Main

Since its emergence in December 2019, SARS-CoV-2 has accounted for more than 30 million cases of coronavirus disease 2019 (COVID-19) worldwide in 9 months3. SARS-CoV-2 is the third novel Betacoronavirus in the past 20 years to cause substantial human disease; however, unlike its predecessors SARS-CoV and Middle East respiratory syndrome coronavirus (MERS-CoV), SARS-CoV-2 is transmitted efficiently from person to person. In the absence of a vaccine, public health measures such as quarantine of newly diagnosed cases, contact tracing, use of face masks and physical distancing have been put into place to reduce transmission4. It is estimated that until 60–70% of the population have immunity, COVID-19 is unlikely to be sufficiently well-controlled for normal human activities to resume. If immunity remains solely dependent on infection, even at a case fatality rate of 1%, more than 40 million people could succumb to COVID-19 globally5. Therefore, rapid development of vaccines against SARS-CoV-2 will be critical for changing the global dynamics of this virus.

The spike (S) protein, a class I fusion glycoprotein analogous to influenza haemagglutinin, respiratory syncytial virus (RSV) fusion glycoprotein (F) and human immunodeficiency virus gp160 (Env), is the major surface protein on the coronavirus virion and the primary target for neutralizing antibodies. S proteins undergo marked structural rearrangement to fuse virus and host cell membranes, enabling delivery of the viral genome into target cells. We previously showed that prefusion-stabilized protein immunogens that preserve neutralization-sensitive epitopes are an effective vaccine strategy for enveloped viruses such as RSV6,7,8,9,10. Subsequently, we identified 2 proline substitutions (2P) at the apex of the central helix and heptad repeat 1 that effectively stabilized MERS-CoV, SARS-CoV and human coronavirus HKU1 S proteins in the prefusion conformation1,11,12. Similar to other prefusion-stabilized fusion proteins, MERS-CoV S(2P) protein was more immunogenic at lower doses than wild-type S protein1. The 2P mutation has similar effects on the stability of S proteins from other betacoronaviruses, suggesting a generalizable approach for designing stabilized-prefusion Betacoronavirus S protein antigens for vaccination. Such generalizability is fundamental to the prototype pathogen approach for pandemic preparedness13,14.

Coronaviruses have long been predicted to have a high probability of causing zoonotic disease and pandemics15,16. As part of our pandemic preparedness efforts, we have studied MERS-CoV as a prototype Betacoronavirus pathogen to optimize vaccine design, dissect the humoral immune response to vaccination, and identify mechanisms and correlates of protection. Achieving an effective and rapid vaccine response to a newly emerging virus requires both the precision afforded by structure-based antigen design and a manufacturing platform to shorten time to product availability. Producing cell lines and clinical-grade subunit protein typically takes more than one year, whereas manufacturing nucleic acid vaccines can be achieved in a matter of weeks17,18. In addition to advantages in manufacturing speed, mRNA vaccines are potently immunogenic and elicit both humoral and cellular immunity19,20,21. We therefore evaluated mRNA formulated in lipid nanoparticles (mRNA–LNP) as a delivery vehicle for MERS-CoV S(2P), and found that transmembrane-anchored MERS-CoV S(2P) mRNA elicited more potent pseudovirus-neutralizing antibody responses than secreted MERS-CoV S(2P) (Extended Data Fig. 1a). Additionally, consistent with protein immunogens, MERS-CoV S(2P) mRNA was more immunogenic than wild-type MERS-CoV S mRNA (Extended Data Fig. 1b). Immunization with MERS-CoV S(2P) mRNA–LNP elicited potent pseudovirus-neutralizing activity with a dose as low as 0.1 μg and protected transgenic mice expressing human DPP4 (288/330+/+)22 against lethal MERS-CoV challenge in a dose-dependent manner, establishing that mRNA encoding S(2P) protein is protective. Notably, a subprotective 0.01 μg dose of MERS-CoV S(2P) mRNA did not cause exaggerated disease following MERS-CoV infection, but instead resulted in partial protection against weight loss followed by full recovery without evidence of enhanced illness (Fig. 1).

Fig. 1: MERS-CoV S-2P mRNA protects mice from lethal challenge.
figure 1

a–d, 288/330+/+ mice were immunized at weeks 0 and 3 with 0.01 (green), 0.1 (blue) or 1 μg (red) MERS-CoV S(2P) mRNA. Control mice were administered phosphate-buffered saline (PBS) (grey). a, Two weeks post-boost, sera were collected from three mice per group and assessed for neutralizing antibodies against MERS m35c4 pseudovirus. b–d, Four weeks post-boost, 12 mice per group were challenged with a lethal dose of mouse-adapted MERS-CoV (m35c4). b, Following challenge, mice were monitored for weight loss. c, d, Two days post-challenge, at peak viral load, lung viral titres (c) and haemorrhage (scored as: 0, no haemorrhage, 4, severe haemorrhage in all lobes) (d) were assessed from five mice per group. In c, d, all dose levels were compared by Kruskal–Wallis analysis of variance (ANOVA) with Dunn’s multiple comparisons test. In b, for weight loss, all comparisons are with PBS control mice at the same time point by two-sided Mann–Whitney U-test. **P < 0.01, ****P < 0.0001. Data are GMT ± geometric s.d. (a, c) or mean ± s.d. (b, d). In c, the dotted line represents assay limit of detection.

Source data

Full size image

SARS-CoV-2 was first identified as the cause of an outbreak of respiratory disease in Wuhan, China in early January 2020. Within 24 h of the release of genomic sequences of SARS-CoV-2 isolates on 10 January 2020, the 2P mutations were substituted into S protein residues 986 and 987 to produce prefusion-stabilized SARS-CoV-2 S(2P) protein for structural analysis23 and serological assay development24,25 in silico, without additional experimental validation. Within 5 days of the release of the sequence, current good manufacturing practice (cGMP) production of mRNA–LNP encoding the SARS-CoV-2 S(2P) as a transmembrane-anchored protein with the native furin cleavage site (mRNA-1273) was initiated in parallel with preclinical evaluation. This led to a first-in-human phase I clinical trial starting on 16 March 2020, 66 days after the viral sequence was released, and a phase II trial 74 days later on 29 May 2020 (Extended Data Fig. 2). Expression and antigenicity of the S(2P) antigen delivered by mRNA was confirmed in vitro before vaccination of the first human participant (Extended Data Fig. 3), and immunogenicity of mRNA-1273 was documented in several mouse strains. The results of those studies are detailed here.

Immunogenicity was assessed in six-week-old female BALB/cJ, C57BL/6J and B6C3F1/J mice by two intramuscular immunizations with 0.01, 0.1 or 1 μg mRNA-1273, separated by a 3-week interval. mRNA-1273 induced dose-dependent specific S-binding antibodies after prime and boost in all mouse strains (Fig. 2a–c). Potent pseudovirus-neutralizing activity was elicited by 1 μg mRNA-1273, reaching reciprocal half-maximal inhibitory concentration (IC50) geometric mean titres (GMTs) of 819 (BALB/cJ), 89 (C57BL/6J) and 1,115 (B6C3F1/J) (Fig. 2d–f). Additionally, mice immunized with 1 μg mRNA-1273 had robust neutralizing antibodies against pseudoviruses that express S protein with the D614G substitution; SARS-CoV-2 expressing the D614G variant of the S protein has recently become dominant around the world2 (Extended Data Fig. 4). To further gauge immunogenicity across a wide dose range, BALB/c mice were immunized with 0.0025–20 μg mRNA-1273, revealing a strong positive correlation between dose-dependent mRNA-1273-elicited binding and pseudovirus-neutralizing antibody responses (Extended Data Fig. 5). BALB/cJ mice that received a single dose of mRNA-1273 were evaluated to ascertain the utility of a single-dose vaccine. S-binding antibodies were induced in mice immunized with one 1 μg or 10 μg dose of mRNA-1273. The 10 μg dose elicited pseudovirus-neutralizing antibody activity that increased between week 2 and week 4, reaching 315 reciprocal IC50 GMT (Extended Data Fig. 6a, b). These data demonstrate that mRNA expressing SARS-CoV-2 S(2P) is a potent immunogen and that pseudovirus-neutralizing activity can be elicited with a single dose.

Fig. 2: mRNA-1273 elicits robust binding and pseudovirus-neutralizing antibody responses in multiple mouse strains.
figure 2

a–f, BALB/cJ (a, d), C57BL/6J (b, e) or B6C3F1/J (c, f) mice (n = 10 per group) were immunized at weeks 0 and 3 with 0.01 (green), 0.1 (blue) or 1 μg (red) mRNA-1273. Control BALB/cJ mice were administered PBS (grey). Sera were collected 2 weeks post-prime (unfilled circles) and 2 weeks post-boost (filled circles) and assessed for SARS-CoV-2 S-specific IgG by enzyme-linked immunosorbent assay (ELISA) (a–c), and for post-boost sera, neutralizing antibodies against homotypic SARS-CoV-2 pseudovirus (d–f). In a–c, time points were compared within each dose level by two-sided Wilcoxon signed-rank test, and doses were compared post-boost by Kruskal–Wallis ANOVA with Dunn’s multiple comparisons test. In d–f, Vaccine groups were compared by two-sided Mann–Whitney U-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Data are presented as GMT ± geometric s.d. Dotted lines represent assay limits of detection.

Source data

Full size image

Vaccine-associated enhanced respiratory disease (VAERD) has been associated with T helper 2 cell (TH2)-biased immune responses in children immunized with whole-inactivated-virus vaccines against RSV and measles virus26,27. A similar phenomenon has also been reported in some animal models with whole-inactivated vaccines and other types of experimental SARS-CoV vaccines28,29,30. We therefore evaluated the balance of TH1 and TH2 cells in immunized mice. We first compared levels of S-specific immunoglobulins, IgG2a and IgG2c, and IgG1—which are surrogates of TH1 and TH2 responses, respectively—elicited by mRNA-1273 with those elicited by immunization with SARS-CoV-2 S(2P) protein using the TLR4 agonist Sigma Adjuvant System (SAS). Both immunogens elicited S-binding antibodies in the IgG2a and IgG1 subclasses, indicating a balanced TH1–TH2 response (Fig. 3a–c, Extended Data Fig. 7). The S-specific IgG-subclass profile following a single dose of mRNA-1273 (Extended Data Fig. 6c) was similar to that observed following two doses. By contrast, TH2-biased responses, with lower IgG2a/IgG1 ratios, were observed in mice immunized with SARS-CoV-2 S(2P) protein formulated in alum (Extended Data Fig. 8a, b). Following restimulation with peptide pools (one pool of overlapping peptides for each S subunit, S1 and S2) covering the entire S protein, splenocytes from mice immunized with mRNA-1273 secreted more IFN-γ (a prototypic TH1 cytokine) than IL-4, IL-5 or IL-13 (classical TH2 cytokines), whereas restimulation with SARS-CoV-2 S(2P) protein with alum adjuvant induced a TH2-biased response (Extended Data Fig. 8c, d). We also directly measured cytokine patterns in vaccine-induced memory T cells by intracellular cytokine staining seven weeks after the boost injection; mRNA-1273-elicited CD4+ T cells re-stimulated with S1 or S2 peptide pools exhibited a TH1-dominant response, particularly at higher immunogen doses (Fig. 3d, e). Furthermore, 1 μg mRNA-1273 induced a robust CD8+ T cell response to the S1 peptide pool (Fig. 3f, g). Together, the IgG subclass and T cell cytokine data demonstrate that immunization with mRNA-1273 elicits balanced TH1 and TH2 responses, in contrast to the TH2-biased response seen when using S protein with alum adjuvant, suggesting that mRNA vaccination avoids TH2-biased immune responses, which have been linked to VAERD.

Fig. 3: Immunizations with mRNA-1273 and S(2P) protein, delivered with TLR4 agonist, elicit S-specific TH1-biased T cell responses.
figure 3

B6C3F1/J mice (n = 10 per group) were immunized at weeks 0 and 3 with 0.01, 0.1 or 1 μg of mRNA-1273 or SARS-CoV-2 S(2P) protein with SAS adjuvant. a–c, Sera were collected two weeks post-boost and assessed by ELISA for SARS-CoV-2 S-specific IgG1, and IgG2a and IgG2c. End-point titres (a, b) and end-point titre ratios of IgG2a plus IgG2c to IgG1 (c) were calculated. Ratios were not calculated for mice for which end-point titres did not reach the lower limit of detection (dotted line; N/A). d–g, Seven weeks post-boost, splenocytes were isolated from five mice per group and restimulated with vehicle or pools of overlapping peptides from SARS-CoV-2 S protein in the presence of a protein transport inhibitor cocktail. After 6 h, intracellular cytokine staining was performed to quantify CD4+ and CD8+ T cell responses. Cytokine expression in the presence of vehicle only was considered as background and subtracted from the responses measured from the S1 and S2 peptide pools for each individual mouse. d, e, Percentage of CD4+ T cells expressing IFN-γ, TNF, IL-2, IL-4 and IL-5 in response to the S1 (d) and S2 (e) peptide pools. f, g, Percentage of CD8+ T cells expressing IFN-γ, TNF and IL-2 in response to the S1 (f) and S2 (g) peptide pools.

Source data

Full size image

Protective immunity was assessed in young adult BALB/cJ mice challenged with mouse-adapted (MA) SARS-CoV-2. SARS-CoV-2 MA contains the substitutions Q498Y/P499T in the receptor-binding domain31. The substitutions enable the virus to bind to the mouse angiotensin-converting enzyme 2 (ACE2) receptor and infect and replicate in the upper and lower respiratory tract32. BALB/cJ mice that received two 1 μg doses of mRNA-1273 were completely protected from viral replication in lungs after challenge 5 or 13 weeks after boost injection (Fig. 4a, Extended Data Fig. 9a). mRNA-1273-induced immunity also resulted in undetectable viral replication in nasal turbinates in 6 out of 7 mice (Fig. 4b, Extended Data Fig. 9b). The efficacy of mRNA-1273 was dose-dependent; two 0.1 μg doses of mRNA-1273 reduced lung viral load by about 100-fold, whereas two 0.01 μg doses reduced lung viral load by about 3-fold (Fig. 4a). Of note, mice challenged 7 weeks after a single dose of 1 or 10 μg mRNA-1273 were also completely protected against lung viral replication (Fig. 4c). Challenging animals immunized with subprotective doses provides an orthogonal assessment of safety signals such as increased clinical illness or pathology. Similar observations with MERS-CoV S(2P) mRNA, mice immunized with subprotective 0.1 or 0.01 μg doses of mRNA-1273 showed no evidence of enhanced lung pathology or excessive mucus production (Fig. 4d). In summary, mRNA-1273 is immunogenic, efficacious and does not produce evidence of VAERD when given at subprotective doses in mice.

Fig. 4: mRNA-1273 protects mice from upper- and lower-airway SARS-CoV-2 infection.
figure 4

a, b, BALB/cJ mice (n = 10 per group) immunized at weeks 0 and 3 with 0.01 μg (green), 0.1 μg (blue) or 1 μg (red) mRNA-1273 or PBS were challenged with SARS-CoV-2 MA five weeks post-boost. c, Other groups were immunized with single doses of 0.1 μg (blue), 1 μg (red) or 10 μg (purple) mRNA-1273 and challenged 7 weeks after immunization. Two days after challenge, at peak viral load, mouse lungs (a, c) and nasal turbinates (b) were collected from five mice per group to measure viral titres. a–c, Data are presented as GMT ± geometric s.d. and dotted lines represent assay limits of detection. Group comparisons were made by Kruskal–Wallis ANOVA with Dunn’s multiple comparisons test. **P < 0.01, ***P < 0.001. d, At days 2 and 4 after challenge, lung sections from 5 mice per group were stained with haematoxylin and eosin, and representative photomicrographs (original magnification ×4 (scale bars, 600 µm) and ×10 (scale bars, 300 µm) as indicated) from each group with detectable virus in lung are shown. Day 2 lungs from PBS control mice demonstrated moderate-to-severe, predominantly neutrophilic inflammation present within and surrounding small bronchioles (arrowheads); alveolar capillaries were markedly expanded by infiltrating inflammatory cells. In the 0.01 μg two-dose group, inflammation was minimal to absent. In the 0.1 μg two-dose group, occasional areas of inflammation intimately associated with small airways (bronchioles) and adjacent vasculature (arrowheads) were seen, primarily composed of neutrophils. In the single-dose 0.1 μg group, there were mild patchy expansions of alveolar septae by mononuclear and polymorphonuclear cells. At day 4, lungs from PBS control mice exhibited moderate-to-marked expansion of alveolar septae (interstitial pattern) with decreased prominence of adjacent alveolar spaces. In the 0.01 μg two-dose group, inflammation was minimal to absent. Lungs in the 0.1 μg two-dose group showed mild, predominantly lymphocytic inflammation, associated with bronchioles and adjacent vasculature (arrowheads). In the single-dose 0.1 μg group there was mild, predominantly lymphocytic inflammation around bronchovascular bundles (arrowheads).

Source data

Full size image

Here we have shown that 1 μg of mRNA-1273 is sufficient to induce robust pseudovirus-neutralizing activity and CD8 T cell responses, balanced TH1–TH2 antibody isotype responses, and protection from viral replication for more than three months following a prime–boost regimen similar to the one being tested in humans. The level of pseudovirus-neutralizing activity induced by 1 μg mRNA-1273 in mice is similar in magnitude to that induced by 100 μg mRNA-1273 in humans33, which is the dose selected for mRNA-1273 to advance into phase III clinical trials. The inclusion of lower subprotective doses demonstrates the dose-dependence of antibody, TH1 CD4 T cell responses and protection, suggesting that immune correlates of protection can be further elucidated. Animal studies supporting candidate SARS-CoV-2 vaccines through clinical trials aim to demonstrate elicitation of potent protective immune responses as well as to show that subprotective responses do not cause VAERD5. Subprotective doses of mRNA-1273 did not prime mice for enhanced immunopathology following challenge. Moreover, the induction of protective immunity following a single dose suggests single-dose administration of this vaccine could be considered in the outbreak setting. These data, combined with immunogenicity data from non-human primates and human participants of early phase I clinical trials, have been used to inform the dose and regimen of mRNA-1273 in advanced clinical efficacy trials.

The COVID-19 pandemic of 2020 is the widely predicted ‘pathogen X event’13,14. Here we provide a paradigm for rapid vaccine development. Combining structure-guided stabilization of the MERS-CoV S protein with a fast, scalable and safe mRNA–LNP vaccine platform has led to a generalizable vaccine solution for Betacoronavirus and a commercial mRNA vaccine delivery platform; these developments enabled a rapid response to the COVID-19 outbreak. This response demonstrates how new technology-driven concepts such as synthetic vaccinology can facilitate a vaccine development programme initiated on the basis of pathogen sequences alone11. This study also provides a proof of concept for the prototype-pathogen approach to pandemic preparedness and response that is predicated on identifying generalizable solutions for medical counter measures within virus families or genera13. Although the response to the COVID-19 pandemic has been unprecedented in its speed and breadth, we envision further improvements in rapid responses to such threats. There are 24 other virus families that are known to infect humans, and sustained investigation of those potential threats will improve our readiness for future pandemics14.

Methods

Data reporting

No statistical methods were used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Pre-clinical mRNA-1273 mRNA and LNP production process

A sequence-optimized mRNA encoding SARS-CoV-2 S(2P) protein was synthesized in vitro using an optimized T7 RNA polymerase-mediated transcription reaction with complete replacement of uridine by N1-methyl-pseudouridine34. The reaction included a DNA template containing the immunogen open reading frame flanked by 5′ untranslated region (UTR) and 3′ UTR sequences and was terminated by an encoded polyA tail. After transcription, the Cap 1 structure was added to the 5′ end using vaccinia capping enzyme (New England Biolabs) and Vaccinia 2′ O-methyltransferase (New England Biolabs). The mRNA was purified by oligo-dT affinity purification, buffer exchanged by tangential flow filtration into sodium acetate, pH 5.0, sterile filtered, and kept frozen at –20 °C until further use.

The mRNA was encapsulated in a lipid nanoparticle through a modified ethanol-drop nanoprecipitation process as described previously20. In brief, ionizable, structural, helper and polyethylene glycol lipids were mixed with mRNA in acetate buffer, pH 5.0, at a ratio of 2.5:1 (lipids:mRNA). The mixture was neutralized with Tris-Cl pH 7.5, sucrose was added as a cryoprotectant, and the final solution was sterile filtered. Vials were filled with formulated LNP and stored frozen at –70 °C until further use. The drug product underwent analytical characterization, which included the determination of particle size and polydispersity, encapsulation, mRNA purity, double stranded RNA content, osmolality, pH, endotoxin and bioburden, and the material was deemed acceptable for in vivo study.

MERS-CoV and SARS-CoV protein expression and purification

Vectors encoding MERS-CoV S-2P1 and SARS-CoV S-2P23 were generated as previously described with the following small amendments. Proteins were expressed by transfection of plasmids into Expi293 cells using Expifectamine transfection reagent (ThermoFisher) in suspension at 37 °C for 4–5 days. Transfected cell culture supernatants were collected, buffer exchanged into 1× PBS, and protein was purified using Strep-Tactin resin (IBA). For proteins used for mouse inoculations, tags were cleaved with addition of HRV3C protease (ThermoFisher) (1% wt/wt) overnight at 4 °C. Size-exclusion chromatography using Superose 6 Increase column (GE Healthcare) yielded final purified protein.

Design and production of recombinant minifibritin foldon protein

A mammalian codon-optimized plasmid encoding foldon inserted minifibritin (ADIVLNDLPFVDGPPAEGQSRISWIKNGEEILGADTQYGSEGSMNRPTVSVLRNVEVLDKNIGILKTSLETANSDIKTIQEAGYIPEAPRDGQAYVRKDGEWVLLSTFLSPALVPRGSHHHHHHSAWSHPQFEK) with a C-terminal thrombin cleavage site, 6×His tag, and Strep-TagII was synthesized and subcloned into a mammalian expression vector derived from pLEXm. The construct was expressed by transient transfection of Expi293 (ThermoFisher) cells in suspension at 37 °C for 5 days. The protein was first purified with a Ni2+-nitrilotriacetic acid resin (GE Healthcare) using an elution buffer consisting of 50 mM Tris-HCl, pH 7.5, 400 mM NaCl and 300 mM imidazole pH 8.0, followed by purification with StrepTactin resin (IBA) according to the manufacturer’s instructions.

Cell lines

HEK293T/17 (ATCC CRL-11268), Vero E6 (ATCC), Huh7.5 cells (provided by D. R. Taylor, US Food and Drug Administration) and ACE2-expressing 293T cells (provided by M. Farzan, Scripps Research Institute) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS, 2 mM glutamine and 1% penicillin–streptomycin at 37 °C and 5% CO2. Vero E6 cells used in plaque assays to determine lung and nasal turbinate viral titres were cultured in DMEM supplemented with 10% Fetal Clone II and 1% antibiotic–antimycotic at 37 °C and 5% CO2. Vero E6 cells used in plaque-reduction neutralization test (PRNT) assays were cultured in DMEM supplemented with 10% Fetal Clone II and amphotericin B (0.25 μg ml−1) at 37 °C and 5% CO2. Lentivirus encoding hACE2-P2A-TMPRSS2 was made to generate A549-hACE2-TMPRSS2 cells, which were maintained in DMEM supplemented with 10% FBS and 1 μg ml−1 puromycin. Expi293 cells were maintained in the manufacturer’s suggested medium. BHK-21/WI-2 cells were obtained from Kerafast and cultured in DMEM with 5% FBS at 37 °C and 6–8% CO2. Cell lines were not authenticated. All cells lines were tested for mycoplasma and remained negative.

In vitro mRNA expression

HEK293T cells were transiently transfected with mRNA encoding SARS-CoV-2 wild-type S or S(2P) protein using a TranIT mRNA transfection kit (Mirus). After 24 h, the cells were collected and resuspended in fluorescence-activated cell sorting (FACS) buffer (1× PBS, 3% FBS, 0.05% sodium azide). To detect surface-protein expression, the cells were stained with 10 μg ml−1 ACE2–Flag (Sigma) or 10 μg ml−1 CR302235 in FACS buffer for 30 min on ice. Thereafter, cells were washed twice in FACS buffer and incubated with FITC–anti-Flag (Sigma) or Alexa Fluor 647–goat anti-human IgG (Southern Biotech) in FACS buffer for 30 min on ice. Live/Dead aqua fixable stain (Invitrogen) were used to assess viability. Data acquisition was performed on a BD LSRII Fortessa instrument (BD Biosciences) and analysed by FlowJo software v.10 (Tree Star).

Mouse models

Animal experiments were carried out in compliance with all pertinent US National Institutes of Health regulations and approval from the Animal Care and Use Committee (ACUC) of the Vaccine Research Center, Moderna Inc., or University of North Carolina at Chapel Hill. For immunogenicity studies, 6- to 8-week-old female BALB/c (Charles River), BALB/cJ, C57BL/6J or B6C3F1/J mice (Jackson Laboratory) were used. mRNA formulations were diluted in 50 μl 1× PBS, and mice were inoculated intramuscularly in the same hind leg for both prime and boost. Control mice received PBS because previous studies have demonstrated the mRNA formulations being tested do not create substantial levels of nonspecific immunity beyond a few days36,37,38. For all SARS-CoV-2 S(2P) protein vaccinations, mice were inoculated intramuscularly with SAS as previously described1. For S(2P) + alum immunizations, SARS-CoV-2 S(2P) protein + 250 μg alum hydrogel was delivered intramuscularly. For challenge studies to evaluate MERS-CoV vaccines, 16- to 20-week-old male and female 288/330+/+mice22 were immunized. Four weeks post-boost, pre-challenge sera were collected from a subset of mice, and the remaining mice were challenged with 5 × 105 PFU of a mouse-adapted MERS-CoV EMC derivative, m35c439. On day 3 post-challenge, lungs were collected and haemorrhage and viral titre were assessed according to previously published methods40. For challenge studies to evaluate SARS-CoV-2 vaccines, BALB/cJ mice were challenged with 105 PFU SARS-CoV-2 MA. This virus contains two mutations (Q498T/P499Y) in the receptor binding domain that enable binding of SARS-CoV-2 S protein to the mouse ACE2 receptor and infection and replication in the upper and lower respiratory tract32. On day 2 post-challenge, lungs and nasal turbinates were collected for viral titre assessment according to previously published methods32. Sample size for animal experiments was determined on the basis of criteria set by institutional ACUC. Experiments were not randomized or blinded.

Histology

Lungs were collected from mice at the indicated study end points and placed in 10% neutral-buffered formalin until adequately fixed. Thereafter, tissues were trimmed to a thickness of 3–5 mm, processed and paraffin embedded. The respective paraffin tissue blocks were sectioned at 5 μm and stained with haematoxylin and eosin. All sections were examined by a board-certified veterinary pathologist using an Olympus BX51 light microscope, and photomicrographs were taken using an Olympus DP73 camera.

ELISA

Nunc Maxisorp ELISA plates (ThermoFisher) were coated with 100 ng per well of protein in 1× PBS at 4 °C for 16 h. Where applicable, to eliminate fold-on-specific binding from MERS-CoV S(2P) or SARS-CoV-2 S(2P) protein-immune mouse serum, 50 μg ml−1 of fold-on protein was added for 1 h at room temperature. After standard washes and blocks, plates were incubated with serial dilutions of heat-inactivated sera for 1 h at room temperature. Following washes, anti-mouse IgG, IgG1 or IgG2a and/or IgG2c–horseradish peroxidase conjugates (ThermoFisher) were used as secondary antibodies, and 3,5,3′5′-tetramethylbenzidine (TMB) (KPL) was used as the substrate to detect antibody responses. End-point titres were calculated as the dilution that emitted an optical density exceeding 4× background (secondary antibody alone).

Lentivirus-based pseudovirus-neutralization assay

The pseudovirus-neutralization assay measures the inhibition of pseudovirus attachment and entry including fusion-inhibiting activity. It is a single-round virus, does not replicate, and does not express the S protein in transduced cells. Therefore, pseudovirus infection will not cause cell-to-cell fusion or plaque formation that can be measured in a classical neutralization assay using live virus. This pseudovirus neutralization assay has been shown to correlate with live virus plaque-reduction neutralization33, and because it does not require BL3 containment, was chosen as the preferred assay for measuring neutralizing activity in these studies. We introduced divergent amino acids, as predicted from translated sequences, into the CMV/R-MERS-CoV EMC S (GenBank: AFS88936) gene41 to generate a MERS-CoV m35c4 S gene39. To produce SARS-CoV-2 pseudoviruses, a codon-optimized CMV/R-SARS-CoV-2 S (Wuhan-1, GenBank: MN908947.3) plasmid was constructed. Pseudoviruses were produced by co-transfection of plasmids encoding a luciferase reporter, lentivirus backbone, and S genes into HEK293T/17 cells (ATCC CRL-11268), as previously described41. For SARS-CoV-2 pseudovirus, human transmembrane protease serine 2 (TMPRSS2) plasmid was also co-transfected42. Pseudoneutralization assay methods have been previously described1,33. In brief, heat-inactivated serum was mixed with pseudoviruses, incubated, and then added to Huh7.5 cells or ACE-2-expressing 293T cells for MERS-CoV and SARS-CoV-2 respectively. Seventy-two hours later, cells were lysed and luciferase activity (in relative light units (RLU)) was measured. Per cent neutralization was normalized considering uninfected cells as 100% neutralization and cells infected with only pseudovirus as 0% neutralization. IC50 titres were determined using a log (agonist) vs normalized-response (variable slope) nonlinear function in Prism v8 (GraphPad).

Recombinant VSV∆G-based pseudovirus neutralization assay

Codon-optimized wild-type (D614) or D614G spike gene (Wuhan-Hu-1 strain, NCBI reference sequence: NC_045512.2) was cloned into pCAGGS vector. To generate VSV∆G-based SARS-CoV-2 pseudovirus, BHK-21/WI-2 cells were transfected with the spike expression plasmid and infected VSV∆G-firefly-luciferase as previously described43. A549-hACE2-TMPRSS2 cells were infected by pseudovirus for 1 h at 37 °C. The inoculum virus or virus-antibody mix was removed after infection. Eighteen hours later, an equal volume of One-Glo reagent (Promega) was added to culture medium for readout using a BMG PHERastar-FS plate reader. The neutralization procedure and data analysis are the same as mentioned above for the lentivirus-based pseudovirus neutralization assay.

PRNT assays

Heat-inactivated sera were diluted in gelatin saline (0.3% (wt/vol) gelatin in PBS supplemented with CaCl2 and MgCl2 to generate a 1:5 dilution of the original specimen, which served as a starting concentration for further serial log4 dilutions terminating in 1:81,920. Sera were combined with an equal volume of SARS-CoV-2 clinical isolate 2019-nCoV USA-WA1-F6/2020 in gelatin saline, resulting in an average concentration of 730 PFU per ml (determined from plaque counts of 24 individual wells of untreated virus) in each serum dilution. Thus, final serum concentrations ranged from 1:10 to 1:163,840 of the original. Virus–serum mixtures were incubated for 20 min at 37 °C, followed by adsorption of 0.1 ml to each of two confluent Vero E6 cell monolayers (in 10 cm2 wells) for 30 min at 37 °C. Cell monolayers were overlaid with DMEM containing 1% agar and incubated for 3 d at 37 °C in humidified 5% CO2. Plaques were enumerated by direct visualization. The average number of plaques in virus + serum (duplicate) and virus-only (24 repeats) wells was used to generate percent neutralization curves according the following formula: 1 – (ratio of mean number of plaques in the presence and absence of serum). The PRNT IC50 titre was defined as the reciprocal serum dilution at which the neutralization curve crossed the 50% threshold.

Intracellular cytokine staining

Mononuclear single-cell suspensions from whole mouse spleens were generated using a gentleMACS tissue dissociator (Miltenyi Biotec) followed by 70-μm filtration and density gradient centrifugation using Fico/Lite-LM medium (Atlanta Biologicals). Cells from each mouse were resuspended in R10 media (RPMI 1640 supplemented with penicillin–streptomycin antibiotic, 10% heat-inactivated FBS, Glutamax and HEPES) and incubated for 6 h at 37 °C with protein transport inhibitor cocktail (eBioscience) under three conditions: no peptide stimulation, and stimulation with two S-protein peptide pools (JPT product PM-WCPV-S-1). Peptide pools were used at a final concentration of 2 μg ml per peptide. Cells from each group were pooled for stimulation with cell stimulation cocktail (eBioscience) as a positive control. Following stimulation, cells were washed with PBS before staining with LIVE/DEAD Fixable Blue Dead Cell Stain (Invitrogen, L23105; 1:800) for 20 min at room temperature. Cells were then washed in FC buffer (PBS supplemented with 2% heat-inactivated FBS and 0.05% NaN3) and resuspended in Fc Block (BD, 553141, clone 2.4G2; 1:100) for 5 min at room temperature before staining with a surface stain cocktail containing the following antibodies: I-A/I-E PE (BD, 557000, clone M5/114.15.2; 1:2,500), CD8a BUV805 (BD, 612898, clone 53-6.7; 1:80), CD44 BUV395 (BD, 740215, clone IM7; 1:800), CD62L BV605 (Biolegend, 104418, clone MEL-14; 1:5,000) and CD4 BV480 (BD, 565634, clone RM4-5; 1:500) in brilliant stain buffer (BD). After 15 min, cells were washed with FC buffer then fixed and permeabilized using the BD Cytofix/Cytoperm fixation/permeabilization solution kit according to the manufacturer’s instructions. Cells were washed in perm/wash solution and stained with Fc Block (5 min at room temperature), followed by intracellular staining (30 min at 4 °C) using a cocktail of the following antibodies: CD3e BUV737 (BD, 741788, clone 17A2; 1:80), IFN-γ BV650 (BD, 563854, clone XMG1.2; 1:500), TNF BV711 (BD, 563944, clone MP6-XT22; 1:80), IL-2 BV421 (BD, 562969, clone JES6-5H4; 1:80), IL-4 Alexa Fluor 488 (Biolegend, 504109, clone 11B11; 1:80) and IL-5 APC (Biolegend, 504306, clone TRFK5; 1:320) in 1× perm/wash diluted with brilliant stain buffer. Finally, cells were washed in perm/wash solution and resuspended in 0.5% PFA–FC stain buffer before running on a Symphony A5 flow cytometer (BD). Analysis was performed using FlowJo software, v.10.6.2 according to the gating strategy outlined in Extended Data Fig. 10. Background cytokine expression in the no-peptide condition was subtracted from that measured in the S1 and S2 peptide pools for each individual mouse.

T cell stimulation and cytokine analysis

Spleens from immunized mice were collected two weeks post-boost. Two-million splenocytes per well (96-well plate) were stimulated in vitro with two peptide libraries, JPT1 and JPT2, (15mers with 11 amino acid overlap) covering the entire SARS-CoV-2 S protein (JPT product PM-WCPV-S-1). Both peptide libraries were used at a final concentration of 1 μg ml−1. After 24 h of culture at 37 °C, the plates were centrifuged and supernatant was collected and frozen at −80 °C for cytokine detection. Measurements and analyses of secreted cytokines from a murine 35-plex kit were performed using a multiplex bead-based technology (Luminex) assay with a Bio-Plex 200 instrument (Bio-Rad) after twofold dilution of supernatants.

Statistical analysis

Geometric means or arithmetic means are represented by the heights of bars, or symbols, and error bars represent the corresponding s.d. Dotted lines indicate assay limits of detection. Two-sided Mann–Whitney U-tests were used to compare two experimental groups and two-sided Wilcoxon signed-rank tests to compare the same animals at different time points. To compare more than two experimental groups, Kruskal–Wallis ANOVA with Dunn’s multiple comparisons tests were applied. In Extended Data Fig. 5a, b, all doses were compared to the 20 μg dose by two-sided Mann–Whitney U-test in a stepwise fashion, such that lowest doses were tested first at α = 0.05 and higher doses were tested only if the lower doses were significant. In Extended Data Fig. 5c, a Spearman correlation test was used to correlate binding antibody titres to pseudovirus-neutralizing antibody titres. Statistical analyses were performed using R v.4.0.0 or Prism v.8 (GraphPad). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.