S gene (3822 nucleotides in length) with varying GC-content, maintaining the maximum codon adaptation index. For each DNA template, we tested six mRNA species with various modified nucleotides. The mRNA candidates (total of 60) displayed a considerable variation in their abilities to express S in HEK 293A cells (Supplementary information, Fig. S1a). Notably, the incorporation of pseudouridine consistently improves the expression of S, regardless of the codon sequence used. For M and E, which are relatively small proteins, we designed one codon-optimized sequence for each and screened for the optimal choice of modified nucleotides. The final mRNAs in vaccines have an optimal combination of codon and modified nucleotides that give the most robust expression (Fig. 1c, d). Previous studies on SARS and MERS have shown that coronavirus VLP assembly requires at least three structural proteins: S, M, and E. Based on our established system, we co-transfected three mRNAs encoding SARS-CoV-2 S, M, and E at a molar ratio of 1:2:2 into cells. All three proteins can be detected by western blotting in culture media. We then purified VLPs through a sucrose gradient and examined the particles under an electron microscope. The VLPs have an average diameter of 100 nm, with the spike protein densely decorating the surface, suggesting that SARS-CoV-2 virus-like particles have formed. We used the well-established lipid nanoparticles (LNPs) to package mRNAs. The mRNA encapsulation efficiency of all three LNP vaccine candidates was greater than 98%, with an average size of 100 nm in diameter (Supplementary information, Fig. S1b, c). All LNPs were able to transfect HEK 293A cells and express antigens of interest, as judged by western blotting (Supplementary information, Fig. S1d). Virus-like particles secreted into culture media of cells transfected with RQ3013-VLP can be detected by western blotting and electron microscopy (Fig. 1b). We then assessed the immunogenicity of each mRNA LNP vaccine candidate in BALB/c mice. A group of mice (n= 10) were immunized intramuscularly with each of the vaccines on day 0 (Fig. 1e). Each dose of vaccine contains 2 μg of RBD mRNA for RQ3011-RBD or 6 μg of S mRNA for RQ3012-Spike. For RQ3013VLP, each treatment delivers 6 μg of S, 2.5 μg of M, and 1.5 μg of E mRNAs. A fourth control group of mice (n= 10) were included in the study, for which 22 μg empty LNP was used as placebo. All groups were boosted on day 21, 3 weeks after the prime injection. No inflammation or other adverse effects were observed at the sites of injection. Sera were collected on days 20 (week 3), 28 (week 4), 42 (week 6), and 56 (week 8). All sera were evaluated for binding to the S ectodomain by enzyme-linked immunosorbent assay (ELISA). Binding antibodies can be detected in mice immunized with RQ3012-Spike and RQ3013-VLP on days 20 after the first injection, while RQ3011-RBD showed marginal stimulation (Fig. 1f). Following a boost, the antibody titers increased dramatically in mice receiving RQ3012Spike or RQ3013-VLP, and peaked at week 3, remaining stable at week 8. A boost did not increase the titer for RQ3011-RBD, which dropped to the level of the placebo group. Notably, mice receiving RQ3013-VLP had the strongest immune response and developed significantly higher titers of S-specific binding antibody than mice receiving RQ3012-Spike. Since we included M and E mRNA in RQ3013-VLP, we analyzed whether M and E induced protein-specific immunoglobulin G (IgG). For that purpose, sub-VLPs consisting of M and E proteins were purified and used for ELISA. No Mor E-specific antibodies were detected in mice vaccinated with RQ3013-VLP (Supplementary information, Fig. S2c). The presence of neutralizing antibodies (NAbs) was evaluated for all groups using our recently established pseudovirus neutralization assay for SARS-CoV-2. We and others have previously demonstrated that NAb titers measured from the vesicular stomatitis virus (VSV) pseudovirus assay correlated well with NAb titers measured from a live SARS-CoV-2 virus assay. In mice receiving RQ3012-Spike, the mean NAb titers (EC50) reached 10,000 at week 4, 1 week after a boost, and peaked at week 6, maintaining relatively stable at week 8 (Fig. 1g and Supplementary information, Fig. S2a, b). In mice receiving RQ3013-VLP, the mean NAb titer rose to 25,028 at week 4, 2.5-fold higher than that in the RQ3012-Spike group. By week 8, the NAb titer was still increasing, with the highest EC50 value of more than 100,000. The differences in NAb titers between RQ3012-Spike and RQ3013-VLP are significant throughout the tested weeks (P= 0.0021 at week 4, P= 0.0042 at week 6, P= 0.0015 at week 8). With


Protein expression of mRNAs and LNPs
The protein expression of mRNAs was tested in HEK 293A cells. mRNA transfection was carried out using lipofectamine 2000 (lipo2K) at a ratio of 1:2 (mRNA: lipo2K). For expression of virus-like particles (VLPs), S, M, and E mRNAs were co-transfected into HEK 293A cells at various molar ratios, and the supernatant was collected 48-hour post-transfection for analysis. For LNP，20µg of LNPs were incubated with HEK 293A cell in one well of a six-well plate for 48 hours, and cells and media were collected for protein analysis. Protein expression was detected by western blotting. Briefly, cells were collected and rinsed with PBS and lysed by lysis buffer (20mM Tris-HCl(pH=7.4), 150mM NaCl, 3mM MgCl2, 1% Triton X-100) supplemented with protease inhibitor. For VLP detection, the culture media was subjected to ultracentrifugation through a 20% sucrose cushion. The pellet was resuspended in PBS. All samples were mixed with SDS loading buffer and separated in a 4-20% gradient SDS-Gel, and transferred to PVDF membranes (ThermoFisher) by Trans-Blot Turbo Transfer System (BioRad). The blots were blocked with 5% non-fat dry milk in TBST and then incubated with appropriate primary antibodies. Signals were detected with HRP-conjugated secondary antibodies and an enhanced chemiluminescence (ECL) detection system.

Negative-stain electron microscopy of VLPs
For VLP purification, cell culture media after 36-hour post-transfection were concentrated using 100-kDa cutoff concentrator (Amicon Ultra-15, Millipore) before being layered on the top of the 30-40-50% (w/v) sucrose gradient in 20 mM HEPES-Na (pH 7.4), 120 mM NaCl. The sucrose solution between 30-40% (w/v) was extracted immediately with a 5 mL syringe after ultracentrifugation for 90 minutes at 4°C, 31,000 rpm (rotor SW32, Beckman). The VLPscontaining solution was buffer-exchanged against PBS three times. To prepare grids for negative-stain transmission electron microscopy (TEM), 5 µL purified VLPs were absorbed for 2 minutes on a glow-discharged carbon-coated grid. The grid was stained in a dropwise manner for 60 seconds and loaded on the Talos L120C microscope (ThermoFisher) for visualization of VLPs.

Mice immunization
All animal experiments were performed under the ethical guidelines of Fudan University. Sixweek-old female BALB/c mice (n=10 for each group) were immunized intramuscularly with 100 µL of LNP vaccine candidates or placebo at week 0. At week 3, all mice received a boost vaccination. Sera were collected from all the mice on days 0, 20, 28, 42, and 56 for analysis of binding and NAb antibody responses.

ELISA for SARS-CoV-2 S-specific IgG
SARS-CoV-2 S-specific antibody responses in immunized sera were determined by enzymelinked immunosorbent assay (ELISA) assay, as previously described. Briefly, 96-well plates were coated with 100 µL of coating buffer containing 100ng/well Spike S1+S2 ECD-His recombinant protein (Sino Biological) at 4°C overnight. Plates were blocked with 2% bovine serum albumin solution in PBST at room temperature for 1 hour. Immunized mice sera were diluted 100-fold as the initial concentration, and then a 5-fold serial dilution of a total of 11 gradients in PBS buffer. PBST washed plates were incubated with serially diluted sera at room temperature for 2 hours. For determination of S-specific antibody response, plates were incubated with goat anti-mouse IgG HRP (Proteintech Cat: SA00001-1) at 37°C for 1 hour and then substrate tetramethylbenzidine (TMB) solution (Invitrogen) was used to develop. The color reaction was quenched with 1N sulfuric acid at about 10 minutes, and the optical density was measured at wavelength 450 nm by Synergy H1 microplate reader (BioTek).

ELISA for SARS-CoV-2 M and E-specific IgG
To prepare antigens for ELISA, we cloned SARS-CoV-2 M and E genes fused by a T2A sequence into the pCDNA3.1 vector. HEK293F cells were used to express proteins. Secreted particles formed by M and E proteins were collected from cell culture by pelleting via ultracentrifugation through a 20% sucrose cushion and further purified through a 30-50% sucrose gradient. Purified M/E particles (6 µg/well) were added to 96-well plates and incubated at 4°C overnight. The ELISA was developed and measured as described above.

Pseudovirus-based neutralization assay
The pseudovirus-based neutralization assay was performed as described 1 . BHK21-hACE2 cells were seeded in 96-well plates. Mice sera were mixed with diluted VSV-SARS-CoV-2-Sdel18 virus (MOI=0.05) and incubated at 37°C for 1h. All samples and viruses were diluted with DMEM (10%FBS), the mixture was added to seeded BHK21-hACE2 cell. After 12h incubation, fluorescence images were obtained by Opera Phenix or Operetta CLS equipment (PerkinElmer). For quantitative determination, fluorescence images were analyzed by Columbus system (PerkinElmer), and the numbers of GFP-activated cells for each well were counted to represent infection performance. The reduction (%) of mice sera treatment GFPactivated cell numbers in comparison with non-treated control well was calculated to show the neutralizing potency.

T cell restimulation assay
Blood was collected through retro-orbital bleeding with heparinized capillary tubes (Kimble Chase). Blood cells were pelleted by centrifuging at 4,000 rpm for 20 minutes at 4°C and resuspended in 1 mL RBC lysis buffer (ThermoFisher). After incubation at room temperature for 5 minutes, the cell suspension was transferred into 10 ml FACS (DPBS supplemented with 0.2% BSA, 2mM EDTA) and spun down at 400 g for 10 minutes at 4 °C. Fresh PBMCs were suspended in FACS, blocked with anti-CD16/32 antibody (BD Biosciences) at 4 °C for 30 minutes and stained with ZombieNIR Fixable Viability Dye (Biolegend) as well as antibodies against mouse CD45 (Biolegend), TCRb (Biolegend), CD4 (ThermoFisher), CD8a (Biolegend) and CD44 (Biolegend) at 4 °C for 30 minutes. Cells were then washed twice with FACS before analyzed on a Gallios flow cytometer (Beckman).

Statistical Analysis
All data was graphed, and statistics performed using GraphPad Prsm8.0. The EC50 values were calculated by non-linear regression. Statistical analyses were carried out by Student's t-test when two groups were analyzed, and by ANOVA when more than two groups were analyzed.