Subunit vaccines are among the safest and most widely used vaccines, and have been highly effective against a variety of infectious diseases, such as hepatitis B, diphtheria, pertussis, tetanus and shingles in diverse age groups2. Therefore, the development of a safe and effective subunit vaccine against SARS-CoV-2 would represent an important step in controlling the COVID-19 pandemic. An essential component of subunit vaccines is the adjuvant, an immune-stimulatory agent that enhances the magnitude, quality and durability of the immune responses induced by vaccination, even with lower doses of antigen3. Alum, the most widely used adjuvant, has been used in billions of vaccine doses over the last century. During the past two decades, novel adjuvants have been developed including the α-tocopherol-containing squalene-based oil-in-water adjuvant AS03 and the TLR9 ligand CpG1018, which are included in licensed vaccines against pandemic influenza and hepatitis B, respectively. In particular, AS03 and CpG1018 are currently being developed as adjuvants for use in candidate subunit SARS-CoV-2 vaccines3,4; however, their capacity to stimulate protective immunity against SARS-CoV-2 remains unknown. In this Article, we evaluate the use of AS03, CpG1018 formulated in alum (CpG-alum), squalene-in-water emulsion (O/W), AS37 and alum as adjuvants for a subunit vaccine in which 60 copies of the SARS-CoV-2 RBD are displayed in a highly immunogenic array using a computationally designed self-assembling protein nanoparticle (RBD–NP)5, and demonstrate protective immunity against SARS-CoV-2 in non-human primates.

Robust and durable neutralizing-antibody responses

To assess the immunogenicity and protective efficacy of RBD–NP vaccination with different adjuvants, we immunized 29 rhesus macaques with 25 μg RBD antigen (71 μg total RBD–NP immunogen) (Extended Data Fig. 1a–d) formulated with one of the five adjuvants: O/W, AS03, AS37, CpG-alum or alum (Extended Data Fig. 1e). Four additional monkeys were administered saline as a control (Extended Data Table 1). All immunizations were administered via the intramuscular route in forelimbs on days 0 and 21. Four weeks after the booster immunization, the monkeys were challenged with SARS-CoV-2 via the intratracheal–intranasal routes. Five of the ten monkeys immunized with RBD–NP-AS03 were not challenged, to enable longitudinal analysis of the durability of the immune responses.

Evaluation of binding-antibody responses showed that spike-specific IgG was detected 21 days after primary immunization in all vaccination groups and the response increased in magnitude after boosting (Fig. 1a). RBD–NP-AS03 induced the highest magnitude (geometric mean titre (GMT) half-maximal effective concentration (EC50) 1:8,551) and RBD–NP-O/W induced the lowest (GMT EC50 1:1,308) on day 42. The I53-50 nanoparticle scaffold also elicited an antibody response in all groups, which correlated with spike-specific antibody responses (Extended Data Fig. 1f, g). Immunization with RBD–NP induced detectable neutralizing-antibody responses against a SARS-CoV-2 spike pseudotyped virus6 in most of the monkeys (except in the RBD–NP-O/W group after primary immunization), which significantly increased in all groups after the booster immunization (Extended Data Fig. 2a). The neutralizing-antibody titres in all groups except the RBD–NP-O/W group were higher than that of 4 samples from convalescent human individuals who had suffered mild-to-moderate COVID-19 (GMT 1:76) and NIBSC control human convalescent plasma (NIBSC code 20/130, neutralizing-antibody titre 1:241) (Extended Data Fig. 2b) assayed simultaneously. Next, we measured neutralizing-antibody responses against authentic SARS-CoV-2 virus using focus reduction neutralization titre (FRNT) assay7, which was used to analyse the recent clinical trials of the Moderna mRNA SARS-CoV-2 vaccine8,9. Consistent with the pseudovirus neutralizing-antibody titres, vaccinations with all adjuvants induced robust neutralizing-antibody titres against live virus after the secondary immunization, with the RBD–NP-AS03 group showing the highest titres (GMT 1:4,145, Fig. 1b). There was a strong correlation between pseudovirus and live-virus neutralizing-antibody titres, as observed in previous studies10,11 (Extended Data Fig. 2c). In addition, there was a robust induction of RBD–NP-specific plasmablast responses four days after secondary immunization (Extended Data Fig. 2d), the magnitude of which correlated with the observed antibody responses (Extended Data Fig. 2e).

Fig. 1: RBD–NP immunization induces robust antibody responses.
figure 1

a, SARS-CoV-2 spike-specific IgG titres in sera following immunization with RBD–NP and the indicated adjuvant measured by ELISA. Boxes represent median, and 25th and 75th percentiles and whiskers show the range. b, c, Serum neutralizing-antibody titres determined by authentic SARS-CoV-2 assay at indicated time points. The black line represents the geometric mean of all data points and numbers show GMT on day 42 (b) and subsequent time points (c). IC50, half-maximal inhibitory concentration. d, Neutralizing-antibody titres against live wild-type (circles) or B.1.1.7 or B.1.351 variant (squares) SARS-CoV-2 virus measured in serum on day 42. Numbers indicate GMT. e, The fold change in neutralizing-antibody titres for the B.1.351 variant versus wild-type SARS-CoV-2. f, Serum neutralizing-antibody titres against the live wild-type (circles) or B.1.351 variant (squares) viruses measured on day 42 or day 154. a, b, e, Two-sided Mann–Whitney rank-sum test. c, f, Two-sided Wilcoxon matched-pairs signed-rank test. Each symbol represents one monkey. n = 4 (O/W), 10 (AS03) and 5 (all other groups).

Source data

To determine the durability of the neutralizing-antibody responses, we monitored 5 monkeys immunized with RBD–NP-AS03 without challenge for 6 months. The pseudovirus neutralizing-antibody titres declined moderately until day 126, but did not differ significantly between days 42 and 126 (Extended Data Fig. 2f). Neutralizing-antibody response measured against the authentic SARS-CoV-2 virus was durably maintained up to day 180 (Fig. 1c). The GMT titres decreased by 7.2-fold between days 42 and 180. Furthermore, we observed little to no reduction in the efficiency of blocking of ACE2 binding to the RBD—a correlate of neutralizing-antibody response12—by sera collected at these time points (Extended Data Fig. 2g). These results demonstrate that the RBD–NP-AS03 immunization induces potent and durable neutralizing-antibody responses.

Neutralization of variants of concern

SARS-CoV-2 variants of concern, B.1.1.7 and B.1.351, first identified in the United Kingdom and South Africa, respectively, have since been found to be circulating globally. We used live virus and pseudovirus neutralization assays to evaluate sera from the immunized monkeys against these variants. Neutralizing-antibody titres against the B.1.1.7 variant were similar to those against the wild-type virus (Fig. 1d, left, Extended Data Fig. 2h); by contrast, titres against the B.1.351 variant were considerably lower (Fig. 1d, right, Extended Data Table 2), as seen in vaccinated humans13,14. Of note, there was a greater reduction of titre in the RBD–NP-AS37 group (median of 16-fold) compared with RBD–NP-AS03 (4.5-fold) and other adjuvant groups (Fig. 1e). These data suggest that the adjuvants not only enhance immunogenicity, but also that different adjuvants vary in their potential to elicit neutralizing antibodies that provide a greater breadth of neutralization. Furthermore, the neutralizing-antibody response against the B.1.351 variant was as durable as that of the responses to the wild-type virus in the RBD–NP-AS03 durability group (Fig. 1e).

Induction of CD4 T cell responses

RBD–NP immunization resulted in an antigen-specific CD4 T cell response but limited CD8 T cell response following ex vivo stimulation of peripheral blood mononuclear cells (PBMCs) with an overlapping peptide pool. RBD-specific CD4 responses were significantly enhanced only after the secondary immunization and were highest in the RBD–NP-AS03 and RBD–NP-CpG-alum groups (Fig. 2a, b). The responses were dominated by IL-2- and TNF-secreting cells (Extended Data Fig. 3a), which remained detectable at day 42 (3 weeks after secondary immunization). There was a low but detectable IL-4 response in both the RBD–NP-AS03 and RBD–NP-CpG-alum groups that peaked on day 28 but declined to near-baseline levels by day 42 (Fig. 2b). Whereas 75% of monkeys in the RBD–NP-alum group and 50% of those in the RBD–NP-O/W group showed induction of RBD-specific CD4 T cells, RBD–NP-AS37 induced a weak T cell response, less than that of the alum group, despite inducing potent antibody responses in all of the monkeys. This is in contrast to findings in mice with the same antigen (L. Grigoryan et al., manuscript in preparation) and in non-human primates with an HIV antigen15. However, a direct comparison with the alum group is confounded by the larger amount of aluminium hydroxide in the alum group compared with the RBD–NP-AS37 group (Extended Data Fig. 1e). We assessed the polyfunctional profile of antigen-specific CD4 T cells expressing IL-2, IFNγ, IL-4 and TNF (Extended Data Fig. 3b). Although IL-2+, TNF+ and IL-2+TNF+ cells formed the majority (around 70%) in all adjuvant groups, differences between the groups were apparent. RBD–NP-AS03 elicited similar proportions of T helper 1 (TH1) and TH2 CD4 T cells, a balanced TH1–TH2 profile. RBD–NP-CpG-alum showed a slightly higher TH1-type response, whereas RBD–NP-alum showed a higher TH2-type response. We further extended our analyses to measure IL-21 and CD154, markers of circulating T follicular helper (TFH)-like cells for their critical role in germinal centre formation and generation of durable B cell responses. We observed detectable IL-21 responses in the RBD–NP-AS03 and RBD–NP-CpG-alum groups (Fig. 2c). All cells secreting IL-21 were CD154+, and the IL-21+CD154+ double-positive cells were significantly higher in the RBD–NP-AS03 and RBD–NP-CpG-alum groups compared with the RBD–NP-AS37 group (Fig. 2d).

Fig. 2: Cell-mediated immune responses to RBD–NP immunization.
figure 2

a, b, RBD-specific CD4 T cell responses measured in blood at indicated time points. CD4 T cells secreting IL-2, IFNγ, or TNF are plotted as TH1-type responses (a) and IL-4-producing CD4 T cells are shown as TH2-type responses (b). c, Flow cytometry plots showing expression of IL-21 and CD154 after ex vivo stimulation with DMSO (no peptide, top) or an overlapping peptide pool spanning the SARS-CoV-2 RBD (bottom). d, RBD-specific CD154+ and CD154+IL-21+ CD4 T cell responses measured in blood at day 28 after inoculation. Boxes show median, 25th and 75th percentiles and whiskers show the range. a, b, Two-sided Wilcoxon matched-pairs signed-rank test. d, Two-sided Mann–Whitney rank-sum test. *P < 0.05, **P < 0.01. Each symbol represents one monkey. n = 4 (O/W), 10 (AS03) and 5 (all other groups).

Source data

We also stimulated PBMCs with a peptide pool spanning the nanoparticle component sequences I53-50A and I53-50B to determine whether RBD–NP immunization induces T cells targeting the nanoparticle scaffold. We observed a significant proportion of CD4 T cells targeting the I53-50 subunits with a response pattern similar to that of the RBD-specific T cells (Extended Data Fig. 3c). The frequencies of nanoparticle-specific CD4 T cells were about threefold higher than those of RBD-specific CD4 T cells (Extended Data Fig. 3d), an observation consistent with the RBD making up approximately one-third of the total peptidic mass of the immunogen.

Protection against SARS-CoV-2 challenge

To assess the protective efficacy, we challenged the monkeys four weeks after secondary immunization with SARS-CoV-2 via the intratracheal and intranasal routes. Two days after challenge, 4 out of 4 control monkeys had a subgenomic viral load (3.1 × 105 to 3.5 × 108 copies of the virus) in the pharyngeal and the nasal compartments. By day 7, the viral load had decreased to baseline levels, consistent with previous studies16,17. All adjuvanted groups, except RBD–NP-O/W, afforded protection from infection (Fig. 3a, Extended Data Fig. 4a). None of the five monkeys in the RBD–NP-AS03 group had detectable viral RNA in pharyngeal swabs at any time and one monkey had a viral load in nasal swabs, at a level approximately 1,000-fold lower than the median in control monkeys (2.2 × 104 versus 2.5 × 107 viral copies). By contrast, viral RNA was detectable in pharyngeal and nasal swabs of all monkeys in the RBD–NP-O/W group, albeit at lower levels than the control group. Only one out of five monkeys in the RBD–NP-CpG-alum group had detectable viral RNA in pharyngeal or nasal swabs. The RBD–NP-AS37 group and the RBD–NP-alum group also showed undetectable viral RNA in both compartments in 3 of the 5 monkeys. We measured the subgenomic viral RNA in bronchoalveolar lavage (BAL) fluid to assess protection in the lung. We used a more sensitive PCR assay to measure the product of the N gene18, as we found only two control monkeys with a positive viral load in the BAL fluid using E subgenomic RNA (sgRNA). Two days after the challenge, all 4 of the 4 control monkeys showed a viral load in the range of 104–106 viral copies. By contrast, none of the monkeys in the vaccinated groups (except one in the RBD–NP-O/W group) showed any detectable virus, suggesting effective protection in the lower respiratory tracts of all vaccinated groups, including the RBD–NP-O/W group. There were no signs of clinical disease in any monkeys, whether or not they had been vaccinated (Extended Data Fig. 4b); however, the control monkeys—but not the vaccinated monkeys—responded with an increase in neutralizing-antibody titres (Extended Data Fig. 4c), consistent with reports that SARS-CoV-2 infection of rhesus macaques results in mild disease19,20.

Fig. 3: Protection against SARS-CoV-2 challenge.
figure 3

a, b, SARS-CoV-2 viral load in nares (a) and BAL fluid (b) measured using subgenomic PCR. The numbers shown on the plots denote number of infected monkeys/total number of monkeys within each group. Statistically significant differences compared with the no-vaccine control group by two-sided Mann–Whitney rank-sum test. *P = 0.01, **P = 0.008. Each symbol represents a monkey. n = 4 for control and O/W, 5 for all other groups.

Source data

Vaccine-associated enhanced respiratory disease (VAERD) has previously been described for respiratory infections with respiratory syncytial virus and SARS-CoV21,22. We evaluated inflammation in the lung tissues of a subset of the tested monkeys using positron emission tomography with computed tomography (PET–CT) on the day of the challenge and four to five days after challenge. Of the 6 monkeys we evaluated (2 unvaccinated, 2 from the AS03 group and 2 from the CpG-alum group, selected randomly), we found inflammation in both control monkeys on day 4 compared with baseline, as measured by enhanced 2-deoxy-2-[18F]fluoroglucose (FDG) uptake. By contrast, only one of the four vaccinated monkeys showed FDG uptake, to a much lower extent than the control monkeys (Extended Data Fig. 4d, e). Additionally, we performed a comprehensive analysis of cytokine responses in all the monkeys one week after challenge, and observed no increase in inflammation in the lungs of the vaccinated monkeys (Extended Data Fig. 5a, b), whereas there were increases of cytokines such as IL-6, IL-8, IFNγ and MCP4 in the lungs of control monkeys (Extended Data Fig. 5c). These data are consistent with an absence of VAERD in these monkeys. However, these results should be interpreted with caution, as VAERD is expected to occur as immunity decreases after immunization.

Immune correlates of protection

Next, we correlated humoral and cellular immune responses measured at peak time points (day 42 for antibody responses and day 28 for T cell responses) with the viral load (nasal or pharyngeal) to determine the putative correlates of protection in an unbiased approach. Neutralizing titres for both live virus and pseudovirus emerged as the top statistically significant correlates of protection (Fig. 4, Extended Data Fig. 6a, b) in both nasal and pharyngeal compartments. Of note, the nanoparticle-specific IL-2+TNF+CD4 T cell response also emerged as a significant correlate of protection in both compartments (Extended Data Fig. 6a–c), the frequencies of which positively correlated with neutralizing-antibody titres (Extended Data Fig. 6d). This is consistent with the possibility that nanoparticle-specific CD4 T cells could offer T cell help to RBD-specific B cells.

Fig. 4: Immune correlates of protection.
figure 4

Spearman’s correlation between peak nasal viral load and neutralizing and binding-antibody responses, the top three parameters from the correlates analysis shown in Extended Data Fig. 6a. Shaded areas represent 95% confidence intervals. Each symbol represents a monkey. n = 4 (control and O/W) and 5 (all other groups).

Source data

In addition to characterizing neutralizing-antibody and T cell responses to vaccination, we sought to understand the humoral functional profile elicited by each adjuvant. Vaccines rapidly induced an increase in different anti-spike antibody isotypes (Extended Data Fig. 7a–c), Fc-receptor-binding (Extended Data Fig. 7d) and antibody-dependent neutrophil phagocytosis (ADNP) (Extended Data Fig. 7e) at day 21 and day 42. To understand how differences in the humoral response could lead to viral breakthrough, we performed a partial least-squares discriminant analysis (PLSDA) on the antibody features measured at day 42, using least absolute shrinkage and selection operator (LASSO) to select features to prevent overfitting (Extended Data Fig. 7f). The PLSDA analysis showed separation between protected and infected monkeys (Extended Data Fig. 7f), marked by an enrichment in IgA, FcR3A and ADNP in the protected monkeys (Extended Data Fig. 7g). Next, we determined the correlation of each measured antibody feature and the peak nasal and pharyngeal viral load to further dissect the antibody features that provide protection against viral break-through. Although the neutralizing-antibody response still represented the strongest correlate of protection, we observed additional functional features, including FcR binding and ADNP, that were negatively correlated with nasal or pharyngeal viral loads (Extended Data Fig. 7h). These data demonstrated an additive role for functional antibody responses in protection. Furthermore, each adjuvant group mounted a distinct profile of antibody response that correlated with protection against the virus (Extended Data Fig. 7i).

RBD–NP versus prefusion spike HexaPro

The data described thus far demonstrate that RBD–NP immunogen when adjuvanted with AS03, AS37, CpG-alum or alum induces robust protective immunity. Next, we compared the immunogenicity of the RBD–NP immunogen to that of HexaPro, a highly stable variant of the prefusion spike trimer23, in either soluble form or attached to a nanoparticle (HexaPro–NP, with 20 HexaPro trimers displayed on the I53-50 nanoparticle). We immunized three cohorts of rhesus macaques with RBD–NP, soluble HexaPro or HexaPro–NP, each with AS03 as adjuvant (Extended Data Fig. 8a, Extended Data Table 1). The RBD–NP-AS03 immunization induced neutralizing-antibody titres consistent with those described in Fig. 1, with a detectable titre on day 21 and a robust increase on day 42. Compared with RBD–NP immunization, immunization with soluble HexaPro or HexaPro–NP induced notably higher neutralizing-antibody titres after a single immunization. However, the neutralizing-antibody titres on day 42 were similar in all three groups (Fig. 5a, Extended Data Fig. 8b). Furthermore, HexaPro-AS03 immunization also elicited cross-reactive neutralizing responses against the SARS-CoV-2 B.1.1.7 and B.1.351 variants (Fig. 5b), similar to the response to RBD–NP immunization (Fig. 1d). Together, these data indicate that the RBD–NP is as potent an immunogen as the highly stable HexaPro, consistent with previous observations that the vast majority of the neutralizing response targets the RBD12. Moreover, these data also suggest that AS03 may be a suitable adjuvant for clinical use with various forms of the spike protein.

Fig. 5: RBD–NP or HexaPro immunization with AS03 elicit similar neutralizing-antibody responses.
figure 5

a, Neutralizing-antibody responses in serum on day 21 and day 42 against SARS-CoV-2 virus. The box plot shows median, 25th and 75th percentiles and the whiskers show the range. Numbers indicate GMT. Statistically significant differences between two groups by two-sided Mann–Whitney rank-sum test. *P = 0.02, **P = 0.006. Open circles denote monkeys from the earlier study shown in Fig. 1. b, Neutralizing-antibody titres measured against live wild-type (circles) or B.1.1.7 or B.1.351 variants (squares) in sera collected on day 42 from monkeys that received soluble HexaPro. Each symbol represents one monkey. n = 13 for RBD–NP and 6 for soluble HexaPro and HexaPro–NP groups.

Source data


Despite the deployment of several million doses of vaccines worldwide8,24,25,26, only a small fraction of the global population has been vaccinated to date. There remains a stark gap between the vaccination rates in different countries, with many developing countries yet to report a single dose. Furthermore, specific subpopulations such as infants and the elderly could benefit from the use of adjuvanted subunit vaccine platforms with a demonstrable history of safety and efficacy in such populations27,28. Here, we have evaluated five adjuvants, and all five induced substantial neutralizing-antibody titres. O/W induced relatively lower neutralizing titres than the other adjuvants; although O/W is an oil-in-water emulsion similar to AS03, it does not contain α-tocopherol, which is thought to be required for the potent antibody responses observed with AS0329. Notably, there were also potent CD4 T cell responses specific to the nanoparticle scaffold.

The different adjuvants produced varying levels of protection against SARS-CoV-2. The neutralizing response was the primary correlate of protection, as previously seen10,16. However, the frequency of nanoparticle-specific IL-2+TNF+ cells were also correlated with protection and neutralizing-antibody titres (Extended Data Fig. 6c, d). Thus, these nanoparticle-scaffold-specific CD4 T cells may provide T cell help to RBD-specific B cells, thereby promoting B cell responses30. In addition, the T cells could provide a complementary mechanism of protection that synergizes with the neutralizing-antibody response31. The adjuvants also induced different TH1–TH2 profiles, with RBD–NP-AS03 stimulating a mixed TH1–TH2 response, whereas RBD–NP-CpG-alum and RBD–NP-AS37 stimulated a TH1-biased response and RBD–NP-alum and RBD–NP-O/W induced TH2-biased responses. We saw no evidence of VAERD in the challenged monkeys.

In addition to evaluating clinically relevant adjuvants, we also compared the immunogenicity of RBD and HexaPro. Our results demonstrate that the RBD–NP immunogen was as potent as HexaPro in inducing neutralizing-antibody titres. Whether differences in immunogenicity become apparent at lower doses of antigen warrants further investigation. Nevertheless, these results demonstrate that AS03 is a potent adjuvant when used with either RBD–NP or HexaPro. It will be of particular interest to the field to evaluate whether the neutralizing responses elicited by RBD–NP or HexaPro-based immunogens induce a broad response not only against the emerging SARS-CoV-2 variants, but also against other coronaviruses.

In summary, this study represents a comprehensive immunological benchmarking of clinically relevant adjuvants for their capacity to enhance the protective immunity of a SARS-CoV-2 vaccine. On the basis of these results, two phase I/II clinical trials (NCT04742738 and NCT04750343) have been initiated by SK Bioscience in collaboration with Coalition for Epidemic Preparedness Innovations (CEPI) for the development of COVID-19 vaccines.


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

Animal subjects and experimentation

Thirty-three male rhesus macaques (Macaca mulatta) of Indian origin, aged 3–9 years, were assigned to the study (Extended Data Table 1). The animals were distributed between the groups such that the age and weight distribution were comparable across the groups. Animals were housed and maintained as per National Institutes of Health (NIH) guidelines at the New Iberia Research Center (NIRC) of the University of Louisiana at Lafayette in accordance with the rules and regulations of the Committee on the Care and Use of Laboratory Animal Resources. The entire study (protocol 2020-8808-15) was reviewed and approved by the University of Louisiana at Lafayette Institutional Animal Care and Use Committee (IACUC). All animals were negative for simian immunodeficiency virus, simian T cell leukaemia virus and simian retrovirus. For the challenge, the animals were transferred to the Regional Biosafety Level 3 facility at the Tulane National Primate Research Center, where the study was reviewed and approved by the Tulane University IACUC (protocol 3918).

RBD-16GS-I53-50 nanoparticle immunogen production

Nanoparticle immunogen components and nanoparticles were produced as previously described in detail5, with the exception that the nanoparticle was in a buffer containing 50 mM Tris pH 8, 150 mM NaCl, 100 mM l-arginine and 5% sucrose.

Nanoparticle biochemical characterization

Dynamic light scattering, negative stain electron microscopy, and maACE2-Fc and CR3022 IgG biolayer interferometry were performed as described previously5.

Adjuvant formulations and immunization

Essai O/W 1849101, a squalene-in-water emulsion (O/W) was provided by Seppic. For each dose, RBD–NP was diluted to 50 μg ml−1 (RBD component) in 250 μl of phosphate buffered saline (PBS) with 150 mM NaCl and mixed with an equal volume of O/W. The dose of O/W was 50% (v/v) AS03 and AS37 were provided by GSK Vaccines. AS03 is an oil-in-water emulsion that contains 11.86 mg α-tocopherol, 10.69 mg squalene, and 4.86 mg polysorbate 80 (Tween-80) in PBS, whereas AS37 is a TLR-7 agonist (200 μg ml−1) adsorbed to aluminium hydroxide (2 mg ml−1). For each dose, RBD–NP was diluted to 50 μg ml−1 (RBD component) in 250 μl of Tris-buffered saline (TBS) and mixed with an equal volume of AS03 or AS37. The dose of AS03 was 50% (v/v) (equivalent of one human dose), AS37 included 50 μg TLR-7 agonist and 0.5 mg aluminium hydroxide. CpG 1018 was provided by Dynavax Technologies at a concentration of 12 mg ml−1. Alum (Alhydrogel 2%) was purchased from Croda Healthcare (batch 0001610348). Of note, we used CpG-alum rather than CpG 1018 (no alum), which is used in Heplisav-B. For each dose of CpG-alum, 25 μg antigen (RBD component) in TBS was mixed with 0.75 mg alum and incubated on ice for 30 min. After 30 min of incubation, 1.5 mg of CpG 1018 was added and mixed rapidly. Each dose contained 1.5 mg CpG 1018 and 0.75 mg alum. For each dose of alum, 25 μg antigen (RBD component) in TBS was mixed with 0.75 mg alum, matching the concentration of alum in the CpG-alum formulation, and incubated on ice for 30 min. Soluble HexaPro or HexaPro–NP used in the experiment to compare RBD–NP versus HexapPro (described in Fig. 5) was diluted to 50 μg ml−1 in 250 μl TBS and mixed with an equal volume of AS03. All immunizations were administered via the intramuscular route in right forelimbs. The volume of each dose was 0.5 ml.

Anti-spike binding ELISA

SARS-CoV-2 spike protein was produced in HEK 293T cells (Atum). Ninety-six-well Corning Costar high-binding plates (Thermo Fisher Scientific) were coated with SARS-CoV-2 spike protein in PBS at a concentration of 0.2 μg per well overnight at 4 °C. On the next day, wells were washed 3 times with PBS, 0.1% Tween 20 (PBS-T) and blocked with PBS-T containing 3% non-fat milk powder for 1 h at room temperature. Wells were then incubated with plasma samples from non-human primates at different dilutions starting at 1:100 in PBS-T containing 1% non-fat milk for 1 h at 37 °C. After washing 3 times with PBS-T, horseradish peroxidase (HRP)-conjugated goat anti-monkey IgG (γ-chain specific, Alpha Diagnostics, 1:4,000 dilution), in PBS-T containing 1% non-fat milk was added and incubated for 1 h at room temperature. Wells were washed 3 times with PBS-T before addition of 3,3′,5,5′-tetramethylbenzidine (TMB) substrate solution. The reaction was stopped after 12 min by addition of 0.16 M sulfuric acid. The optical density (OD) at 450 nanometers was measured with a Tecan Infinite M Nano Plus microplate reader.

Anti-I53-50 ELISA

The protocol was adapted from Tiller et al.32. In brief, recombinant I53-50 protein nanoparticles, SARS-CoV-2 S2P trimers, or goat anti-human IgG (Jackson ImmunoResearch, 109-005-044) were immobilized on 96-well Nunc MaxiSorp (Thermo Fisher Scientific) plates (2 μg ml−1, 50 μl per well). After 1 h incubation at room temperature, plates were blocked with 200 μl TBS, 2% (w/v) BSA and 0.05% (v/v) Tween 20 for 1 h. Plates were washed 3 times in TBST with a plate washer (BioTek), and 50 μl of 1:5 serial dilutions starting at 1:100 of non-human primate sera in TBST incubated for 1 h in wells with I53-50 or spike protein. In wells with anti-human IgG capture antibody, human IgG control (SinoBiological, HG1K) was serially diluted from 0.5–500 ng ml−1 in TBST in triplicate and 50 μl of each dilution incubated for 1 h. Plates were washed 3 times in TBST, then HRP-conjugated goat anti-monkey IgG (Alpha Diagnostics, 70021) was diluted 1:5,000 in 2% BSA in TBST and 50 μl was incubated in each well for 30 min. Plates were washed 3 times in TBST and 100 μl TMB (SeraCare) was added to each well for 2 min. The reaction was quenched by adding 100 μl of 1 N HCl. Plates were immediately read at 450 nm on a SpectraMax M5 plate reader (Molecular Devices). The data were plotted with GraphPad Prism. A sigmoidal 4PL curve was fit with x being log (concentration) to determine the EC50 values. A logarithmic equation fit to the linear portion of the sigmoidal curve of the human IgG control was used to calculate concentration (in mg ml−1) of IgG in sera for anti-I53-50 and anti-spike titres. All steps were performed at ambient temperature.

Pseudovirus production and neutralization assay

Pseudovirus production has been described in Walls et al.6. In brief, MLV-based SARS-CoV-2 spike pseudotyped viruses were prepared as previously6,33,34 except that the SARS-CoV-2 spike construct contained the D614G mutation and a truncation of the C-terminal 21 residues11,35.

For neutralization assays, HEK-hACE2 cells were cultured in DMEM with 10% FBS (Hyclone) and 1% penicillin-streptomycin with 8% CO2 in a 37 °C incubator on 96 well plates coated with poly-l-lysine (Sigma). To coat plates, 40 μl poly-l-lysine (Sigma) was incubated with rotation for 5 min. Poly-l-lysine was removed, plates were dried for 5 min then washed once with water before plating cells. The following day, cells were checked to be at 80% confluence. In a half-area 96-well plate a 1:3 serial dilution of sera was made in DMEM in 22 μl final volume. Twenty-two microlitres of pseudovirus was then added to the serial dilution and incubated at room temperature for 30–60 min at room temperature. The medium on the HEK-hACE2 plate was removed and 40 μl of the sera–virus mixture was added to the cells and incubated for 2 h at 37 °C with 8% CO2. Following incubation, 40 μl 20% FBS and 2% PenStrep containing DMEM was added to the cells. Following 48–72 h infection, One-Glo-EX (Promega) was added to the cells in half culturing volume (40 μl added) and incubated in the dark for 5 min before reading on a Varioskan LUX plate reader (Thermo Fisher Scientific). Measurements were performed on all sera samples from each group in at least duplicates. Relative luciferase units were plotted and normalized in Prism (GraphPad) using a zero value of cells alone and a 100% value of 1:2 virus alone. Nonlinear regression of log(inhibitor) versus normalized response was used to determine IC50 values from curve fits. The human convalescent samples assayed in parallel were obtained from individuals aged 37–67 years, all of whom had mild-to-moderate disease with fever, cough, chills, shivering, runny nose, muscle aches, trouble breathing and fatigue as symptoms. The use of samples was approved by the University of Washington Human Subjects Division Institutional Review Board (IRB00009810).

FRNT assay

Neutralization assays with authentic SARS-CoV-2 virus were performed as previously described36. Plasma or serum were serially diluted (threefold) in serum-free Dulbecco’s modified Eagle’s medium (DMEM) in duplicate wells and incubated with 100–200 focus forming units (FFU) infectious clone-derived SARS-CoV-2-mNG virus37 at 37 °C for 1 h. The antibody–virus mixture was added to VeroE6 cell (C1008, ATCC, CRL-1586) monolayers seeded in 96-well blackout plates and incubated at 37 °C for 1 h. After incubation, the inoculum was removed and replaced with pre-warmed complete DMEM containing 0.85% methylcellulose. Plates were incubated at 37 °C for 24 h. After 24 h, methylcellulose overlay was removed, cells were washed twice with PBS and fixed with 2% paraformaldehyde in PBS for 30 min at room temperature. Following fixation, plates were washed twice with PBS and foci were visualized on a fluorescence ELISPOT reader (CTL ImmunoSpot S6 Universal Analyzer) and counted using Viridot38. The neutralization titres were calculated as follows: 1 − (mean number of foci in the presence of sera/number of foci at the highest dilution of the respective serum sample). Each specimen was tested in two independent assays performed at different times. The FRNT-mNG50 titres were interpolated using a 4-parameter nonlinear regression in GraphPad Prism 8.4.3. Samples with an FRNT-mNG50 value that was below the limit of detection were plotted at 10. For these samples, this value was used in fold-reduction calculations.

ACE2 blocking assay

Antibodies blocking the binding of SARS-CoV-2 spike RBD to ACE2 were detected with a V-PLEX SARS-CoV-2 Panel 2 (ACE2) Kit (Meso Scale Diagnostics) according to the manufacturer’s instructions. Serum samples from non-human primates were analysed in duplicate at a dilution of 1:100 and per cent inhibition was calculated as (1 –average sample ECL signal/average ECL signal of calibrator 7) × 100.

Pseudovirus neutralization assay against UK B.1.1.7 variant

A neutralization assay comparing the ability of sera from vaccinated animals to neutralize wild-type (with D614G in spike) SARS-CoV-2 versus the B.1.1.7 variant were performed using a pseudotyped virus-neutralization assay as previously reported39, with minor modifications. In brief, mutations were introduced into a plasmid expressing codon-optimized spike of the Wuhan-1 strain that contains the D614G mutation using site-directed mutagenesis. Pseudovirions were produced in HEK 293T/17 cells by co-transfection of a lentivirus backbone plasmid, a spike-expressing plasmid, and a firefly luciferase reporter gene plasmid. Pseudotyped viruses were titrated in 293T/ACE2.MF cells for median tissue culture infectious dose (TCID50) and used for neutralization assay. Viruses were incubated with serial diluted serum samples at 37 °C for 1 h, and subsequently added to cells and incubated for 66–72 h. Luminescence was measured using a GloMax Navigator luminometer (Promega). Neutralization titres are the inhibitory dilution (ID) of serum samples at which relative luminescence unit (RLU) readings were reduced by either 50% (ID50) or 80% (ID80) compared to virus control wells after subtraction of background RLUs.

FRNT assay against the variants of concern

The wild-type infectious clone SARS-CoV-2 (icSARS-CoV-2), derived from the 2019-nCoV/USA_WA1/2020 strain, was propagated in VeroE6 cells (ATCC) and sequenced37. The B.1.1.7 variant (SARS-CoV-2/human/USA/CA_CDC_5574/2020) was isolated from a residual nasopharyngeal swab collected from a patient in San Diego, CA, propagated in Vero cells and sequenced. The RSA B.1.351 variant was isolated as previously described40. Our laboratory plaque-isolated the virus on VeroE6 cells followed by a single round of propagation on VeroE6 cells (multiplicity of infection 0.05), aliquoted to generate a working stock and sequenced. Viral titres were determined by focus-forming assay on VeroE6 cells. Viral stocks were stored at −80 °C until use.

FRNT assays were performed as previously described for the wild-type FRNT assay. The assay with each variant was performed simultaneously with wild-type controls. The samples were diluted 3-fold in 8 serial dilutions using DMEM in duplicates with an initial dilution of 1:10 in a total volume of 60 μl. Serially diluted samples were incubated with an equal volume of wild-type or variant SARS-CoV-2 (100–200 foci per well) at 37 °C for 1 h in a round-bottomed 96-well culture plate. The antibody–virus mixture was then added to Vero cells and incubated at 37 °C for 1 h. After incubation, the antibody–virus mixture was removed and 100 μl of prewarmed 0.85% overlay was added to each well. Plates were incubated at 37 °C for 24 h. After 24 h, methylcellulose overlay was removed, and cells were washed 3 times with PBS. Cells were then fixed with 2% paraformaldehyde in PBS (Electron Microscopy Sciences) for 30 min. Following fixation, plates were washed twice with PBS and 100 μl of permeabilization buffer (0.1% BSA, saponin in PBS), was added to the fixed Vero cells for 20 min. Cells were incubated with an anti-SARS-CoV spike primary antibody directly conjugated to biotin (CR3022-biotin) for 1 h at room temperature. Next, the cells were washed three times in PBS and avidin–HRP was added for 1 h at room temperature followed by three washes in PBS. Foci were visualized using TrueBlue HRP substrate (KPL, 5510-0050) and imaged on an ELISPOT reader (CTL).

Intracellular cytokine staining assay

Antigen-specific T cell responses were measured using the intracellular cytokine staining assay. Live frozen PBMCs were revived, counted and resuspended at a density of 106 live cells per ml in complete RPMI (RPMI supplemented with 10% FBS and antibiotics). The cells were rested overnight at 37 °C in a CO2 incubator. Next morning, the cells were counted again, resuspended at a density of 15 × 106 per ml in complete RPMI and 100 μl of cell suspension containing 1.5 × 106 cells was added to each well of a 96-well round-bottomed tissue culture plate. Each sample was treated with three conditions, no stimulation, a peptide pool spanning the RBD region of spike at a concentration of 1.2 μg ml−1 of each peptide and a peptide pool spanning the I53-50A, and I53-50B components of the nanoparticle scaffold (1.2 μg ml−1 of each peptide) in the presence of 1 μg ml−1 of anti-CD28 (clone CD28.2, BD Biosciences) and anti-CD49d (clone 9F10, BD Biosciences) as well as anti-CXCR3 and anti-CXCR5. The peptides were custom synthesized to 90% purity using GenScript, a commercial vendor. All samples contained 0.5% (v/v) DMSO in total volume of 200 μl per well. The samples were incubated at 37 °C in CO2 incubators for 2 h before addition of 10 μg ml−1 brefeldin A. The cells were incubated for an additional 4 h. The cells were washed with PBS and stained with Zombie UV fixable viability dye (Biolegend). The cells were washed with PBS containing 5% FCS, before the addition of surface antibody cocktail. The cells were stained for 20 min at 4 °C in 100 μl volume. Subsequently, the cells were washed, fixed and permeabilized with cytofix/cytoperm buffer (BD Biosciences) for 20 min. The permeabilized cells were stained with intracellular cytokine staining antibodies for 20 min at room temperature in 1× perm/wash buffer (BD Biosciences). Cells were then washed twice with perm/wash buffer and once with staining buffer before acquisition using the BD Symphony Flow Cytometer and the associated BD FACS Diva software. All flow cytometry data were analysed using Flowjo software v10 (TreeStar Inc.).

Viral challenge

Animals were inoculated via the intratracheal and intranasal routes with a total of 3.2 × 106 plaque-forming units of SARS-CoV-2, isolate USA WA1/2020 (accession: MN985325). The virus stock was generated by expansion of a seed stock on Vero E6 cells and titred by plaque assay on Vero E6 cells. It was deep sequenced and found to contain no polymorphisms at greater than 5% of reads relative to the original patient isolate. The furin cleavage site, a site with frequent culture adaptation in Vero E6 cells, harboured no polymorphisms at greater than 1% of sequence reads in this stock.

Sampling of nares and pharynges

The monkeys were anaesthetized and placed in dorsal recumbency or a chair designed to maintain an upright posture. The pharynx was visualized using a laryngoscope. A sterile swab was gently rubbed and rolled across the lateral surfaces of the pharynx for approximately 5 s, including the tonsillar fossa and posterior pharynx. Care was taken to avoid touching the soft palate, uvula, buccal mucosa, tongue or lips. After all pertinent surfaces had been sampled, the swab was removed and placed into either culture medium or an appropriate container for transport. The pharyngeal swabs were done before the nasal swabs to reduce blood contamination from the nasal cavity into the pharyngeal area.

Sterile swabs were gently inserted into the nares. Once inserted, the sponge or swab was rotated several times within the cavity or region and immediately withdrawn.

BAL collection and processing

The animals were anaesthetized using Telazol and placed in a chair designed specifically for the proper positioning for BAL procedures. A local anaesthetic (2% lidocaine) may be applied to the larynx at the discretion of the veterinarian. A laryngoscope was used to visualize the epiglottis and larynx. A feeding tube was carefully introduced into the trachea after which the stylet was removed. The tube was advanced further into the trachea until slight resistance was encountered. The tube was slightly retracted and the syringe attached. Aliquots of warmed normal saline were instilled into the bronchus. The saline was aspirated between each lavage before a new aliquot was instilled. When the procedure was complete, the monkey was placed in right lateral recumbency. The monkey was carefully monitored, with observation of the heart rate, respiratory rate and effort, and mucous membrane colour. An oxygen facemask may be used following the procedure at the discretion of the veterinarian. The monkey was returned to its cage, positioned on the cage floor in right lateral recumbency and was monitored closely until recovery is complete.

The BAL samples were filtered twice via 100-μl strainers and collected in 50-ml centrifuge tubes. The samples were centrifuged at 300g for 10 min at 4 °C. The supernatant was transferred into new tubes, aliquoted and stored at −80 °C until RNA isolation. The cells were washed, lysed for red-blood cells using ammonium-chloride-potassium (ACK) lysis buffer and live-frozen in 90% FBS with 10% DMSO.

Viral load

Quantitative PCR with reverse transcription (RT–qPCR) was performed as described previously41. RT–qPCR for the subgenomic (sg) RNA encoding the envelope (E) protein was performed as described42 and RT–qPCR for the sgRNA encoding the nucleocapsid (N) protein was performed using the same cycling conditions as used for the E sgRNA using an unpublished assay provided by D. Hartigan-O’Connor and J. Dutra (U. California-Davis). Primers and probes for the N sqRNA qRT–PCR were as follows: forward 5′-CGATCTCTTGTAGATCTGTTCTC-3′, reverse 5′-GGTGAACCAAGACGCAGTAT-3′, probe 5′-FAM-TAACCAGAATGGAGAACGCAGTGGG-BHQ1-3′. Both PCRs were run in a 20 μl volume containing 5 μl sample, 900 nM primers, 250 nM probe with TaqPath 1-step RT-qPCR master mix, CG (Thermo Fisher Scientific). The PCR conditions were 2 min at 25 °C for uracil N-glycosylase incubation, 15 min at 50 °C for reverse transcription, 2 min at 95 °C (Taq activation), followed by 40 cycles of 95 °C for 3 s (denaturation) and 60 °C for 30 s (annealing and elongation).

PET–CT administration, acquisition and data collection

The animals were anaesthetized and brought to the PET–CT suite where they were monitored and prepared for imaging. An intravenous catheter is placed and the animals were intubated and placed on a gas anaesthetic (isoflurane). FDG was administered as an intravenous bolus at a dose of 0.5 mCi kg−1 in the animal preparatory room. The catheter was flushed, and the animals were transferred to the PET–CT imaging room. Images were acquired on a Mediso LFER 150 PET–CT (Mediso Medical Imaging Systems). The animals were then placed on the table in a ‘head-in-supine’ position with heat support. Scout CT images of side and top views were obtained for positioning purposes and preferred scanning ranges. The number of fields of view (FOV) was determined depending on the size of the animal (each FOV covers 15 cm and takes 10 min to obtain with PET). A CT scan was captured at 80 kVp and 1 mA with a time range of 1–5 min depending on the FOV. Breath holds were performed during the CT scan on animals that can be imaged in one FOV. A breath hold lasts for the majority of the CT scan which is approximately 45–60 s. PET images were obtained following FDG uptake time (45–60 min) and the CT scan. Once the images were captured, the animal’s fluids were discontinued and the animal was removed from isoflurane. When swallowing reflexes returned, the animal was extubated and returned to its home cage. Images were reconstructed using Nucline software with the following parameters: Mediso Tera-Tomo 3D algorithm, 8 iterations, 9 subsets, voxel size 0.7 mm.

PET–CT data analysis

PET–CT images were analysed using OsiriX MD or 64-bit (v.11, Pixmeo). Before analysis, the PET images were Gaussian smoothed in OsiriX and smoothing was applied to raw data with a 3 × 3 matrix size and a matrix normalization value of 24. Whole lung FDG uptake was measured by first creating a whole lung region-of-interest (ROI) on the lung in the CT scan by creating a 3D growing region highlighting every voxel in the lungs between −1024 and −500 Hounsfield units. This whole lung ROI is copied and pasted to the PET scan and gaps within the ROI are filled in using a closing ROI brush tool with a structuring element radius of 4. All voxels within the lung ROI with a standard uptake value (SUV) below 1.5 are set to zero and the SUVs of the remaining voxels are summed for a total lung FDG uptake (total inflammation) value. Total FDG uptake values were normalized to back muscle FDG uptake that was measured by drawing cylinder ROIs on the back muscles adjacent to the spine at the same axial level as the carina (SUVCMR; cylinder-muscle-ratio)43. PET quantification values were organized in Microsoft Excel. 3D images were created using the 3D volume rendering tool on OsiriX MD.

Luminex isotype and FcR binding assay

To determine relative concentrations of antigen-specific antibody isotypes and Fc receptor binding activity, a Luminex isotype assay was performed as previously described44. Antigens (SARS-CoV-2 spike, RBD, S1, S2, HKU1 RBD and OC43 RBD) were covalently coupled to Luminex microplex carboxylated bead regions (Luminex Corporation) using NHS-ester linkages with Sulfo-NHS and EDC (Thermo Fisher Scientific) according to manufacturer recommendations. Immune complexes were formed by incubating antigen-coupled beads with diluted samples. Mouse-anti-rhesus antibody detectors were then added for each antibody isotype (IgG1, IgG2, IgG3, IgG4 and IgA (NIH Nonhuman Primate Reagent Resource supported by AI126683 and OD010976)). Tertiary anti-mouse-IgG detector antibodies conjugated to phycoerythrin (PE) were then added. FcR binding was quantified similarly by using recombinant non-human primate FcRs (FcγR2A-1, FcγR2A-2 and FcγR3A (Duke Protein Production Facility)) conjugated to PE as secondary detectors. Flow cytometry was performed using an iQue (Intellicyt) and an S-LAB robot (PAA), and analysis was performed on IntelliCyt ForeCyt (v 8.1).

Systems serology

To quantify antibody functionality of plasma samples, bead-based assays were used to measure antibody-dependent cellular phagocytosis (ADCP), ADNP and antibody-dependent complement deposition (ADCD), as previously described45,46,47,48. SARS-CoV-2 spike protein (HexaPro antigen (from E. Ollmann Saphire, La Jolla Institute for Immunology)) was coupled to fluorescent streptavidin beads (Thermo Fisher) and incubated with serum samples to allow antibody binding to occur. For ADCP, cultured human monocytes (THP-1 cell line) were incubated with immune complexes, during which phagocytosis occurred. For ADNP, primary PMBCs were isolated from whole blood using an ACK lysis buffer. After phagocytosis of immune complexes, neutrophils were stained with an anti-CD66b Pacific Blue detection antibody (Biolegend) before flow cytometry. For ADCD, lyophilized guinea pig complement (Cedarlane) was reconstituted according to manufacturer’s instructions and diluted in a gelatin veronal buffer with calcium and magnesium (Boston BioProducts). After antibody-dependent complement deposition occurred, C3 bound to immune complexes was detected with FITC-Conjugated Goat IgG Fraction to Guinea Pig Complement C3 (MP Biomedicals). For quantification of antibody-dependent natural killer (NK) cell activation, diluted plasma samples were incubated in Nunc MaxiSorp plates (Thermo Fisher Scientific) coated with antigen. Human NK cells were isolated the evening before using RosetteSep Human NK cell Enrichment cocktail (Stemcell Technologies) from healthy buffy coat donors and incubated overnight with human recombinant IL-15 (STEMCELL Technologies). NK cells were incubated with immune complexes, CD107a PE-Cy5 (BD), Golgi stop (BD) and brefeldin A (Sigma-Aldrich). After incubation, cells were stained using anti-CD16 APC-Cy7 (BD), anti-CD56 PE-Cy7 (BD) and anti-CD3 Pacific Blue (BD), and then fixed (Perm A, Life Tech). Intracellular staining using anti-IFNγ FITC (BD) and anti-MIP-1β PE (BD) was performed after permeabilizing the NK cells with Perm B (Thermo Fisher). Flow cytometry acquisition of all assays was performed using an iQue (IntelliCyt) and a S-LAB robot (PAA). For ADCP, phagocytosis events were gated on bead-positive cells. For ADNP, neutrophils were identified by gating on CD66b+ cells, phagocytosis was identified by gating on bead-positive cells. A phagocytosis score for ADCP and ADNP was calculated as (percentage of bead-positive cells) × (MFI of bead-positive cells) divided by 10,000. ADCD quantification was reported as MFI of FITC-anti-C3. For antibody-dependent NK activation, NK cells were identified by gating on CD3, CD16+ and CD56+ cells. Data were reported as the percentage of cells positive for CD107a, IFNγ and MIP-1β.

Statistics and data visualization

The difference between any two groups at a time point was measured using a two-tailed nonparametric Mann–Whitney unpaired rank-sum test. The difference between time points within a group was measured using a Wilcoxon matched-pairs signed-rank test. All correlations were Spearman’s correlations based on ranks. All the statistical analyses were performed using GraphPad Prism v.9.0.0 or R version 3.6.1. All the figures were made in GraphPad Prism or R and organized in Adobe Illustrator.

Reporting summary

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