Elicit: Impact of Tozinameran mRNA Design on Antigenicity

Skip to main content

Impact of Tozinameran mRNA Design on Antigenicity

Research reportView only

Create alertChat

May 5, 2026

How does Tozinameran's mRNA design (e.g., prefusion-stabilized spike) affect antigenicity?

Tozinameran's prefusion-stabilized spike design (K986P/V987P mutations) and N1-methylpseudouridine modifications enhance antigenicity by concentrating immune responses on neutralizing epitopes and prolonging antigen expression, producing neutralizing antibody titers 10-18 times higher than natural infection despite recognizing fewer total epitopes.

Abstract

Tozinameran’s mRNA design achieves high antigenicity through multiple synergistic elements centered on prefusion-stabilized spike protein presentation. The K986P and V987P mutations lock the spike in a conformation where approximately 20% of trimers adopt the one-RBD ‘up’ state, authentically presenting the ACE2 binding site and neutralizing epitopes. This design elicits neutralizing antibody titers 10-18 times higher than convalescent sera in non-human primates and preserves neutralization against some variants, supported by clinical efficacy exceeding 95%. N1-methylpseudouridine modification enhances mRNA stability and translation while reducing innate immune sensing, contributing to the superiority of nucleoside-modified vaccines (>94% efficacy) over unmodified designs (48% efficacy). Additional elements including optimized UTRs and TENT5A-mediated poly(A) tail extension prolong antigen production and enhance immunogenicity.

Critically, the prefusion-stabilized design produces a distinct immune profile compared to natural infection: vaccine-elicited sera recognize fewer total linear epitopes (11 vs 37) but achieve higher neutralizing potency by focusing responses on functionally critical determinants like K417, E484, and N501 residues. The 26 epitopes recognized exclusively by convalescent sera map predominantly to post-fusion conformational states masked by prefusion stabilization, representing non-neutralizing targets. This epitope focusing represents an intentional design feature prioritizing neutralizing antibody quality over total antibody diversity. The design also generates strong TH1-type CD4+ and IFNγ+ CD8+ T-cell responses with broader epitope coverage from the full-length spike compared to RBD-only constructs. Advanced stabilization approaches like the S6P mutation further enhance durability and cross-variant neutralization, while self-amplifying platforms demonstrate sustained immune activation through prolonged antigen expression.

Methods

We analyzed 10 sources from an initial pool of 200, using 9 screening criteria. Each paper was reviewed for 6 key aspects that mattered most to the research question. More on methods

Records from Elicit search

n = 200

Papers screened using: Tozinameran Focus, mRNA Design Features, Antigenicity Outcomes, Study Design, Research Context, Comparative Analysis, Beyond Efficacy/Safety Only, mRNA Vaccine Relevance, Adequate Study Type

n = 200

Papers screened out

n = 190

Papers included for extraction

n = 10

Press enter or space to select a node.You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.

Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.

Paper search

We performed a semantic search across over 138 million academic papers from the Elicit search engine, which includes all of Semantic Scholar and OpenAlex.

We ran this query: “How does Tozinameran’s mRNA design (e.g., prefusion-stabilized spike) affect antigenicity?”

The search returned 200 total results from Elicit.

We retrieved 200 papers most relevant to the query for screening.

Screening

We screened in sources based on their abstracts that met these criteria:

We considered all screening questions together and made a holistic judgement about whether to screen in each paper.

Data extraction

We asked a large language model to extract each data column below from each paper. We gave the model the extraction instructions shown below for each column.

Extract all mRNA design characteristics relevant to Tozinameran/BNT162b2 antigenicity, including:

Extract all measures of antigenicity and immune response specifically for Tozinameran-type mRNA vaccines, including:

Extract the specific relationships between mRNA design features and antigenicity outcomes for Tozinameran, including:

Extract comparative data between different mRNA vaccine designs or between vaccine-induced and natural immunity, focusing on:

Extract mechanistic explanations for how Tozinameran’s mRNA design affects antigenicity, including:

Extract the experimental model and population studied, including:

Results

Characteristics of Included Studies

Study

Full text retrieved?

Study type

Species

Primary focus

Sample size/characteristics

Annette B. Vogel et al., 2020

Yes

Preclinical

Mice, rhesus macaques

BNT162b2 design and immunogenicity

Male rhesus macaques, 2-4 years old

Annette B. Vogel et al., 2021

Yes

Preclinical

Mice, rhesus macaques

BNT162b1 vs BNT162b2 comparison

Male rhesus macaques, 2-4 years old

Lizhou Zhang et al., 2023

Yes

Preclinical

Mice

Component comparison (Pfizer vs Moderna)

7-week-old female BALB/c mice

Y. Bong et al., 2025

Yes

Preclinical

Mice

S6P mutation effects

Balb/c and K18-hACE2 transgenic mice, 6-8 weeks old

Sascha Hein et al., 2021

Yes

Observational (human sera)

Human

BNT162b2 vs CVnCoV comparison

Healthcare workers (BNT162b2), phase I participants (CVnCoV)

Annette B. Vogel et al., 2020a

Yes

Preclinical

Mice, rhesus macaques

BNT162b vaccine candidates

Male rhesus macaques, 2-4 years old

Sascha Hein et al., 2021a

Yes

Observational (human sera)

Human

Epitope recognition patterns

Vaccinated and convalescent individuals

P. Krawczyk et al., 2025

Yes

Preclinical

Mice

Re-adenylation mechanisms

6-14 week old mice

Kate S. Levine et al., 2025

No (abstract only)

Clinical trial (Phase III)

Human

ARCT-154 vs BNT162b2

Participants with 3 prior mRNA vaccine doses

Rein Verbeke et al., 2021

Yes

Review

Multiple (review)

mRNA vaccine design and performance

Phase 3 clinical trial data

toof

Pageof

The included studies spanned preclinical models (mice and non-human primates), human observational studies, clinical trials, and systematic reviews. Six studies used full text, while one was available as abstract only. Studies employed diverse vaccination schedules, with preclinical studies typically using prime-boost regimens at days 0 and 21, while human studies examined booster responses in previously vaccinated individuals.

mRNA Design Features of Tozinameran

Tozinameran (BNT162b2) incorporates several molecular design elements intended to enhance antigenicity. The vaccine encodes a full-length spike protein stabilized in its prefusion conformation through K986P and V987P substitutions, also referred to as P2 mutations. Structural analysis revealed that approximately 20% of expressed spike trimers adopt the one-RBD ‘up’, two-RBD ‘down’ conformation, which authentically presents the ACE2 binding site and other neutralizing epitopes.

The mRNA contains nucleoside modifications, specifically N1-methylpseudouridine (m1Ψ), which reduces innate immune sensing and enhances RNA translation. The construct includes a 100-nucleotide poly(A) tail, though one study noted BNT162b2 has a composite poly(A) tail structure. The 5’ untranslated region (UTR) derives from human hemoglobin α-globin (HBA1), while the 3’ UTR originates from AES-mtRNR1. Additional optimization includes non-coding sequence elements and a linker (A30LA70) to improve RNA stability and translational efficiency.

Comparative design studies revealed differences between Tozinameran and other mRNA vaccines. BNT162b2 employs the ionizable lipid ALC-0315, enzymatically incorporated Cap 1 for mRNA capping, and a prefusion-stabilized spike configuration. An alternative design, BNT162b1, encodes a soluble, secreted trimerized receptor-binding domain (RBD-foldon) rather than the full-length spike. Advanced stabilization approaches, such as the S6P mutation (incorporating KV986PP plus four additional proline substitutions at F817P, A892P, A899P, and A942P), have been explored in next-generation candidates.

Effects on Antigenicity

Antibody Responses

Tozinameran elicited robust antibody responses across preclinical and clinical studies. In rhesus macaques, prime-boost vaccination produced neutralizing geometric mean titers 10.2 to 18.0 times higher than those in convalescent human serum panels. The vaccine induced dose-dependent increases in pseudovirus neutralization titers in mice, with neutralizing antibodies remaining detectable at 3 months post-vaccination in humans.

Comparative analyses demonstrated superiority over some platforms. BNT162b2-elicited sera showed higher titers and better neutralizing capacity compared to CVnCoV-elicited sera. The vaccine preserved neutralizing antibody activity against the B.1.1.7 variant, though neutralizing capacity was significantly impacted against the B.1.351 variant. Anti-RBD titers from vaccine-elicited sera exceeded those from convalescent sera.

Design elements differentially affected antibody production. The ionizable lipid component influenced neutralizing antibody generation, with SM-102 (used in Moderna’s vaccine) producing higher neutralizing activity at days 14 and 35 post-vaccination compared to ALC-0315. UTR sequences from Pfizer-BioNTech’s 5’ region and Moderna’s 3’ region demonstrated superior performance in supporting mRNA translation and antibody production.

T-Cell Responses

The vaccine generated strong cellular immunity characterized by TH1-type CD4+ and IFNγ+ CD8+ T-cell responses in both mice and rhesus macaques. This TH1-biased response profile is favorable for vaccine safety and efficacy. CD4+ T-cell responses were identified in all COVID-19 patients and correlated with IgG and IgA antibody titers. The vaccine facilitated efficient presentation of peptide fragments to CD8+ T cells through intracellular processing.

The full-length prefusion-stabilized spike design (BNT162b2) provided a broader T-cell epitope range compared to the RBD-only construct (BNT162b1), contributing to the selection of BNT162b2 for clinical development. The balanced immune response included secretion of both TH1 cytokines (IFN-γ, IL-2, TNF-α) and TH2 cytokines (IL-4, IL-5, IL-13) in certain contexts.

Epitope Recognition and Structural Considerations

A critical finding emerged regarding epitope recognition patterns. Peptide array analyses identified 37 linear epitopes across the spike protein, with 26 epitopes almost exclusively recognized by convalescent sera but not vaccine-elicited sera. These 26 epitopes predominantly mapped to regions masked in the prefusion structure but exposed in the post-fusion conformation. Specifically, three epitopes in the conserved central helix were only exposed post-fusion.

The prefusion stabilization achieved through K986P and V987P mutations reduced structural flexibility and conferred resistance to proteolytic processing. While this stabilization enhanced presentation of the ACE2 binding site and other key neutralizing epitopes, it simultaneously masked epitopes accessible during natural infection. Vaccine-elicited sera strongly recognized linear epitopes harboring K417, E484, and N501 residues, which are critical for ACE2 binding.

The structural design resulted in approximately 20% of spike trimers adopting the one-RBD ‘up’ conformation, creating a dynamic equilibrium between RBD positions. This equilibrium allowed exposure of epitopes that would otherwise remain buried, enhancing immune recognition while maintaining the prefusion-stabilized state.

mRNA Stability and Processing

The cellular fate of vaccine mRNA significantly influenced antigenicity. Studies revealed that mRNA-1273 (Moderna) underwent more potent poly(A) tail re-adenylation by TENT5A compared to BNT162b2, correlating with higher membrane association and antigen production. Re-adenylation extended poly(A) tails from approximately 100 to 200 nucleotides, stabilizing mRNA and enhancing translation. TENT5A deficiency reduced specific immunoglobulin production in mice immunized with mRNA vaccines.

The capping strategy influenced re-adenylation efficiency, with enzymatically incorporated Cap 1 producing several-fold more antigen than CleanCap-equipped mRNA. The m1Ψ modification enhanced mRNA stability and translation while reducing innate immune signaling, improving overall antigen presentation. However, varying N1-methylpseudouridine content at wobble positions showed little effect on neutralizing antibody production.

Comparative Design Effects

Direct comparisons between BNT162b1 and BNT162b2 revealed that while both candidates elicited strong immune responses, BNT162b2 provided superior protection of the lower respiratory tract from viral RNA. BNT162b2 was selected for clinical advancement based on greater tolerability and broader T-cell epitope coverage, despite comparable immunogenicity.

Comparison with CVnCoV, an unmodified mRNA vaccine encoding a stabilized spike, highlighted the importance of nucleoside modification. BNT162b2 and Moderna (both using modified nucleosides) achieved over 94% efficacy, while CVnCoV demonstrated 48% efficacy. The modified RNA in BNT162b2 (30 μg dose) showed higher efficacy and immune responses compared to unmodified RNA in CVnCoV (12 μg dose). Reactogenicity of unmodified RNA limited achievable dosing in CVnCoV, potentially impacting antigenicity.

Component-level comparisons demonstrated that UTR selection affected translation efficiency. Pfizer-BioNTech’s 5’ UTR outperformed Moderna’s counterpart, while Moderna’s 3’ UTR exceeded Pfizer-BioNTech’s version. These UTR differences contributed to overall antigen expression levels and subsequent immune responses.

Advanced stabilization strategies, exemplified by the S6P mutation, enhanced immunogenicity beyond the standard P2 design. The S6P mutation (incorporating six proline substitutions) elicited significantly stronger B and T cell responses with more durable neutralizing antibodies compared to S2P-based vaccines. This enhanced stability preserved the prefusion conformation more effectively, leading to improved immune recognition and broader cross-variant neutralization.

Synthesis: Reconciling Epitope Diversity with Neutralizing Potency

The studies revealed an apparent paradox: prefusion-stabilized spike designs like BNT162b2 produce fewer recognized linear epitopes than natural infection yet elicit higher neutralizing antibody titers than convalescent sera. This divergence reflects fundamental differences in antigen presentation rather than a deficiency in either response.

Mechanism of Differential Epitope Exposure

The prefusion stabilization through K986P and V987P mutations constrains spike protein flexibility, preventing conformational transitions that would expose post-fusion epitopes. During natural infection, spike proteins undergo fusion-mediated conformational changes, exposing epitopes in the central helix and other regions that become accessible only in post-fusion states. The 26 epitopes preferentially recognized by convalescent sera map predominantly to these conformational-dependent regions. In contrast, the vaccine-stabilized prefusion conformation maintains epitopes critical for ACE2 binding and neutralization in an accessible state, while masking conformational epitopes.

Functional Consequences: Quality Over Quantity

The reduced epitope diversity in vaccine-elicited responses does not indicate inferior immunity; rather, it reflects focused recognition of neutralizing determinants. Vaccine-elicited antibodies target the RBD and S1 domain with high affinity, particularly epitopes containing K417, E484, and N501—residues essential for ACE2 binding. This focused response produces neutralizing titers 10-18 times higher than convalescent sera, demonstrating that epitope quality (neutralizing capacity) outweighs quantity (total epitopes recognized).

Natural infection presents both prefusion and post-fusion spike conformations, generating antibodies to a broader epitope repertoire that includes non-neutralizing targets. While this broader response may provide advantages for variant recognition, it dilutes the neutralizing antibody fraction. The vaccine’s prefusion-only presentation concentrates immune responses on functionally critical epitopes.

Design Element Contributions to Antigenicity

Multiple design elements synergistically enhance antigenicity through distinct mechanisms:

The m1Ψ modification operates primarily through enhanced mRNA stability and reduced innate immune interference, increasing antigen expression duration and quantity. This nucleoside modification proved critical; unmodified mRNA vaccines showed substantially lower efficacy (48% vs >94%) despite encoding similar antigens, likely due to rapid degradation and inflammatory responses that limited dosing.

UTR sequences regulate translation initiation and mRNA stability, with optimal combinations (Pfizer’s 5’ UTR and Moderna’s 3’ UTR) providing superior translation efficiency. The poly(A) tail undergoes enzymatic extension by TENT5A in macrophages, extending from 100 to 200 nucleotides and stabilizing membrane-associated mRNA to prolong antigen production. However, BNT162b2 shows less potent re-adenylation than mRNA-1273, potentially due to differences in membrane association efficiency or capping strategies.

The prefusion stabilization itself provides structural advantages beyond epitope presentation. The K986P and V987P mutations lock the spike in a conformation that authentically presents neutralizing epitopes while maintaining the dynamic RBD equilibrium (~20% ‘up’) necessary for immune recognition. More extensive stabilization (S6P mutation with six proline substitutions) further enhances this effect, producing stronger and more durable responses by improving thermal stability and expression levels.

Implications for Variant Protection

The prefusion-focused epitope recognition has mixed implications for variant responses. While preserved neutralization against B.1.1.7 demonstrates some cross-variant protection, significantly reduced neutralization of B.1.351 reveals limitations. Variants with mutations in conserved prefusion epitopes (particularly K417, E484, N501) can escape vaccine-elicited immunity more readily than they escape convalescent immunity, which targets additional epitopes in the central helix and other regions.

However, advanced designs incorporating broader stabilization (S6P) or self-amplifying platforms (ARCT-154) demonstrate enhanced cross-variant neutralization. The sustained antigen production from self-amplifying mRNA moves immune profiles toward activating phenotypes with broad antigenic coverage, suggesting that prolonged antigen exposure may partially compensate for reduced epitope diversity by allowing affinity maturation against variant epitopes.

Trade-offs in Design Optimization

Design choices involve inherent trade-offs between different immune response characteristics. Maximum neutralizing potency (achieved through strict prefusion stabilization) comes at the cost of epitope breadth. Higher mRNA doses and enhanced stability (through modifications, optimal UTRs, and re-adenylation) increase antigen expression but may alter the balance between humoral and cellular responses. More extensive proline stabilization (S6P vs S2P) improves durability but requires validation for potential effects on protein processing and presentation.

The selection of BNT162b2 over BNT162b1 illustrates these trade-offs: the full-length spike provides broader T-cell epitope coverage and better respiratory tract protection, while the RBD-only construct offers high-avidity ACE2 binding through flexible tethering. The choice prioritized epitope breadth and T-cell responses over maximizing RBD-focused neutralization.

Conclusion

Tozinameran’s mRNA design achieves high antigenicity through prefusion stabilization that optimally presents neutralizing epitopes, nucleoside modifications that enhance expression, and structural elements that support mRNA stability and translation. The reduced linear epitope recognition compared to natural infection represents a feature rather than limitation—focusing immune responses on functionally critical neutralizing determinants while sacrificing recognition of conformational epitopes with limited protective value. This design philosophy prioritizes neutralizing antibody quality and T-cell breadth over total antibody diversity, an approach validated by clinical efficacy exceeding 95% despite narrower epitope targeting than natural infection.

References

Annette B. Vogel, I. Kanevsky, Y. Che, K. Swanson, A. Muik, and 56 more\ (2020).A prefusion SARS-CoV-2 spike RNA vaccine is highly immunogenic and prevents lung infection in non-human primates. bioRxiv

Rein Verbeke, I. Lentacker, S. D. De Smedt, H. Dewitte\ (2021).The dawn of mRNA vaccines: The COVID-19 case. Journal of Controlled Release

Annette B. Vogel, I. Kanevsky, Y. Che, K. Swanson, A. Muik, and 69 more\ (2021).BNT162b vaccines protect rhesus macaques from SARS-CoV-2. Nature

Lizhou Zhang, Kunal R. More, Amrita Ojha, C. B. Jackson, Brian D. Quinlan, and 5 more\ (2023).Effect of mRNA-LNP components of two globally-marketed COVID-19 vaccines on efficacy and stability. npj Vaccines

Y. Bong, David Brown, Ezra Chung, N. Ananthaswamy, Renxiang Chen, and 10 more\ (2025).S6P mutation in Delta and Omicron variant spike protein significantly enhances the efficacy of mRNA COVID-19 vaccines. Frontiers in Immunology

Sascha Hein, Marie-Luise Herrlein, Ines Mhedhbi, Daniela Bender, Vanessa Haberger, and 11 more\ (2021).Analysis of BNT162b2‐ and CVnCoV‐elicited sera and of convalescent sera toward SARS‐CoV‐2 viruses. Allergy. European Journal of Allergy and Clinical Immunology

Annette B. Vogel, I. Kanevsky, Y. Che, K. Swanson, A. Muik, and 69 more\ (2020).BNT162b vaccines are immunogenic and protect non-human primates against SARS-CoV-2. bioRxiv

Sascha Hein, N. Benz, Jonathan Eisert, Marie-Luise Herrlein, D. Oberle, and 4 more\ (2021).Comirnaty-Elicited and Convalescent Sera Recognize Different Spike Epitopes. Vaccines

P. Krawczyk, M. Mazur, Wiktoria Orzeł, Olga Gewartowska, Sebastian Jeleń, and 20 more\ (2025).Re-adenylation by TENT5A enhances efficacy of SARS-CoV-2 mRNA vaccines. Nature

Kate S. Levine, Ross Blanc, Qixin Wang, Hadar Malca, R. Sekulovich, and 6 more\ (2025).Sustained Humoral Activation through self-amplifying mRNA Vaccination Enhances Longitudinal Antibody Function in a Phase III Trial. bioRxiv

Download BIBDownload RISDownload TXT

Report

Status

Gather sources

200 sources found

Details

Screen sources

10 sources included

Details

Extract data

60 data points extracted

Details

Generate report

Save PDF

BIBLaTeX, ZoteroRISZotero, MendeleyTXTAPA BibliographyPDFPDFDOCXMicrosoft Word

Chat

Got some follow-up questions?

Sign up or sign in to chat with this report.

Back

BNT162b vaccines protect rhesus macaques from SARS-CoV-2

Annette B. Vogel, I. Kanevsky, Y. Che, K. Swanson, A. Muik, M. Vormehr, L. Kranz, Kerstin C. Walzer, Stephanie Hein, Alptekin Güler, Jakob Loschko, M. Maddur, Ayuko Ota-Setlik, Kristin Tompkins, J. Cole, B. G. Lui, Thomas Ziegenhals, Arianne Plaschke, D. Eisel, Sarah C. Dany, Stephanie Fesser, Stephanie Erbar, F. Bates, D. Schneider, B. Jesionek, Bianca Sänger, A. Wallisch, Yvonne Feuchter, H. Junginger, S. A. Krumm, André P. Heinen, P. Adams-Quack, J. Schlereth, Stefan Schille, Christoph Kröner, Ramón de la Caridad Güimil Garcia, T. Hiller, Leyla Fischer, R. Sellers, S. Choudhary, O. Gonzalez, F. Vascotto, Matthew R. Gutman, J. Fontenot, Shannan Hall-Ursone, K. Brasky, M. Griffor, Seungil Han, A. Su, J. Lees, Nicole L Nedoma, E. Mashalidis, P. Sahasrabudhe, Charles Y. Tan, D. Pavliakova, Guy Singh, Camila R. Fontes-Garfias, M. Pride, Ingrid L Scully, Tara Ciolino, Jennifer Obregon, M. Gazi, R. Carrion, Kendra J Alfson, W. Kalina, D. Kaushal, P. Shi, T. Klamp, Corinna Rosenbaum, A. Kuhn, Ö. Türeci, P. Dormitzer, K. Jansen, U. Şahin

Nature·

2021·

602 citations

SourceDOI

Plain textPDF

Searching for PDF

Unable to find PDF from source

Back

mRNA Design Features

- Spike protein configuration: Full-length transmembrane spike glycoprotein (BNT162b2), prefusion-stabilized; Soluble, secreted trimerized receptor-binding domain (RBD-foldon) (BNT162b1) - Specific stabilization mutations: K986P/V987P (P2 mutations) for BNT162b2 - Nucleoside modifications: N1-methylpseudouridine (m1Ψ) - Poly(A) tail characteristics and length: 100 nucleotides - Codon optimization details: Optimized non-coding sequence elements and linker (A30LA70) for improved RNA stability and translational efficiency

Antigenicity Outcomes

- Antibody responses: High neutralizing titers and binding antibodies in mice and rhesus macaques. - Epitope recognition patterns and specificity: Not specifically mentioned. - T-cell responses: Strong T-helper-1 (Th1) CD4+ and IFNγ+ CD8+ T-cell responses. - Cross-reactivity against variants: Not mentioned. - Duration of immune response: Not mentioned. - Cellular uptake and processing measures: Not mentioned. - Immune breadth or diversity: Not specifically mentioned.

Design-Antigenicity Effects

- Direct comparisons: BNT162b1 (RBD-foldon) vs. BNT162b2 (prefusion-stabilized spike) - Quantitative effects: BNT162b2 elicits higher neutralizing titers (8.2 to 18.2 times that of convalescent serum) - Design elements enhancing antigenicity: Prefusion-stabilized spike in BNT162b2; RBD-foldon in BNT162b1 - Dose-response relationships: Not explicitly mentioned - Time-dependent effects: Not explicitly mentioned - Trade-offs: Greater tolerability and broader T-cell epitope range with BNT162b2

Vaccine Comparisons

- BNT162b1 vs BNT162b2: Both induce strong antibody and T-cell responses, but BNT162b2 provides better protection against SARS-CoV-2 challenge in rhesus macaques. - BNT162b2 was selected for further clinical trials due to greater tolerability and broader T-cell epitope coverage. - Vaccine-induced vs convalescent immunity: Vaccine-induced immunity is significantly higher than natural immunity. - No direct comparisons with other mRNA vaccines (Moderna, CureVac) or cross-platform comparisons with non-mRNA vaccines are provided.

Mechanistic Insights

- Structural biology insights: BNT162b1 encodes a soluble, secreted RBD-foldon; BNT162b2 encodes a full-length spike protein stabilized in its prefusion conformation (S(P2)). The RBD-foldon has flexible tethering, and S(P2) has a dynamic equilibrium between RBD 'up' and 'down' states. - Cellular processing mechanisms: The m1Ψ-modification enhances mRNA stability and translation efficiency while reducing innate immune sensing. - Epitope exposure patterns: The RBD-foldon authentically presents the ACE2 binding site, and the prefusion-stabilized spike has a dynamic equilibrium affecting epitope exposure. - Role of specific design elements: Prefusion stabilization and flexible tethering enhance immune recognition by affecting epitope exposure and binding affinity to ACE2. - Molecular explanations for observed antigenicity differences: The design elements influence antigenicity by affecting epitope exposure and binding affinity. - Mechanistic models proposed: The dynamic equilibrium between RBD 'up' and 'down' states and the flexible tethering mechanism contribute to antigenicity.

Study Model

- Species: Mouse, Rhesus macaque - Study design: Preclinical - Sample characteristics: Male rhesus macaques, 2-4 years old - Vaccination schedule: Intramuscular injections on Days 0 and 21 with 30 or 100 µg doses - Time points of immune assessment: Day 14 after first dose, Day 28 after second dose - Relevant study limitations: Small number of animals used

A safe and effective vaccine against COVID-19 is urgently needed in quantities that are sufficient to immunize large populations. Here we report the preclinical development of two vaccine candidates (BNT162b1 and BNT162b2) that contain nucleoside-modified messenger RNA that encodes immunogens derived from the spike glycoprotein (S) of SARS-CoV-2, formulated in lipid nanoparticles. BNT162b1 encodes a soluble, secreted trimerized receptor-binding domain (known as the RBD-foldon). BNT162b2 encodes the full-length transmembrane S glycoprotein, locked in its prefusion conformation by the substitution of two residues with proline (S(K986P/V987P); hereafter, S(P2) (also known as P2 S)). The flexibly tethered RBDs of the RBD-foldon bind to human ACE2 with high avidity. Approximately 20% of the S(P2) trimers are in the two-RBD 'down', one-RBD 'up' state. In mice, one intramuscular dose of either candidate vaccine elicits a dose-dependent antibody response with high virus-entry inhibition titres and strong T-helper-1 CD4

A safe and effective vaccine against COVID-19 is urgently needed in quantities sufficient to immunise large populations. We report the preclinical development of two BNT162b vaccine candidates, which contain lipid-nanoparticle (LNP) formulated nucleoside-modified mRNA encoding SARS-CoV-2 spike glycoprotein-derived immunogens. BNT162b1 encodes a soluble, secreted, trimerised receptor-binding domain (RBD-foldon). BNT162b2 encodes the full-length transmembrane spike glycoprotein, locked in its prefusion conformation (P2 S). The flexibly tethered RBDs of the RBD-foldon bind ACE2 with high avidity. Approximately 20% of the P 2S trimers are in the two-RBD 'down,' one-RBD 'up' state. In mice, one intramuscular dose of either candidate elicits a dose-dependent antibody response with high virus-entry inhibition titres and strong T H 1 CD4 + and IFNγ + CD8 + T-cell responses. Prime/boost vaccination of rhesus macaques with BNT162b candidates elicits SARS-CoV-2 neutralising geometric mean titres 8.2 to 18.2 times that of a SARS-CoV-2 convalescent human serum panel. The vaccine candidates protect macaques from SARS-CoV-2 challenge, with BNT162b2 protecting the lower respiratory tract from the presence of viral RNA and with no evidence of disease enhancement. Both candidates are being evaluated in phase 1 trials in Germany and the United States [1] [2] [3] . BNT162b2 is being evaluated in an ongoing global, pivotal Phase 2/3 trial (NCT04380701, NCT04368728).

Due to the shattering impact of the coronavirus disease 2019 (COVID-19) pandemic on human health and society, multiple collaborative research programs have been launched, generating new insights and progress in vaccine development. Soon after emerging in December 2019, the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) was identified as a β-coronavirus with high sequence similarity to bat-derived SARS-like coronaviruses 4, 5 . Fast pandemic vaccine availability is critical, and the rapid globalised response is mirrored by the upload of over 212,000 viral genome sequences as of November 23, 2020, to GISAID (Global Initiative on Sharing All Influenza Data).

The trimeric spike glycoprotein (S) of SARS-CoV-2 is a key target for virus neutralising antibodies 6 and the prime candidate for vaccine development. S binds its cellular receptor, human angiotensin converting enzyme 2 (ACE2), through a receptor-binding domain (RBD), which is part of S1, its N-terminal furin cleavage fragment 7, 8 . On S, the RBDs have 'up' positions, in which the receptor binding sites and their dense Article cluster of neutralising epitopes are exposed, and 'down' positions, in which the receptor binding sites are buried, but some S neutralising epitopes on and off the RBDs remain available [9] [10] [11] [12] . S rearranges to translocate the virus into cells by membrane fusion 9, 13 . The C-terminal furin cleavage fragment, S2, contains the fusion machinery 14 .

Messenger RNA technology allows versatile vaccine antigen design and highly scalable, fast manufacturing. With efficient lipid-nanoparticle (LNP) formulation processes, RNA vaccines are highly suited to rapid development and pandemic supply 15, 16 . RNA generated from DNA templates by a highly productive, cell-free in vitro transcription process is molecularly well defined and free of animal-origin materials. Here, we report the preclinical development of the LNP formulated N 1 -methyl-pseudouridine (m1Ψ) nucleoside-modified mRNA (modRNA) BNT162b vaccine candidates that encode SARS-CoV-2 S-derived immunogens (Fig. 1a ). The m1Ψ-modification dampens innate immune sensing and, together with optimised non-coding sequence elements, increases efficiency of RNA translation in vivo [16] [17] [18] . Vaccines based on modRNA have proven immunogenic for several viral targets 19, 20 .

Both BNT162b vaccines are being evaluated in phase 1 clinical trials in the US (NCT04368728) and Germany (NCT04380701, EudraCT: 2020-001038-36); BNT162b2 is being evaluated in a pivotal, global, phase 2/3 safety and efficacy study [1] [2] [3] .

Construct design and analysis of expressed antigens

BNT162b1 RNA encodes the RBD with the SARS-CoV-2 S signal peptide (SP) fused to its N-terminus to enable ER translocation and secretion and with the trimerisation domain (foldon) of T4 fibritin 21 fused to its C-terminus for multimeric display; BNT162b2 RNA encodes full-length S, stabilised in the prefusion conformation by the mutation of residues 986 and 987 to proline (P2 S; Fig. 1a ) 10, 22, 23 . Both RNAs have single, sharp microfluidic capillary electrophoresis profiles, consistent with their calculated lengths, indicating high purity and integrity (Fig. 1b ). Robust expression of RBD-foldon or P2 S was detectable by flow cytometry upon transfection of HEK293T cells with BNT162b1 RNA or BNT162b2 RNA, respectively, formulated as LNPs or mixed with a transfection reagent (Extended Data Fig. 1a ). In transfected cells, BNT162b1-encoded RBD and BNT162b2-encoded P2 S localised to the secretory pathway as shown by immunofluorescence microscopy (Extended Data Fig. 1b ). A main band of RBD-containing protein with an apparent MW >75 kDa was detected in the medium of BNT162b1 RNA-transfected cells (together with lesser quantities of a faster migrating species) by western blot under denaturing and non-denaturing conditions, consistent with secretion of trimeric RBD-foldon (predicted MW 88.4 kD; Extended Data Fig. 1c ).

For further structural characterisation, the RBD-foldon and P2 S antigens were expressed from DNA corresponding to the RNA coding sequences. The RBD-foldon was purified from the medium of transfected Expi293F cells by affinity capture with the ACE2-peptidase domain (PD) immobilised on agarose beads, leaving little residual RBD-foldon uncaptured from the medium. Evidence that the RBD-foldon has three RBDs flexibly tethered to a central hub was obtained by electron microscopy (EM), which revealed a variety of conformations (Fig. 1c ). The trimerised RBD bound to the human ACE2 peptidase domain (PD) with an apparent K D of <5 pM, which is 1,000-fold the reported K D of 5 nM for monomeric RBD and consistent with the avidity effect of multivalent binding enabled by the flexible tethering (Extended Data Fig. 1d ). Although the flexibility of the RBD-foldon precluded direct structural analysis at high resolution, one RBD per trimer could be immobilised by binding to a complex of ACE2 and the B 0 AT1 neutral amino acid transporter, which ACE2 chaperones, when that complex was in the previously reported closed conformation (Fig. 1d ) 8 . The size and symmetry of the RBD-foldon/ACE2/B 0 AT1 ternary complex aided image reconstruction by electron cryomicroscopy (cryo-EM), and the structure of the RBD in the complex was determined to 3.24 Å resolution (Fig. 1e , Extended Data Table 1 and Supplementary Fig. 2 ). One copy of the RBD was resolved for each bound trimer. The binding interface between the resolved RBD and the ACE2 extracellular domain was fitted to a previously reported structure and showed good agreement 7 . The high avidity binding to ACE2 and well-resolved structure in complex with ACE2 demonstrate that the recombinant RBD-foldon authentically presents the ACE2 binding site targeted by many SARS-CoV-2 neutralising antibodies 11, 24 .

The trimeric P2 S was affinity purified from detergent solubilised protein via the C-terminal TwinStrep tag. P2 S bound the human ACE2-PD and a human anti-RBD neutralising antibody B38 with high affinity (apparent K D 1 nM for each, Extended Data Fig. 1e , f) 25 . Structural analysis by cryo-EM produced a 3.29 Å nominal resolution mass density map, into which a previously published atomic model 10 was fitted and rebuilt (Fig. 1f ; Extended Data Fig. 2a , b and Extended Data Table 1 ). The rebuilt model showed good agreement with reported structures of prefusion full-length wild type S and its ectodomain with P2 mutations 9, 10 . Three-dimensional classification of the dataset showed a class of particles that was in the one RBD 'up' (accessible for receptor binding), two RBD 'down' (closed) conformation and represented 20.4% of the trimeric molecules (Fig. 1g , Extended Data Fig. 2c ). The remainder were in the all RBD 'down' conformation. The RBD in the 'up' conformation was less well resolved than other parts of the structure, suggesting conformational flexibility and a dynamic equilibrium between RBD 'up' and RBD 'down' states, as also suggested by others 9, 26 . The binding and structural analyses indicate that the BNT162b2 RNA sequence encodes a recombinant P2 S that can authentically present the ACE2 binding site and other epitopes targeted by SARS-CoV-2 neutralising antibodies.

BNT162b-elicted immunogenicity in mice

To study vaccine immunogenicity, B-and T-cell responses were characterised in a series of experiments in BALB/c mice after a single intramuscular (IM) immunisation with 0.2, 1, or 5 µg of BNT162b vaccines, or buffer control. One immunisation with either candidate induced high dose level-dependent RBD-and S1-binding serum IgG titres (Fig. 2a ; Extended Data Fig. 3a-d ), which increased more steeply for BNT162b2. On day 28 after one immunisation with 5 µg BNT162b1 or BNT162b2, RBD-binding geometric mean endpoint titres were 752,680 or 434,560, respectively. Polyclonal IgG elicited by either candidate had strong apparent binding affinity for a recombinant RBD target antigen (geometric mean apparent K D 717 pM for BNT162b1 and 993 pM for BNT162b2), with a low apparent off-rate and a high apparent on-rate (Extended Data Fig. 3e ). Serum samples from buffer-immunised control animals had no detectable RBD-or S1-specific IgG (Fig. 2a, b and Extended Data Fig. 3a-d ), and neither did serum samples from animals immunised up to two times with equivalent LNP-formulated modRNA that encoded a SARS-CoV-2 irrelevant antigen (not shown).

Virus entry inhibition by BNT162b immunised mouse serum was measured with a vesicular stomatitis virus (VSV)-based SARS-CoV-2 pseudovirus neutralisation assay. Like the antigen-specific IgG geometric mean titres (GMTs), fifty percent pseudovirus neutralisation (pVNT 50 ) GMTs increased steadily after immunisation with 5 µg of either candidate, reaching 1,056 for BNT162b1 and 296 for BNT162b2 on Day 28 after immunisation (Fig. 2b , Extended Data Fig. 3f, g ). A random selection of samples was tested in a SARS-CoV-2 virus neutralisation assay, demonstrating strong correlation of pseudovirus and SARS-CoV-2 neutralisation (Pearson correlation of 0.9479 between the tests (Extended Data Fig. 3h ). In summary, each candidate induced a high functional antibody response in mice, with BNT162b1 inducing higher titres after one immunisation.

Characterisation of antigen-specific splenic T-cell responses in mice 12 and 28 days after BNT162b vaccine immunisation revealed a high fraction of CD4 + and CD8 + T cells that produced IFNγ and CD8 + cells

that produced IL-2, as shown by enzyme linked immunospot assay (ELISpot) or intracellular cytokine staining (ICS) flow cytometry analysis after ex vivo restimulation with a full-length S peptide pool (Fig. 2c , Extended Data Fig. 4a, b ). Total splenocytes harvested on Day 28 and re-stimulated with the full-length S peptide pool secreted high levels of the T H 1 cytokines IL-2 or IFNγ and minute or undetectable levels of the T H 2 cytokines IL-4, IL-5 or IL-13, as measured in multiplex immunoassays (Fig. 2d ). Overall, the patterns of CD4 + and CD8 + T-cell responses were similar for the two vaccine candidates, with a somewhat stronger IFNγ-producing CD8 + T-cell response in BNT162b2-immunised mice.

Vaccine-induced effects on the proliferation and dynamics of immune cell populations were assessed in injection site draining lymph nodes (dLNs), to evaluate the principal immune-educated compartments for proficient T-and B-cell priming, as well as in blood and spleen, to evaluate systemic vaccine effects. Higher numbers of plasma cells, class switched IgG1-and IgG2a-positive B cells, and germinal center B cells were observed in dLNs, and higher numbers of class switched IgG1-positive and germinal centre B cells were observed in spleens of mice 12 days after immunisation with 5 µg of either vaccine as compared to control (Extended Data Fig. 4c, d ). Vaccine-immunised mice had significantly fewer circulating B cells than control mice as measured in blood at Day 7 post-immunisation (Extended Data Fig. 4e ), which may imply that B-cell homing to lymphoid compartments contributed to augmented B-cell counts in dLN and spleen.

The dLNs from BNT162b1-or BNT162b2-immunised mice also displayed significantly elevated counts of CD8 + and CD4 + T cells, which were most pronounced for T follicular helper (T FH ) cells, including ICOS + subsets that are essential for germinal centre formation (Extended Data Fig. 4c ). Both BNT162b vaccines increased T FH cell counts in the spleen and blood, while an increase in circulating CD8 + T cells was only detected in BNT162b2-immunised mice (Extended Data Fig. 4d, e ). In aggregate, these data indicate a strong induction of SARS-CoV-2 pseudovirus neutralisation titres and systemic CD8 + and T H 1-driven CD4 + T-cell responses by both modRNA vaccine candidates, with a somewhat more pronounced cellular response to BNT162b2.

BNT162b-elicted immunogenicity in rhesus macaques

To assess the immunogenicity of BNT162b1 and BNT162b2 in non-human primates, groups of six male, 2-4 year old rhesus macaques were immunised IM with 30 or 100 µg of BNT162b1, BNT162b2, or saline control on Days 0 and 21. RBD-binding IgG was readily detectable by Day 14 after Dose 1, and levels increased further 7 days after Dose 2 (Day 28; Fig. 3a ). On Day 28, geometric mean RBD-binding IgG concentrations (GMCs) were 20,962 units (U)/mL (30 µg dose level) and 48,575 U/mL (100 µg dose level) for BNT162b1 and 23,781 U/mL (30 µg dose level) and 26,170 U/mL (100 µg dose level) for BNT162b2. For comparison, the RBD-binding IgG GMC of a panel of 38 SARS-CoV-2 convalescent human sera (HCS) was 602 U/mL, lower than the GMC of immunised rhesus macaques after one or two doses.

Fifty percent virus neutralisation GMTs, measured by a SARS-CoV-2 neutralisation assay 27 (not a pseudovirus neutralisation assay), were detectable in the sera of most BNT162b1-immunised rhesus macaques by Day 21 after Dose 1 and in all BNT162b2-immunised macaques by Day 14 after Dose 1 (Fig. 3b ). There was a strong boosting effect, with comparable GMTs elicited by BNT162b1 (768 for 30 µg and 1,714 for 100 µg) or BNT162b2 (962 for 30 µg or 1,689 for 100 µg), measured in sera drawn 7 or 14 days after Dose 2. For BNT162b2, sera were available up to Day 56 after Dose 1 (28 days after Dose 2), and robust GMTs of 285 for 30 µg and 283 for 100 µg dose levels persisted to that time point. For comparison, the neutralisation GMT of the human convalescent serum was 94, substantially lower than the GMTs of rhesus macaque sera drawn 21 or 35 days after Dose 2.

S-specific T-cell responses of the BNT162b2-or saline-immunised rhesus macaques were analysed using peripheral blood mononuclear cells (PBMCs) collected before immunisation and at the times indicated after Doses 1 and 2. ELISpot demonstrated strong IFNγ but minimal IL-4 responses after Dose 2 (Fig. 3c, d , and Extended Data Fig. 5a ). ICS confirmed that BNT162b2 elicited a high frequency of CD4 + T cells that produced IFNγ, IL-2, or TNF but a low frequency of CD4 + T cells that produced IL-4, indicating a T H 1-biased response (Extended Data Fig. 5b and c ). ICS also demonstrated that BNT162b2 elicited circulating S-specific CD8 + T cells that produced IFNγ (Extended Data Fig. 5d ).

BNT162b-elicted protection in rhesus macaques

Forty-one to fifty-five days after Dose 2, 6 of the 2-4 year old rhesus macaques that had been immunised with 100 µg BNT162b1 and 6 that had been immunised with 100 µg BNT162b2 were challenged with 1.05 × 10 6 plaque forming units of SARS-CoV-2 (strain USA-WA1/2020), split equally between intranasal and intratracheal routes, as previously described (Extended Data Fig. 6 , Extended Data Table 2 ) 28 . In addition, nine age-matched macaques (controls) that had been mock-immunised with saline received the same SARS-CoV-2 challenge, and 6 age-matched macaques (sentinels), 3 of which had been immunised with 30 µg BNT162b2, were mock-challenged with cell culture medium. Nasal, oropharyngeal (OP), and rectal swabs were collected, and bronchoalveolar lavage (BAL) was performed at the times indicated (Extended Data Table 2 ). Samples were tested for SARS-CoV-2 RNA (genomic RNA and subgenomic transcripts) by reverse-transcription quantitative polymerase chain reaction (RT-qPCR). All personnel performing clinical, radiological, histopathological, or RT-qPCR evaluations were blinded to the group assignments of the macaques.

Viral RNA was detected in BAL fluid from 7 of the 9 control macaques on Day 3, from 4 of 8 on Day 6 after challenge (with 1 indeterminant result), and from none of the 6 that underwent BAL at the end of project (EOP, Days 7-23 after challenge; Fig. 4a ). Viral RNA was detected in the BAL fluid of 2 of 6 BNT162b1-immunised macaques on day 3 after challenge and from none thereafter. At no time point sampled was viral RNA detected in BAL fluid from the BNT162b2-immunised and SARS-CoV-2 challenged macaques.

In nasal swabs obtained on the day after challenge, viral RNA was detected from control-immunised macaques (4 of 9) and BNT162b2immunised macaques (5 of 6) but not from BNT162b1-immunised macaques (Fig. 4b ). In subsequent nasal swabs, viral RNA was detected from some of the control-immunised macaques on each sampling (5 of 9 on Day 3, 4 of 9 on Day 6, and 2 of 9 on Days 7-23), from some BNT162b1-immunised macaques on only 1 sampling (2 of 6 on Day 6), and from none of the BNT162b2-immunised macaques on any sampling. Similar patterns were seen in OP and rectal swabs, with viral RNA more often detected in control-immunised macaques than in BNT162b1-or BNT162b2-immunised macaques and with more persistence of viral RNA in rectal swabs than in OP swabs (Extended Data Fig. 7a, b ).

At the time of challenge, SARS-CoV-2 neutralising titres ranged from 208 to 1,185 in the BNT162b1-immunised animals and from 260 to 1,004 in the BNT162b2-immunised animals. Neutralising titres were below the limit of detection in the control animals (Fig. 4c, d ). The control animals responded to infectious virus challenge with an increase in SARS-CoV-2 neutralising titres, consistent with an immune response to viral infection. However, there was no trend toward increasing SARS-CoV-2 neutralising titres in response to viral challenge in the BNT162b1-immunised or BNT162b2-immunised animals, consistent with immunisation suppressing SARS-CoV-2 infection. The maximum SARS-CoV-2 neutralising titre elicited by virus challenge of control rhesus macaques remained below 150 through the time of necropsy, whereas all immunised animals maintained neutralising titres greater than 150 throughout the challenge experiment.

None of the challenged animals, whether immunised or not, showed clinical signs of illness (Extended Data Fig. 7c-f ). Radiographic abnormalities were generally minimal or mild and were not consistently associated with viral challenge (Extended Data Fig. 8a, b ). Histopathology of necropsy specimens obtained 7-8 days after challenge revealed

localised areas of pulmonary inflammation that were limited in extent even in the control animals challenged after mock immunisation with saline (Extended Data Fig. 8c ). We conclude that the 2-4 year old male rhesus macaque challenge model is primarily a SARS-CoV-2 infection model rather than a COVID-19 disease model.

Discussion

We demonstrate that BNT162b1 or BNT162b2, LNP-formulated, m1Ψ nucleoside-modified mRNAs that encode secreted, trimerised SARS-CoV-2 RBD or prefusion-stabilised S, respectively, induce strong antigen-specific immune responses in mice and rhesus macaques. The RBD-foldon coding sequence directs the expression and secretion of a flexible, trimeric protein that binds ACE2 with high affinity and has structurally intact ACE2 receptor binding sites. Protein expressed from DNA with the BNT162b2-encoded P2 S amino acid sequence was confirmed to be in the prefusion conformation by cryo-EM. This analysis showed that the antigenically important RBD can assume the 'up' conformation, with the receptor binding site, rich in neutralising epitopes, accessible in a proportion of the molecules 24 . The alternative states observed likely reflect a dynamic equilibrium between RBD 'up' and 'down' positions 10, 26 . Binding of expressed and purified P2 S to ACE2 and a neutralising monoclonal antibody further demonstrates its conformational and antigenic integrity.

In mice, a single sub-microgram immunisation with either BNT162b candidate rapidly induced high antibody titres that inhibited pseudovirus entry in the range of or above recently reported neutralising titres elicited by other SARS-CoV-2 vaccine candidates 29, 30 . The candidates also induced strong T FH and T H 1 type CD4 + T-cell responses, the latter thought to be a more general effect of LNP-formulated modRNA vaccines against SARS-CoV-2 30 . Both CD4 + T-cell types are known to support antigen-specific antibody generation and maturation. In some animal models of respiratory virus infection, a T H 2 type CD4 + T-cell response has been associated with vaccine-associated enhanced respiratory disease 31, 32 . Therefore, a T H 1 type response to immunisation is preferred, as it may reduce the theoretical risk of enhanced pulmonary disease during subsequent viral infection. Immunisation with the vaccine candidates triggered redistribution of B cells from the blood to lymphoid tissues, where antigen presentation occurs. In humans, T FH cells in the circulation after vaccination with a VSV-vectored Ebola vaccine candidate have been correlated with a high frequency of antigen-specific antibodies 33 . After vaccination of mice with BNT162b1 or BNT162b2, high numbers of T FH were present in both blood and LN, a potential correlate for the generation of a strong adaptive B-cell response in germinal centres. In addition to eliciting favourable CD4 + T-cell responses, both BNT162b1 and BNT162b2 elicit CD8 + T-cell responses in mice, with BNT162b2 appearing to be somewhat more efficient at eliciting antigen-specific cytotoxic IFNγ CD8 + T cells.

BNT162b1 and BNT162b2 elicit immune profiles in rhesus macaques similar to those observed in mice. Seven days after Dose 2 of 100 µg administered to macaques, during the expansion phase of the antibody response, neutralising GMTs elicited by either candidate reached approximately 18-times the GMT of a human SARS-CoV-2 convalescent serum panel. Neutralising GMTs declined by Day 56 (35 days after Dose 2), consistent with the contraction phase, but remained well above the GMT of the panel. The duration of the study was not long enough to assess the rate of decline during the plateau phase of the antibody response. As it had in mice, BNT162b2 elicted a strongly T H 1-biased CD4 + T-cell response and IFNγ + CD8 + T-cell response in rhesus macaques.

Limitation and clearance of virus infections is promoted by the interplay of neutralising antibodies that eliminate infectious particles with CD8 + T cells that target intracellular virus reservoirs. CD8 + T cells may also reduce the influx of monocytes into infected lung tissue, which can be associated with undesirable IL-6 and TNF production and impaired antigen presentation 34, 35 . The responses elicited by the vaccine candidates reflect a pattern favourable for vaccine safety and efficacy, providing added reassurance for clinical translation 36 . The contributions of the individual immune effector systems to human protection from SARS-CoV-2 are not yet understood. Therefore, it appears prudent to develop COVID-19 vaccines that enlist concomitant cognate B cells, CD4 + T cells, and CD8 + T-cell responses.

Both candidates protected 2-4 year old rhesus macaques from infectious SARS-CoV-2 challenge, with reduced detection of viral RNA in immunised animals compared to those that received saline. Immunisation with BNT162b2 provided particularly strong RT-qPCR evidence for lower respiratory tract protection, as demonstrated by the absence of detectable SARS-CoV-2 RNA in serial BAL samples obtained starting 3 days after challenge. The lack of serological response to the SARS-CoV-2 challenge in BNT162b1-or BNT162b2-immunised macaques, despite a neutralising response to challenge in control-immunised macaques, suggests suppression of infection by the vaccine candidates. Clinical signs of disease were absent, and radiological and pathological abnormalities were generally mild after challenge. As in other published reports of immunisation and SARS-CoV-2 challenge of non-human primates (NHPs), there was no evidence of vaccine-mediated enhancement of viral replication, disease, or pathology 37, 38 . The interpretation of the vaccine-mediated NHP protection is limited by the small number of animals and the inherent limitations of animal models. Nevertheless, these preclinical results provided key support for the immunisation of large numbers of clinical trial participants with BNT162b2.

The selection of BNT162b2 over BNT162b1 for further clinical testing was largely driven by greater tolerability of BNT162b2 with comparable immunogenicity in clinical trials 3 and the broader range and MHC-diversity of T-cell epitopes on the much larger full-length spike. A global, pivotal, phase 3 safety and efficacy study of immunisation with BNT162b2 (NCT04368728) is ongoing and may answer those open questions that cannot be addressed by preclinical models.

Ethics statement

All mouse studies were performed at BioNTech SE, and protocols were approved by the local authorities (local welfare committee) and conducted according to Federation of European Laboratory Animal Science Associations recommendations. Study execution and housing were in compliance with the German Animal Welfare Act and Directive 2010/63/ EU. Mice were kept in individually ventilated cages in a 12 h light/dark cycle in a controlled environmental conditions (22 °C ± 2 °C, 45% to 65% relative humidity) under specific-pathogen-free (SPF) conditions. Food and water was available ad libitum. Only animals with an unobjectionable health status were selected for testing procedures.

Immunisations for the NHP study were performed at the University of Louisiana at Lafayette-New Iberia Research Centre (NIRC), which is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC, Animal Assurance #: 000452).

Protein and peptide reagents

Purified recombinant SARS-CoV-2 RBD (Sino Biological) or trimeric S protein (Acro Biosystems) was used as a target for western blot, and the RBD tagged with a human Fc (Sino Biological) was used in ELISA to detect SARS-CoV-2 S-specific IgG. A recombinant SARS-CoV-2 RBD containing a C-terminal Avitag™ (Acro Biosystems) was used as a target antigen in Luminex immunoassays. Purified recombinant SARS-CoV-2 S1 including a histidine tag (Sino Biological) was used in ELISA to detect SARS-CoV-2 S-specific IgG in mice. Purified recombinant SARS-CoV-2 S1 and RBD with histidine tags (both Sino Biological) were used for surface plasmon resonance (SPR) spectroscopy. A peptide pool of 15-mer peptides overlapping by 11 amino acids covering the full length S protein was used for re-stimulation in ELISpot, cytokine profiling and intracellular cytokine staining followed by flow cytometry. An irrelevant peptide (SPSYVYHQF, derived from gp70 AH-1 39 ) or a CMV peptide pool was used as control for ELISpot assays. All peptides were obtained from JPT Peptide Technologies.

Human convalescent serum panel

A panel of human SARS-CoV-2/COVID-19 convalescent sera described in previously published studies [1] [2] [3] was used as a benchmark for non-human primate serology. The sera (n=38) had been drawn from donors 18-83 years of age at least 14 days after PCR-confirmed diagnosis and at a time when the participants were asymptomatic. Most serum donors had outpatient (35/38) or inpatient (1/38) COVID-19; two of thirty-eight had asymptomatic SARS-CoV-2 infections. Sera were obtained from Sanguine Biosciences (Sherman Oaks, CA), the MT group (Van Nuys, CA) and Pfizer Occupational Health and Wellness (Pearl River, NY).

Cell culture

Human embryonic kidney (HEK)293T and Vero 76 cells (both ATCC) were cultured in Dulbecco's modified Eagle's medium (DMEM) with GlutaMAX™ (Gibco) supplemented with 10% fetal bovine serum (FBS [Sigma-Aldrich]). Cell lines were tested for mycoplasma contamination after receipt, before expansion and cryopreservation. For studies including NHP samples, Vero 76 and Vero CCL81 cells (both ATCC) were cultured in DMEM (Gibco) containing 2% HyClone fetal bovine and 100 U/mL penicillium/streptomycin (Gibco). Expi293F™ cells were grown in Expi293™ media and transiently transfected using Expi-Fectamine™293 (all from Thermo Fisher Scientific).

In vitro transcription and purification of RNA

Antigens encoded by BNT162b vaccine candidates were designed on a background of S sequences from SARS-CoV-2 isolate Wuhan-Hu-1 (Gen-Bank: MN908947.3). The DNA template for the BNT162b1 RNA is a DNA fragment encoding a fusion protein of the SARS-CoV-2 S signal peptide (SP, amino acids 1-16), the SARS-CoV-2 S RBD, and the T4 bacteriophage fibritin trimerisation motif 21 ('foldon'). The template for the BNT162b2 RNA is a DNA fragment encoding SARS-CoV-2 S (GenBank: MN908947) with K986P and V987P mutations. BNT162b1 and BNT162b2 DNA templates were cloned into a plasmid vector with backbone sequence elements (T7 promoter, 5′ and 3′ UTR, 100 nucleotide poly(A) tail) interrupted by a linker (A30LA70, 10 nucleotides) for improved RNA stability and translational efficiency 17, 40 . The DNA was purified, spectrophotometrically quantified, and in vitro transcribed by T7 RNA polymerase in the presence of a trinucleotide cap1 analogue ((m 2 7, 3'-O )Gppp(m 2'-O )ApG; TriLink) and with N 1 -methylpseudouridine-5'-triphosphate (m1ΨTP; Thermo Fisher Scientific) replacing uridine-5'-triphosphate (UTP) 41 . RNA was purified using magnetic particles 42 . RNA integrity was assessed by microfluidic capillary electrophoresis (Agilent Fragment Analyser), and the concentration, pH, osmolality, endotoxin level and bioburden of the solution were determined.

Lipid-nanoparticle formulation of the RNA

Purified RNA was formulated into LNPs using an ethanolic lipid mixture of ionisable cationic lipid and transferred into an aqueous buffer system via diafiltration to yield an LNP composition similar to one previously described 43 . The LNP contains RNA, an ionisable lipid, ((4-hydroxybutyl) azanediyl)bis(hexane-6,1-diyl)bis(2-hexyldecanoate)), a PEGylated lipid, 2-[(polyethylene glycol)-2000]-N,N-ditetradecylacetamide, and two structural lipids (1,2-distearoyl-sn-glycero-3-phosphocholine [DSPC] and cholesterol). The vaccine candidates were stored at -70 to -80 °C at a concentration of 0.5 mg/mL.

Transfection of HEK cells

HEK293T cells were transfected with 1 µg RiboJuice transfection reagent-mixed BNT162b1 RNA or BNT162b2 RNA or with the vaccine candidates BNT162b1 (LNP-formulated BNT162b1 RNA) or BNT162b2 (LNP-formulated BNT162b2 RNA) by incubation for 18 hours. Non-LNP formulated mRNA was diluted in Opti-MEM medium (Thermo Fisher Scientific) and mixed with the transfection reagent according to the manufacturer's instructions (RiboJuice, Merck Millipore).

Western blot analysis of size fractions of the medium of BNT162b1 RNA transfected cells

Medium from cultured HEK293T cells were collected. After 13-fold concentration via Vivaspin 20 centrifugal concentrators with a molecular weight cut off of 10 kDa, supernatants were applied to a preparative HiLoad ® 16/600 Superdex ® 200 pg column (both Sigma Aldrich). The column was run at 29.8 cm/h in phosphate buffered saline (PBS), and 500 µL fractions were collected (Supplementary Fig. 1 ). The gel filtration column was calibrated with well defined protein standards separated under identical conditions in a second run. Size fractioned FBS-free medium from BNT162b1 RNA-transfected HEK293T cells was analysed by denaturing (95 °C) and non-denaturating (no-heating) PAGE using 4-15% Criterion™ TGX Stain-Free™ Gel (Bio-Rad) and western blot. Transfer to a nitrocellulose membrane (Bio-Rad) was

W performed using a semi-dry transfer system (Trans-Blot Turbo Transfer System, Bio-Rad). Blotted proteins were detected with a monoclonal antibody that recognizes SARS-CoV-2 S1 (SinoBiological) and a secondary anti-rabbit horse radish peroxidase (HRP)-conjugated antibody (Sigma Aldrich). Blots were developed with Clarity Western ECL Substrate (Bio-Rad) and imaged with a Fusion FX Imager (Vilber) using the Image Lab software version 6.0.

Vaccine antigen detection by flow cytometry

Transfected HEK293T cells were stained with Fixable Viability Dye (eBioscience). After fixation (Fixation Buffer, Biolegend), cells were permeabilised (Perm Buffer, eBioscience) and stained with a monoclonal antibody that recognizes SARS-CoV-2 S1 (SinoBiological). Cells were acquired on a FACSCanto II flow cytometer (BD Biosciences) using BD FACSDiva software version 8.0.1 and analysed by FlowJo software version 10.6.2 (FlowJo LLC, BD Biosciences).

Localization of expressed vaccine antigens by immunofluorescence

Transfected HEK293T cells were fixed in 4% paraformaldehyde (PFA) and permeabilised in PBS/0.2% Triton X-100. Free binding sites were blocked and cells incubated with a rabbit monoclonal antibody that recognizes the SARS-CoV-2 S1 subunit (SinoBiological), an anti-rabbit IgG secondary antibody ( Jackson ImmunoResearch), labelled lectin HPA (Thermo Fisher Scientific) and concanavalin A (Fisher Scientific GmbH). DNA was stained with Hoechst (Life Technologies). Images were acquired with a Leica SP8 confocal microscope and Application Suite LAS-X Version 3.1.5.

SARS-CoV-2 RBD-foldon and P2 S expression and purification

To express the RBD-foldon encoded by BNT162b1 for ACE2 binding analysis and electron cryomicroscopy, DNA corresponding to the RNA coding sequence was cloned into the pMCG1309 vector. A plasmid encoding amino acids 1-615 of human ACE2 with C-terminal His-10 and Avi tags was generated for transient expression of the ACE2 peptidase domain (ACE2 PD) in Expi293F cells. The ACE2/B 0 AT1 complex was produced by co-expression of two plasmids in Expi293F cells, one of them encoding ACE2 amino acids 1-17 followed by haemagglutinin and Strep II tags and ACE2 amino acids 18-805, and the other containing a methionine followed by a FLAG tag and amino acids 2-634 of human B 0 AT1. Secreted ACE2 PD was isolated from conditioned cell culture medium using Nickel Excel resin (GE Healthcare) followed by gel filtration chromatography on a Superdex200 10/30 column (GE Healthcare) in PBS. Approximately 5 mg of purified ACE2 PD was covalently attached per 1 mL of 4% beaded agarose by amine coupling using AminoLink Plus resin (Thermo Fisher Scientific).

The RBD-trimer was purified from conditioned medium by affinity capture with the ACE2 PD crosslinked agarose and was eluted from the resin with 3 M MgCl 2 . Following dialysis, the protein was concentrated and purified by gel filtration using a Superdex200 10/300 column in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)-buffered saline (HBS) with 10% glycerol. Purification of the ACE2/B 0 AT1 complex was based on the procedure described previously 8 . To form the ACE2/ B 0 AT1/RBD-trimer complex, ACE2/B 0 AT1 aliquots were combined with purified RBD-foldon diluted in size exclusion chromatography buffer (25 mM Tris pH 8.0, 150 mM NaCl, 0.02% glyco diosgenin) for a 3:1 molar ratio of RBD-trimers to ACE2 protomers. After incubation at 4 °C for 30 minutes, the sample was concentrated and resolved on a Superose 6 Increase 10/300 GL column. Peak fractions containing the complex were pooled and concentrated.

To express SARS-CoV-2 P2 S encoded by BNT162b2 for characterisation by size exclusion chromatography, ACE2-PD binding, monoclonal antibody binding, and electron cryomicroscopy, a gene encoding the full length of SARS-CoV-2 (GenBank: MN908947) with two prolines substituted at residues 986 and 987 (K986P and V987P) followed with a C-terminal HRV3C protease site and a TwinStrep tag was cloned into a modified pcDNA3.1(+) vector with the CAG promoter. The TwinStrep-tagged P2 S was expressed in Expi293F cells.

Purification of the recombinant protein was based on a procedure described previously, with minor modifications 9 . Upon cell lysis, P2 S was solubilised in 1% NP-40 detergent. The TwinStrep-tagged protein was then captured with StrepTactin Sepharose HP resin in 0.5% NP-40. P2 S was further purified by size-exclusion chromatography and eluted as three distinct peaks in 0.02 % NP-40 as previously reported 9 . (Chromatogram not shown.) A peak that consists of intact P2 S migrating at around 150 kDa, as well as dissociated S1 and S2 subunits (which co-migrate at just above 75 kDa), was used in the structural characterisation. Spontaneous dissociation of the S1 and S2 subunits occurs throughout the course of protein purification, starting at the point of detergent-mediated protein extraction, so that P2 S preparations also contain dissociated S1 and S2.

Binding kinetics of the RBD-foldon trimer and P2 S to immobilised human ACE2 and a neutralizing monoclonal antibody by biolayer interferometry

Binding of purified RBD-foldon to the human ACE2 peptidase domain (ACE2 PD) and of NP-40 solubilised, purified P2 S to ACE2-PD and human neutralising monoclonal antibody B38 25 was measured by biolayer interferometry at 25 o C on an Octet RED384 (FortéBio). RBD-foldon binding was measured in 10 mM HEPES pH 7.5, 150 mM NaCl and 1 mM ethylenediaminetetraacetic acid (EDTA). P2 S binding was measured in 25 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA and 0.02% NP-40. Avi-tagged human ACE2 PD was immobilised on streptavidin-coated sensors; B38 antibody was immobilised on protein G-coated sensors. For a RBD-foldon concentration series, binding data were collected for 600 seconds of association and 900 seconds of dissociation. For a P2 S concentration series, after initial baseline equilibration of 120 seconds, the sensors were dipped in a 10 µg/mL solution of Avi-tagged ACE2-PD or B38 mAb for 300 seconds to achieve capture levels of 1 nM using the threshold function. Then, after another 120 seconds of baseline, binding data were collected for 300 seconds of association and 600 seconds of dissociation.

Biolayer interferometry data were collected with Octet Data Acquisition software version 10.0.0.87 and processed using ForteBio Data Analysis software version 10.0. Data were reference subtracted and fit to a 1:1 binding model with R 2 value greater than 0.96 for the RBD and 0.95 for P2 S. Potential avidity effects for the RBD-foldon and potential ongoing dissociation of S1 from P2 S could make the actual binding events more complicated than represented by 1:1 binding model. Therefore, we report apparent kinetics and affinity (P2 S) or avidity (RBD-foldon) of binding as calculated using Octet Data Analysis Software v10.0 (For-téBio). For the RBD-foldon, the dissociation rate of interaction (k d ) with ACE2-PD was slower than the limit of measurement of the instrument, and the apparent minimum binding avidity (K D ) was estimated using an assumed dissociation rate k d of 1 × 10 -6 s -1 .

Electron microscopy of negatively stained RBD-foldon trimers

Purified RBD-foldon in 4 µL was applied to a glow-discharged copper grid overlaid with formvar and amorphous carbon (Ted Pella). Negative staining was performed with Nano-W organotungstate stain (Nanoprobes) according to the manufacturer's protocol. The sample imaged using an FEI TF-20 microscope operating at 200 kV, with a magnification of 62,000x and defocus of -2.5 µm. Micrographs were contrast transfer function (CTF)-corrected in RELION using CTFFIND-4.1 44 . A small manually picked dataset was used to generate 2D references for auto-picking. The resulting particle set was subjected to 2D classification in RELION 3.0.6 45 .

Cryo-EM of the ACE2/B 0 AT1/RBD-trimer complex

Cryo-EM was performed using a Titan Krios operating at 300 keV equipped with a Gatan K2 Summit direct electron detector in

super-resolution mode at a magnification of 165,000x, for a magnified pixel size of 0.435 Å at the specimen level.

Purified ACE2/B 0 AT1/RBD-trimer complex at 6 mg/mL in 4 µL was applied to gold Quantifoil R1.2/1.3 200 mesh grids glow discharged in residual air for 30 seconds at 20 mA using a Pelco Easiglow. The sample was blotted using a Vitrobot Mark IV for 5 seconds with a force of -3 before being plunged into liquid ethane cooled by liquid nitrogen. In total, 7,455 micrographs were collected from a single grid. Data were collected over a defocus range of -1.2 to -3.4 µm with a total electron dose of 52.06 e -/Å 2 fractionated into 40 frames over a 6-second exposure for 1.30 e -/Å 2 /frame. Initial motion correction was performed in Warp 46 , during which super-resolution data were binned to give a pixel size of 0.87 Å. Corrected micrographs were imported into RELION 3.1-beta 45 for CTF estimation with CTFFIND-4.1 44 .

Particles were picked using the LaPlacian-of-Gaussian particle picking algorithm as implemented in RELION and extracted with a box size of 450 pixels. References obtained by 2D classification were used for a second round of reference-based auto-picking, yielding a dataset of 715,356 particles. Two of the three RBDs of each particle (the two not constrained by binding to ACE2/B 0 AT1) exhibited diffuse density in 2D classification that reflected high particle flexibility, consistent with the conformational flexibility of RBD trimers observed by negative stain EM (Fig. 1c, d ). This flexibility precluded the inclusion of all three RBDs in the final structural solution. Particle heterogeneity was filtered out with 2D and 3D classification with a mask size of 280 Å to filter out the diffuse density of the two non-ACE2-bound RBD copies in each RBD-trimer, yielding a set of 87,487 particles, which refined to 3.73 Å with C2 symmetry. Refinement after subtraction of micelle and B 0 AT1 density from the particles yielded an improved map of 3.24 Å. The atomic model from PDB ID 6M17 8 was rigid-body fitted into the 3.24 Å density and then flexibly fitted to the density using real-space refinement in Phenix 47 alternating with manual building in Coot 48 . The microscope was operated for image acquisition using SerialEM software version 3.8.0 beta 49 . Validation of this model is shown in Supplementary Fig. 2 . Data collection, 3D reconstruction and model refinement statistics are listed in Extended Data Table 1 .

Cryo-EM of P2 S

For TwinStrep-tagged P2 S, 4 µL purified protein at 0.5 mg/mL were applied to gold Quantifoil R1.2/1.3 300 mesh grids freshly overlaid with graphene oxide. The sample was blotted using a Vitrobot Mark IV for 4 seconds with a force of -2 before being plunged into liquid ethane cooled by liquid nitrogen. 27,701 micrographs were collected from two identically prepared grids. Data were collected from each grid over a defocus range of -1.2 to -3.4 µm with a total electron dose of 50.32 and 50.12 e -/Å 2 , respectively, fractionated into 40 frames over a 6-second exposure for 1.26 and 1.25 e -/Å 2 /frame. On-the-fly motion correction, CTF estimation, and particle picking and extraction with a box size of 450 pixels were performed in Warp 46 , during which super-resolution data were binned to give a pixel size of 0.87 Å. A total of 1,119,906 particles were extracted. All subsequent processing was performed in RELION 3.1-beta 45 . Particle heterogeneity was filtered out with 2D and 3D classification, yielding a set of 73,393 particles, which refined to 3.6 Å with C3 symmetry. 3D classification of this dataset without particle alignment separated out one class with a single RBD up, representing 15,098 particles. The remaining 58,295 particles, in the three RBD 'down' conformation, were refined to give a final model at 3.29 Å. The atomic model from PDB ID 6XR8 9 was rigid-body fitted into the map density, then flexibly fitted to the density using real-space refinement in Phenix 47 alternating with manual building in Coot 48 . The cryo-EM model validation is provided in Extended Data Fig. 2 , the full cryo-EM data processing workflow, and the model refinement statistics in Extended Data Table . 1.

Immunisation

Mice. Female BALB/c mice ( Janvier; 8-12 weeks) were randomly allocated to groups. BNT162b1 and BNT162b2 diluted in PBS with 300 mM sucrose (Fig. 2a-c and Extended Data Fig. 3 ; Fig. 2e and Extended Data Fig. 4a for BNT162b2) or 0.9% NaCl (Fig. 2d , Extended Data Fig. 4b-e ; Fig. 2e and Extended Data Fig. 4a for BNT162b1) were injected into the gastrocnemius muscle at a volume of 20 µL under isoflurane anaesthesia. PBS with 300 mM sucrose or 0.9% NaCl served as buffer controls, respectively.

Rhesus macaques (Macaca mulatta). Male rhesus macaques (2-4 years old) were randomly assigned to receive BNT162b1 or BNT162b2 on Days 0 and 21 or saline control on Days 0 and 21 or 35. Vaccine was administered in 0.5 mL by IM injection in the left quadriceps muscle. Animals were anesthetised with ketamine HCl (10 mg/kg; IM) during immunisation and were monitored for adequate sedation.

Phlebotomy and tissue preparation

Mice. Peripheral blood was collected from the retro-orbital venous plexus under isoflurane anaesthesia or vena facialis without anaesthesia. For flow cytometry, blood was heparinised. For serum generation, blood was centrifuged for 5 min at 16,000 x g, and the serum was immediately used for downstream assays or stored at -20 °C. Spleen single-cell suspensions were prepared in PBS by mashing tissue against the surface of a 70 µm cell strainer (BD Falcon). Erythrocytes were removed by hypotonic lysis. Popliteal, inguinal and iliac lymph nodes were pooled, cut into pieces, digested with collagenase D (1 mg/mL; Roche) and passed through cell strainers.

Rhesus macaques (Macaca mulatta). Serum was obtained before, 6 hours after, and 1, 14, 21, 28, 35 and 42 days after immunisation with BNT162b1, BNT162b2, or saline (Extended Data Table 2 ). For BNT162b2 and challenge cohort 3 controls, serum was also obtained on Day 56, and PBMCs were obtained before immunisation and on Days 7, 28, and 42, except that PBMCs were not obtained from the challenge cohort 3 control animals on Day 28. Blood for serum and PBMCs was collected in compliance with animal protocol 2017-8725-023 approved by the NIRC Institutional Animal Care and Use Committee. Animals were anesthetised with ketamine HCl (10 mg/kg; IM) during blood collection and were monitored for adequate sedation.

Analysis of S1-and RBD-specific serum IgG

Mice. MaxiSorp plates (Thermo Fisher Scientific) were coated with recombinant S1 or RBD (1 µg/mL) in sodium carbonate buffer, and serum-derived, bound IgG was detected using a horseradish peroxidase (HRP)-conjugated secondary antibody and tetramethylbenzidine (TMB) substrate (Biotrend). Data collection was performed using a BioTek Epoch reader and Gen5 software version 3.0.9. For concentration analysis, an IgG mouse isotype control was used in parallel in a serial dilution, and the sample signals were correlated to a standard curve of the isotype control.

Rhesus macaques (Macaca mulatta), humans. Recombinant SARS-CoV-2 S1 containing a C-terminal Avitag™ (Acro Biosystems) was bound to streptavidin-coated Luminex microspheres. Bound rhesus macaque or human anti-S1 antibodies present in the serum were detected with a fluorescently labelled goat anti-human polyclonal secondary antibody ( Jackson ImmunoResearch). Data were captured as median fluorescent intensities (MFIs) using a Bioplex200 system (Bio-Rad) and converted to U/mL antibody concentrations using a reference standard consisting of 5 pooled human COVID-19 convalescent serum samples (obtained >14 days PCR diagnosis, from the panel described above), diluted in antibody depleted human serum with arbitrary assigned concentrations of 100 U/mL and accounting for the serum dilution factor.

Surface plasmon resonance spectroscopy of polyclonal mouse immune sera

Binding kinetics of murine S1-and RBD-specific serum IgG to recombinant S1 and RBD was determined using a Biacore T200 device (Cytiva) with 10 mM Hepes, 150 mM NaCl, 3 mM EDTA, 0.05% v/v surfactant P20 (HBS-EP running buffer, BR100669, Cytiva) at 25 °C. Carboxyl

A C C E L E R A T E D A R T I C L E

P R E V I E W groups on the CM5 sensor chip matrix were activated with a mixture of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimidehydrochloride (EDC) and N-hydroxysuccinimide (NHS) to form active esters for the reaction with amine groups. Anti-mouse IgG Fc-antibody ( Jackson ImmunoResearch) was diluted in 10 mM sodium acetate buffer pH 5 (30 µg/mL) for covalent coupling to immobilisation level of ~10,000 response units (RU). Free N-hydroxysuccinimide esters on the sensor surface were deactivated with ethanolamine.

Mouse serum was diluted 1:50 in HBS-EP buffer and applied at 10 µL/min for 30 seconds to the active flow cell for capture by immobilised antibody, while the reference flow cell was treated with buffer. Binding analysis of captured murine IgG antibodies to S1-His or RBD-His (Sino Biological Inc.) was performed using a multi-cycle kinetic method with concentrations ranging from 25 to 400 nM or 1.56 to 50 nM, respectively. An association period of 180 seconds was followed by a dissociation period of 600 seconds with a constant flow rate of 40 µL/min and a final regeneration step. Apparent binding kinetics for the captured polyclonal IgG were calculated using a global kinetic fit model (1:1 Langmuir, Biacore T200 Evaluation Software Version 3.1, Cytiva).

VSV-SARS-CoV-2 S pseudovirus entry inhibition assay by serum IgG in mice

A recombinant replication-deficient vesicular stomatitis virus (VSV) vector that encodes green fluorescent protein (GFP) instead of VSV-G (VSVΔG-GFP) was pseudotyped with SARS-CoV-2 S according to published pseudotyping protocols 50, 51 . In brief, HEK293T/17 monolayers transfected to express SARS-CoV-2 S truncated of the C-terminal cytoplasmic 19 amino acids (SARS-CoV-2-S-CΔ19) were inoculated with VSVΔG-GFP vector (rescued from pVSVΔG-GFP plasmid expression vector; Kerafast Inc.). After incubation for 1 h at 37 °C, the inoculum was removed, and cells were washed with PBS before medium supplemented with anti-VSV-G antibody (clone 8G5F11, Kerafast Inc.) was added to neutralise residual input virus. VSV/SARS-CoV-2 pseudovirus-containing medium was harvested 20 h after inoculation, 0.2 µm filtered and stored at -80 °C.

Vero-76 cells were seeded in 96-well plates. Serial dilutions of mouse serum samples were prepared and pre-incubated for 10 min at room temperature with VSV/SARS-CoV-2 pseudovirus suspension (4.8 × 10 3 infectious units [IU]/mL) before transferring the mix to Vero-76 cells. Inoculated Vero-76 cells were incubated for 20 h at 37 °C. Plates were placed in an IncuCyte Live Cell Analysis system (Sartorius) and incubated for 30 min prior to the analysis (IncuCyte 2019B Rev2 software). Whole well scanning for brightfield and GFP fluorescence was performed using a 4× objective. The 50% pseudovirus neutralisation titre (pVNT 50 ) was reported as the reciprocal of the highest dilution of serum still yielding a 50% reduction in GFP-positive infected cell number per well compared to the mean of the no serum pseudovirus positive control. Each serum sample dilution was tested in duplicates.

IFNγ and IL-4 ELISpot

Mice. ELISpot assays were performed with mouse IFNγ ELISpot PLUS kits according to the manufacturer's instructions (Mabtech). A total of 5 × 10 5 splenocytes was ex vivo restimulated with the full-length S peptide mix (0.1 µg/mL final concentration per peptide) or controls (gp70-AH1 [SPSYVYHQF] 39 , 4 µg/mL; concanavalin A [ConA], 2 µg/mL [Sigma]). Streptavidin-alkaline phosphatase (ALP) and 5-bromo-4-chloro-3′-indolyl phosphate (BCIP)/nitro blue tetrazolium (NBT)-plus substrate were added, and spots counted using an ELISpot plate reader (ImmunoSpot® S6 Core Analyzer [CTL]). Spot numbers were evaluated using ImmunoCapture Image Acquisition Software V7.0 and ImmunoSpot 7.0.17.0 Professional. Spot counts denoted too numerous to count by the software were set to 1,500. For T-cell subtyping, CD8 + T cells and CD4 + T cells were isolated from splenocyte suspensions using MACS MicroBeads (CD8a [Ly-2] and CD4 [L3T4] [Miltenyi Biotec]) according to the manufacturer's instructions. CD8 + or CD4 + T cells (1 × 10 5 ) were subsequently re-stimulated with 5 × 10 4 syngeneic bone marrow-derived dendritic cells loaded with full-length S peptide mix (0.1 µg/mL final concentration per peptide), or cell culture medium as control. Purity of isolated T-cell subsets was determined by flow cytometry to calculate spot counts per 1 × 10 5 CD8 + or CD4 + T cells.

Rhesus macaques (Macaca mulatta). Rhesus macaque PBMCs were tested with commercially available NHP IFNγ and IL-4 ELISpot assay kits (Mabtech). Cryopreserved rhesus macaque PBMCs were thawed in pre-warmed AIM-V media (Thermo Fisher Scientific) with Benzonase (EMD Millipore). For IFNγ ELISpot, 1.0 x 10 5 PBMCs and for IL-4 ELISpot, 2.5 x 10 5 PBMCs were stimulated ex vivo with 1 µg/mL of the full-length S overlapping peptide mix. Tests were performed in triplicate wells and medium containing dimethyl sulphoxide (media-DMSO), a CMV peptide pool and phytohemagglutinin (PHA; Sigma) were included as controls. After 24 h for IFNγ and 48 h for IL-4, streptavidin-HRP and 3-amino-9-ethylcarbazole (AEC) substrate (BD Bioscience) were added and spots counted using a CTL ImmunoSpot S6 Universal Analyzer (CTL). Results shown are background (Medium-DMSO) subtracted and normalised to SFC/10 6 PBMCs.

Cell-mediated immunity by flow cytometry

Mice. For T-cell analysis in peripheral blood, erythrocytes from 50 µL freshly drawn blood were lysed (ammonium-chloride-potassium [ACK] lysing buffer [Gibco]), and cells were stained with Fixable Viability Dye (eBioscience) and primary antibodies in the presence of Fc block in flow buffer (Dulbecco's phosphate-buffered saline [Gibco] supplemented with 2% fetal calf serum (FCS), 2 mM ethylenediaminetetraacetic acid [both Sigma] and 0.01% sodium azide [Morphisto]). After staining with secondary biotin-coupled antibodies in flow buffer, cells were stained extracellularly against surface markers with directly labelled antibodies and streptavidin in Brilliant Stain Buffer Plus (BD Bioscience) diluted in flow buffer. Cells were washed with 2% RotiHistofix (Carl Roth), fixed (Fix/Perm Buffer, FoxP3/Transcription Factor Staining Buffer Set [eBioscience]) and permeabilised (Perm Buffer, FoxP3/Transcription Factor Staining Buffer Set [eBioscience]) overnight. Permeabilised cells were intracellularly treated with Fc block and stained with antibodies against transcription factors in Perm Buffer.

For T-cell analysis in lymphoid tissues, 1 × 10 6 lymph node cells (for BNT162b1) or 1.5 × 10 6 lymph node cells (for BNT162b2) and 4 × 10 6 spleen cells were stained for viability and extracellular antigens with directly labelled antibodies. Fixation, permeabilisation and intracellular staining was performed as described for blood T-cell staining.

For B-cell subtyping in lymphoid tissues, 2.5 × 10 5 lymph node and 1 × 10 6 spleen cells were treated with Fc block, stained for viability and extracellular antigens as described for blood T-cell staining and fixed with 2% RotiHistofix overnight.

For intracellular cytokine staining of T cells from BNT162b1immunised mice, 1 x 10 6 lymph node and 4 x 10 6 spleen cells were ex vivo restimulated with 0.2 µg/mL final concentration per peptide of full-length S peptide mix. For intracellular cytokine staining of T cells from mice immunised with BNT162b2, 4 x 10 6 spleen cells were ex vivo restimulated with 0.5 µg/mL final concentration per peptide of full-length S peptide mix or cell culture medium (no peptide) as control. The cells were restimulated for 5 hours in the presence of GolgiStop and GolgiPlug (both BD Bioscience) for 5 hours. Cells were stained for viability and extracellular antigens as described for lymphoid T-cell staining. Cells were fixed with 2% RotiHistofix and permeabilised overnight. Intracellular staining was performed as described for blood T-cell staining.

Mouse cells were acquired on a BD Symphony A3 or BD Celesta (B-cell subtyping) flow cytometer (BD Bioscience) using BD FACSDiva software version 9.1 or 8.0.1.1, respectively, and analysed with FlowJo 10.6 (FlowJo LLC, BD Biosciences).

Rhesus macaques (Macaca mulatta). For intracellular cytokine staining in T cells, 1.5 x 10 6 PBMCs were stimulated with the full-length S

Article peptide mix at 1 µg/mL (concentration of all peptides, combined), Staphyloccocus enterotoxin B (SEB; 2 µg/mL) as positive control, or 0.2% DMSO as negative control. GolgiStop and GolgiPlug (both BD Bioscience) were added. Following 37 °C incubation for 12 to 16 h, cells were stained for viability and extracellular antigens after blocking Fc binding sites with directly labelled antibodies. Cells were fixed, permeabilised with BDCytoFix/CytoPerm solution (BD Bioscience), and intracellular staining was performed in the permeabilisation buffer for 30 min at room temperature. Cells were washed, resuspended in 2% FBS/PBS buffer and acquired on an LSR Fortessa. Data were analysed by FlowJo 10.4.1 (FlowJo LLC, BD Biosciences). Results shown are background (media-DMSO) subtracted.

Cytokine profiling in mice by bead-based immunoassay

Mouse splenocytes were re-stimulated for 48 h with full-length S peptide mix (0.1 µg/mL final concentration per peptide) or cell culture medium (no peptide) as control. Concentrations of IFNγ, IL-2, IL-4, IL-5 and (for splenocytes from BNT162b2-immunised mice) IL-13 in supernatants were determined using a bead-based, 11-plex T H 1/T H 2 mouse ProcartaPlex multiplex immunoassay (Thermo Fisher Scientific) according to the manufacturer's instructions. Fluorescence was measured with a Bioplex200 system (Bio-Rad) and analysed with Pro-cartaPlex Analyst 1.0 software (Thermo Fisher Scientific). Values below the lower limit of quantification (LLOQ) were set to zero.

SARS-CoV-2 neutralisation by rhesus macaque (Macaca mulatta) sera

The SARS-CoV-2 neutralisation assay used a previously described strain of SARS-CoV-2 (USA_WA1/2020) that had been rescued by reverse genetics and engineered by the insertion of an mNeonGreen (mNG) gene into open reading frame 7 of the viral genome 27 . This reporter virus generates similar plaque morphologies and indistinguishable growth curves from wild-type virus. Viral master stocks were grown in Vero E6 cells as previously described 52 . When testing human convalescent serum specimens, the fluorescent neutralisation assay produced comparable results to the conventional plaque reduction neutralisation assay. Serial dilutions of heat-inactivated sera were incubated with the reporter virus (2 x 10 4 plaque forming units [PFU] per well) to yield an approximately 10-30% infection rate of the Vero CCL81 monolayer for 1 h at 37 °C before inoculating Vero CCL81 cell monolayers (targeted to have 8,000 to 15,000 cells in the central field of each well at the time of seeding, one day before infection) in 96-well plates to allow accurate quantification of infected cells. Cell counts were enumerated by nuclear stain (Hoechst 33342), and fluorescent virus-infected foci were detected 16-24 hours after inoculation with a Cytation 7 Cell Imaging Multi-Mode Reader (BioTek) with Gen5 Image Prime version 3.09. Titres were calculated in GraphPad Prism version 8.4.2 by generating a 4-parameter (4PL) logistical fit of the percent neutralisation at each serial serum dilution. The 50% neutralisation titre (VNT 50 ) was reported as the interpolated reciprocal of the dilution yielding a 50% reduction in fluorescent viral foci.

SARS-CoV-2 challenge of rhesus macaques (Macaca mulatta)

The SARS-CoV-2 inoculum was obtained from a stock of 2.1 × 10 6 PFU/mL previously prepared at Texas Biomedical Research Institute (San Antonio, TX), aliquoted into single use vials, and stored at -70 °C. The working virus stock was generated from two passages of the SARS-CoV-2 USA-WA1/2020 isolate (a 4 th passage seed stock purchased from BEI Resources; NR-52281) in Vero E6 cells. The virus was confirmed to be SARS-CoV-2 by deep sequencing that demonstrated identity to a published SARS-CoV-2 sequence (GenBank accession number MN985325.1).

BNT162b1-immunised (n=6), BNT162b2-immunised (n=6), and age-matched saline-immunised (n=9) male rhesus macaques (control) were challenged with 1.05 × 10 6 plaque forming units of SARS-CoV-2 USA-WA1/2020 isolate, split equally between the intranasal (IN; 0.25 mL) and intratracheal (IT; 0.25 mL) routes as previously described 28 . Sentinel age-and sex-matched animals (n=6) were mock challenged with DMEM supplemented with 10% FCS IN (0.25 mL) and IT (0.25 mL). The macaques were challenged or mock challenged at the times relative to immunisation indicated in Extended Data Fig. 6 and Extended Data Table 2 .

Twelve to nineteen days prior to challenge, animals were moved from the NIRC, in New Iberia, LA, where they had been immunised, to the animal biosafety level 3 facility at SNPRC (in San Antonio, TX). Animals were monitored regularly by a board-certified veterinary clinician for rectal body temperature, weight and physical examination. Specimen collection was performed under tiletamine zolazepam (Telazol) anaesthesia as described 28 . Bronchoalveolar lavage (BAL), nasal, OP and rectal swab collection, X-ray and CT examinations and necropsy were performed at the times indicated in Extended Data Fig. 6 and Extended Data Table 2 . The 3 control animals in challenge cohort 3 and 3 sentinel animals were not necropsied to allow their subsequent re-challenge (control) or challenge (sentinel). BAL was performed by instilling 20 mL of saline 4 times. These washings were pooled, aliquoted and stored frozen at -70 °C.

SARS-CoV-2 viral RNA quantification by reverse-transcription quantitative polymerase chain reaction

To detect and quantify SARS-CoV-2 in NHP, viral RNA was extracted from BAL fluid and from nasal, OP, and rectal swabs as previously described [53] [54] [55] and tested by RT-qPCR as previously described 28 . Briefly, 10 µg yeast tRNA and 1 × 10 3 PFU of MS2 phage (Escherichia coli bacteriophage MS2, ATCC) were added to each thawed sample, and RNA extraction performed using the NucleoMag Pathogen kit (Macherey-Nagel). The SARS-CoV-2 RT-qPCR was performed on extracted RNA using a CDC-developed 2019-nCoV N1 assay on a QuantStudio 3 instrument (Applied Biosystems). The cut-off for positivity (limit of detection, LOD) was established at 10 gene equivalents (GE) per reaction (800 GE/mL). Samples were tested in duplicate. One BAL specimen from the challenge cohort 2 control group obtained on Day 6 after challenge and one nasal swab from the BNT162b1-immunised group obtained on Day 1 after challenge had, on repeated measurements, viral RNA levels on either side of the LLOD. These specimens were categorised as indeterminate and excluded from the graphs and the analysis.

Radiology

Thoracic radiographs and computed tomography (CT) scans were performed under anesthesia as previously described 28 . For radiographic imaging, 3-view thoracic radiographs (ventrodorsal, right and left lateral) were obtained at the times relative to challenge indicated in Extended Data Table 2 . The animals were anesthetized using Telazol (2-6 mg/kg) and maintained by inhaled isoflurane delivered through a Hallowell 2002 ventilator anesthesia system (Hallowell, Pittsfield, MA). Animals were intubated to perform end inspiratory breath-hold using a remote breath-hold switch. Lung field CT images were acquired using Multiscan LFER150 PET/CT (MEDISO Inc., Budapest, Hungary) scanner. Image analysis was performed using 3D ROI tools available in Vivoquant (Invicro, Boston, MA). Images were interpreted by a board-certified veterinary radiologist blinded to treatment groups. Scores were assigned to a total of 7 lung regions on a severity scale of 0-3 per region, with a maximum severity score of 21. Pulmonary lesions evident prior to challenge, or those which could not be unequivocally attributed to the viral challenge (such as atelectasis secondary to recumbency and anesthesia) received a score of "0".

Histopathology

Lung histopathology is reported on necropsies performed on 2-4 year old male rhesus macaques at the times after challenge indicated in Extended Data Fig. 6 and Extended Data Table 2 . Necropsy, tissue processing, and histology were performed by SNPRC in San Antonio,

Samples were fixed in 10% neutral buffered formalin and processed routinely into paraffin blocks. Tissue blocks were sectioned to 5 µm and stained with hematoxylin and eosin. Microscopic evaluation of 7 lung tissue sections per animal (1 sample of each lobe on L & R) was performed blindly by SNPRC and Pfizer pathologists. Lungs were evaluated using a semi-quantitative scoring system with inclusion of cell types and/or distribution as appropriate. Inflammation score was based on area of tissue in section involved: 0 = normal; 1=<10%; 2=11-30%; 3=30-60%; 4= 60-80%; 5=>80%. Each lobe received an individual score, and the final score for each animal was reported as the mean of the individual scores. The pathologists were unblinded to the group assignments after agreement on diagnoses. As indicated in Extended Data Fig. 6 and Extended Data Table 2 , the BNT162b1-immunised and control macaques were challenged and necropsied in parallel (challenge cohorts 1 and 2), and the BNT162b2-immunised rhesus macaques were immunised and challenged subsequently (challenge cohort 3).

Statistics and reproducibility

No statistical methods were used to predetermine group and samples sizes (n). All experiments were performed once. P-values reported for RT-qPCR analysis were determined by nonparametric analysis (Friedman's test) based on the ranking of viral RNA shedding data within each day. PROC RANK and PROC GLM from SAS® 9.4 were used to calculate the p-values. All available post-challenge BAL fluid and nasal, OP, and rectal swab samples from the necropsied animals and all available post-challenge samples through Day 10 from the animals not necropsied were included in the analysis. Indeterminate results were excluded from this analysis. All remaining analyses were two-tailed and carried out using GraphPad Prism 8.4. 7 | Viral RNA detection in oropharyngeal (OP) and rectal swabs and clinical signs in BNT162b-immunised rhesus macaques after challenge with infectious SARS-CoV-2. Rhesus macaques immunised with 100 µg of BNT162b1 or BNT162b2 (n=6 each) and macaques immunised with saline or not immunised (Control, n=9), as described in Fig. 4 , Extended Data Fig. 6 , and Extended Data Table 2 , were challenged with 1.05 × 10 6 total plaque forming units (PFU) of SARS-CoV-2 split equally between the intranasal (IN) and intratracheal (IT) routes. Additional macaques (sentinel, n=6) were mock-challenged with cell culture medium. a,b, Viral RNA levels were detected by RT-qPCR in a, OP swabs and b, rectal swabs. Ratios above data points indicate the number of viral RNA positive animals among all animals providing evaluable samples in a group. Heights of bars indicate geometric mean of viral RNA copies; whiskers indicate geometric standard deviations. Every symbol represents one animal. Dotted lines indicate the lower limits of detection (LLODs). Values below the LLOD were set to ½ the LLOD. The two-sided statistical significance by Friedman's non-parametric test of differences in viral RNA detection between 6 BNT162b1-immunised and 6 contemporaneously control-immunised animals (challenge cohorts 1 and 2) after challenge was p < 0.0001 for OP swabs and p = 0.1179 for rectal swabs; between 6 BNT162b2immunised animals and 3 contemporaneously control-immunised animals (challenge cohort 3) after challenge, the statistical significance was p = 0.0007 for OP swabs and p = 0.2209 for rectal swabs. 8 | Radiographic signs and pulmonary histopathology of rhesus macaques immunised with BNT162b1 or BNT162b2 and challenged with SARS-CoV-2. Rhesus macaques were immunised with BNT162b1, BNT162b2, or saline (control) and challenged with SARS-CoV-2, and a sentinel group was challenged with cell culture medium. The disposition of the animals for immunisation, infectious challenge, imaging, and necropsy are described in Figs. 3 and 4 , Extended Data Fig. 6 , and Extended Data Table 2 . Three-view thoracic radiographs (ventrodorsal, right and left lateral) and lung field CT images were obtained. The animals were anesthetised and intubated to perform end inspiratory breath-hold. Images were interpreted by two board-certified veterinary radiologists blinded to treatment groups. Scores were assigned to 7 lung regions on a severity scale of 0-3 per region, with a maximum severity score of 21. Pulmonary lesions evident prior to challenge or those which could not be unequivocally attributed to the viral challenge (such as atelectasis secondary to recumbency and anesthesia) received a score of "0". a, Thoracic radiograph scores. b, Lung field CT scores. Each dot represents the summed radiograph or CT scores for the 7 lung lobes of a single animal. Two veterinary pathologists blindly performed microscopic evaluation of formalin fixed, hematoxylin and eosin stained lung tissue sections from each of 7 lobes from each macaque that had been necropsied on Day 7 or 8. Inflammation scores were assigned by consensus between the pathologists on a scale of 1-5 based on the area of involvement. c, Pulmonary histopathology scores. Each dot represents the mean inflammation area score from the 7 lung lobes of an individual animal. For a, b, and c, the height of each bar indicates the arithmetic mean of the radiograph, CT, or histopathology score, respectively, for the animals in each group, and whiskers indicate standard deviations. All macaques in the BNT162b1, BNT162b2, and control challenge groups were challenged with SARS-CoV-2. Macaques in the sentinel challenge group were mock challenged. 2 "-" indicates no immunisation. 3 Challenge cohort 2 was challenged with SARS-CoV-2 or mock challenged one week after challenge cohort 1. Challenge cohort 3 was challenged with SARS-CoV-2 or mock challenged 6 weeks after challenge cohort 2. 4 All macaques were challenged with SARS-CoV-2 or mock challenged, according to their challenge group. The entry for "Days from Dose 2 to SARS-COV-2 or mock challenge" for macaques that were not immunised is "-". Clinical trial registration This manuscript describes preclinical studies. See the clinical reports for background to the referenced clinical data.

Study protocol

This manuscript describes preclinical studies. See the clinical reports for background to the referenced clinical data.

Data collection

This manuscript describes preclinical studies. See the clinical reports for background to the referenced clinical data.

Outcomes

This manuscript describes preclinical studies. See the clinical reports for background to the referenced clinical data.

Flow Cytometry Plots

Confirm that:

The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).

The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).

All plots are contour plots with outliers or pseudocolor plots.

A numerical value for number of cells or percentage (with statistics) is provided.

Methodology Sample preparation

Mice: For flow cytometry or ELISpot, blood or cell suspensions from lymph node and spleen were used directly or after peptide stimulation. Peripheral blood was collected from the retro-orbital venous plexus or vena facialis under isoflurane anaesthesia. Spleen single-cell suspensions were prepared in PBS by mashing tissue against the surface of a 70 μm cell strainer (BD Falcon) using the plunger of a 3-mL syringe (BD Biosciences). Erythrocytes were removed by hypotonic lysis. Popliteal, inguinal and iliac lymph nodes were pooled, cut into pieces, digested with collagenase D (1 mg/mL; Roche) and passed through cell strainers. For intracellular stains, cells were fixed and permeabilized using the eBioscience™ Foxp3/ Transcription Factor Staining Buffer Set. NHP: Blood for serum and PBMCs was collected in compliance with animal protocol 2017-8725-023 approved by the NIRC Institutional Animal Care and Use Committee. Animals were anesthetised with ketamine HCl (10 mg/kg; IM) during blood collection and immunisation, and monitored for adequate sedation.

Instrument

Cell culture cells were acquired on a FACSCanto II flow cytometer (BD Biosciences). Mouse cells were acquired on a BD Symphony A3 or BD Celesta (B-cell subtyping) flow cytometer (BD Bioscience). NHP cells were analyzed on a LSR Fortessa X-20

Software

Cell culture cells were analyzed using BD FACSDiva software version 8.0.1 and analysed by FlowJo software version 10.6.2 (FlowJo LLC, BD Biosciences). Mouse cells were analyzed using BD FACSDiva software version 9.1 or 8.0.1.1, respectively, and analysed with FlowJo 10.6 (FlowJo LLC, BD Biosciences). NHP cells were analyzed using FlowJo (10.4.1).

Cell population abundance

Sorted CD4 CD8 T cells from mouse were confimed following magnetic bead separation.

Gating strategy

The gating strategies are detailed in the supplementary information. Mouse: Flow cytometry gating strategy for the identification of IFNγ, IL-2, and TNF secreting CD8+ T cells in the mouse spleen. was performed after CD8+ T cells were gated within single, viable lymphocytes, excluding CD4+ T cells. Flow cytometry gating strategy for identification of TFH cells, activated T cells and B cells in lymph nodes and the spleen was performed by CD3+CD19-T cells gating within single, viable lymphocytes. CD4+ and CD8+ T cells were gated from CD3+ cells; TFH cells were gated from CD4+ T cells and defined as CD4+ T-bet-GATA3-CD44+ CD62L-PD-1+ CXCR5+ cells. Flow cytometry gating strategy for the identification of B cells in lymph nodes and the spleen was done by gating activated B cells within single, viable lymphocytes defined as IgD-Dump (CD4, CD8, F4/80, GR-1)-cells. Plasma cells (PC) were gated from activated B cells and defined as CD138+ B220low/-cells. Switched B cells were gated from non-PC and defined as CD19+ CD138-IgM-. Germinal centre (GC) and IgG1+ and IgG2a+ B cells were gated from switched B cells and defined as CD19+ IgM-CD38-CD95+ and CD19+ IgM-IgG1+/IgG2a+, respectively. Flow cytometry gating strategy for the identification of T cells, B cells and TFH cells in peripheral blood was performed by gating CD3+ CD19-T cells within single, viable lymphocytes. CD4+ and CD8+ T cells were gated from CD3+ CD19-cells. TFH cells were gated from CD4+ T cells and defined as CD4+ T-bet-GATA3-CD44+ CD62L-PD-1+ CXCR5+ cells.

Rhesus macaque: Flow cytometry gating strategy for identification of spike-specific SARS-CoV-2 modRNA vaccine BNT162b2-induced T cells started with events acquired with a constant flow stream and fluorescence intensity, viable cells, lymphocytes and single events were identified and gated. Within singlet lymphocytes, CD20-CD3+ T cells were identified and gated into CD4+ T cells and CD8+ T cells. Antigen-specific CD4+ T cells were identified by gating on CD154 and cytokine-positive cells, and CD8+ T cells were identified by gating on CD69 and cytokine-positive cells. The antigen-specific cells were used for further analysis. Flow cytometry gating strategy for the identification of HEK293T cells transfected with BNT162b1 or BNT162b2, or BNT162b1-RNA or BNT162b2-RNA using a transfection reagent or no RNA (control) was performed by gating S1+ HEK293T cells within single, viable HEK293T cells. Tick this box to confirm that a figure exemplifying the gating strategy is provided in the Supplementary Information.

Data availability

The SARS-CoV-2 isolate Wuhan-Hu-1 (GenBank: MN908947.3) is the genetic background of the BNT162b antigens. The cryo-EM maps and atomic coordinates have been deposited to the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) with accession numbers EMD-23211 and PDB 7L7F for the ACE2/B0AT1/RBD-foldon complex and EMD-23215 and PDB 7L7K for P2 S. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Data Policy information about availability of data

All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:

-Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. The vaccine sequence is based on GenBank: MN908947.3 For P2 S, the atomic model from PDB ID 6XR8 was rigid-body fitted into the map density (https://www.rcsb.org/structure/6XR8).

The cryo-EM maps and atomic coordinates have been deposited to the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) with accession numbers EMD-23211 and PDB 7L7F for the ACE2/B0AT1/RBD-foldon complex and EMD-23215 and PDB 7L7K for P2 S.

annex

Author contributions U.S. conceived and conceptualised the work and strategy. S.H., S.C.D., A.A.H.S., C.G., R.d.l.C.G.G., and M.C.G. designed primers, performed oligosynthesis, cloned constructs and performed protein expression experiments. T.Z., S.F., J.S. and A.N.K. developed, planned, performed and supervised RNA synthesis and analysis. E.H.M. purified P2 S. N.L.N. purified RBD-trimer and ACE2 PD. J.A.L. developed ACE2/B 0 AT1/RBD-trimer formation and purified the complex. P.V.S. developed and performed biolayer interferometry experiments. J.A.L. and S.H. performed electron microscopy and solved the structure of the complex. Y.C. supervised the structural and biophysical characterisation and analysed the structures. A.M. and B.G.L. performed surface plasmon resonance spectroscopy. A.G., S.A.K, S.S., T.H., L.F. and F.V. planned, performed and analysed in vitro studies. F.B., T.K., C.R. managed formulation strategy. A.B.V., M.V., L.M.K., K.C.W. designed mouse studies, analysed and interpreted data. A.P., S.E., D.P. and G.S. performed and analysed the S1-and RBD-binding IgG assays. M.G. designed and optimized MS2-SARS-nCoV-2-N1 RT-qPCR assay. M.G., R.C., Jr., and K.J.A. Funding BioNTech is the Sponsor of the study, and Pfizer is its agent. BioNTech and Pfizer are responsible for the design, data collection, data analysis, data interpretation, and writing of the report. The corresponding authors had full access to all the data in the study and had final responsibility for the decision to submit the data for publication. This study was not supported by any external funding at the time of submission.

Competing interests

Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics

For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed

The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly

The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated Our web collection on statistics for biologists contains articles on many of the points above.

Software and code

Policy information about availability of computer code Data collection Western Blot: Fusion FX Imager (Vilber), Image Lab software version 6.0 For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Field-specific reporting

Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences

Behavioural & social sciences Ecological, evolutionary & environmental sciences

For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.

Sample size

No sample-size calculation was performed.

For mouse studies, a two-sided test with the hypothesis that the mean is a given value, with the shift to be detected a multiple of the standard deviation was taken into account. For α=0.05 and a desired power of 80%, a group size of n = 8 is required to find significant differences between groups. For non-human primate studies, sample size was limited by availability of suitable animals.

Data exclusions No data was excluded.

Replication

Replication was not attempted. Independent studies and analysis methods to analyse immune responses were performed.

Randomization Mice or NHP were randomly allocated to groups. No formal randomization was done.

Blinding

Studies were performed unblinded as analyis results could not be manipulated by interpretation. Thoracic radiographs and computed tomography scan images were interpreted by a board-certified veterinary radiologist blinded to treatment groups. Histopathology slides were read by veterinary pathologists blinded to treatment groups.

Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study.

Wild animals

The study did not involve wild animals.

Field-collected samples The study did not involve animals collected from the field. Note that full information on the approval of the study protocol must also be provided in the manuscript.

Recruitment

This manuscript describes preclinical studies. See the clinical reports for background to the referenced clinical data.

Ethics oversight

This manuscript describes preclinical studies. See the clinical reports for background to the referenced clinical data.

Note that full information on the approval of the study protocol must also be provided in the manuscript.

Acknowledgements

Acknowledgements We thank T. Garretson and D. Cooper for advice on and M. Cutler for coordination of NHP serology studies.We thank R. F. Sommese and K. F. Fennell for technical assistance for molecular cloning and cell-based binding.Valuable support and assistance M. Dvorak, M. Drude, F. Zehner, T. Lapin, B. Ludloff, S. Hinz, F. Bayer, J. Scholz, A.L. Ernst, T. Sticker and S. Wittig resulted in a rapid availability of oligonucleotides and DNA templates.E. Boehm, K. Goebel, R. Frieling, C. Berger, S. Koch, T. Wachtel, J. Leilich, M. Mechler, R. Wysocki, M. Le Gall, A. Czech and S. Klenk carried out RNA production and analysis.Without their commitment during this pandemic situation, this vaccine candidate could not have been transferred to non-clinical studies in light speed.B. Weber, J. Vogt, S. Krapp, K. Zwadlo, J. Mottl, J. Mühl and P. Windecker supported the mouse studies and serological analysis with excellent technical assistance.We thank radiologists A. K. Voges and E. Clemmons for interpreting radiographs for the nonhuman primate study and S. Ganatra for radiology services.We thank E. Romero for veterinary services and K.A. Soileau and the staff of the New Iberia Research Centre for non-human primate care.We thank E.J. Dick for veterinary pathology.We thank Polymun Scientific for excellent formulation services as well as Acuitas Therapeutics for fruitful discussions.We thank S. Wigge and C. Lindemann for scientific writing support and manuscript review.

Failed to load PDF:

Paper sources

Abstract screening pilot

Abstract screening results

Extraction pilot

Extraction results

Research report

Modify setup

StripeM-Inner